dataset
stringclasses 2
values | data_source
stringclasses 1
value | prompt
listlengths 2
2
| ability
stringclasses 1
value | reward_model
dict | extra_info
dict |
|---|---|---|---|---|---|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Which title is used for the leader of the major opponent to Freemasonry? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "the Bishop of Rome, who is known as the Pope",
"style": "rule"
}
|
{
"answer": "the Bishop of Rome, who is known as the Pope",
"evidence_list": [],
"index": 311,
"need_tools_kwargs": true,
"question": "Which title is used for the leader of the major opponent to Freemasonry?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "the Bishop of Rome, who is known as the Pope"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "the Bishop of Rome, who is known as the Pope"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What's the record label of the Addicted to Music creator? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Kontor Records",
"style": "rule"
}
|
{
"answer": "Kontor Records",
"evidence_list": [],
"index": 315,
"need_tools_kwargs": true,
"question": "What's the record label of the Addicted to Music creator?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Kontor Records"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Kontor Records"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "When did the author of A Defence of Poetry die? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "8 July 1822",
"style": "rule"
}
|
{
"answer": "8 July 1822",
"evidence_list": [],
"index": 319,
"need_tools_kwargs": true,
"question": "When did the author of A Defence of Poetry die?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "8 July 1822"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "8 July 1822"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Which year did the founder of Regius Professor of Surgery die? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "1820",
"style": "rule"
}
|
{
"answer": "1820",
"evidence_list": [],
"index": 323,
"need_tools_kwargs": true,
"question": "Which year did the founder of Regius Professor of Surgery die?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "1820"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "1820"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What Chinese city shares a border with the city that hosted the 2009 ELEXCON fair? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Dongguan",
"style": "rule"
}
|
{
"answer": "Dongguan",
"evidence_list": [],
"index": 326,
"need_tools_kwargs": true,
"question": "What Chinese city shares a border with the city that hosted the 2009 ELEXCON fair?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Dongguan"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Dongguan"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who did Miley play in the episode 'Achy Jakey Heart'? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Miley Stewart",
"style": "rule"
}
|
{
"answer": "Miley Stewart",
"evidence_list": [],
"index": 329,
"need_tools_kwargs": true,
"question": "Who did Miley play in the episode 'Achy Jakey Heart'?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Miley Stewart"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Miley Stewart"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "When did the person who said that poets are the unacknowledged legislators of world die? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "8 July 1822",
"style": "rule"
}
|
{
"answer": "8 July 1822",
"evidence_list": [],
"index": 332,
"need_tools_kwargs": true,
"question": "When did the person who said that poets are the unacknowledged legislators of world die?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "8 July 1822"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "8 July 1822"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "How many episodes are there of the series that the episode I'll Always Remember You is a part of? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "98",
"style": "rule"
}
|
{
"answer": "98",
"evidence_list": [],
"index": 333,
"need_tools_kwargs": true,
"question": "How many episodes are there of the series that the episode I'll Always Remember You is a part of?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "98"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "98"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the sibling of the person who wrote the script for the 1996 Crucible movie adaption? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Joan Maxine Copeland",
"style": "rule"
}
|
{
"answer": "Joan Maxine Copeland",
"evidence_list": [],
"index": 336,
"need_tools_kwargs": true,
"question": "Who is the sibling of the person who wrote the script for the 1996 Crucible movie adaption?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Joan Maxine Copeland"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Joan Maxine Copeland"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Which record label owns the series where Achy Jakey Heart belongs? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Walt Disney Records",
"style": "rule"
}
|
{
"answer": "Walt Disney Records",
"evidence_list": [],
"index": 339,
"need_tools_kwargs": true,
"question": "Which record label owns the series where Achy Jakey Heart belongs?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Walt Disney Records"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Walt Disney Records"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "When was the company that repaired the Grand Trunk Road from Calcutta to Delhi established in India? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "31 December 1600",
"style": "rule"
}
|
{
"answer": "31 December 1600",
"evidence_list": [],
"index": 342,
"need_tools_kwargs": true,
"question": "When was the company that repaired the Grand Trunk Road from Calcutta to Delhi established in India?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "31 December 1600"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "31 December 1600"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What year did the person who described poets as the unacknowledged legislators of mankind die? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "8 July 1822",
"style": "rule"
}
|
{
"answer": "8 July 1822",
"evidence_list": [],
"index": 348,
"need_tools_kwargs": true,
"question": "What year did the person who described poets as the unacknowledged legislators of mankind die?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "8 July 1822"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "8 July 1822"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What country holds the Embassy of Turkey, to the country that Serdar Denktas is from? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "TRNC",
"style": "rule"
}
|
{
"answer": "TRNC",
"evidence_list": [],
"index": 351,
"need_tools_kwargs": true,
"question": "What country holds the Embassy of Turkey, to the country that Serdar Denktas is from?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "TRNC"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "TRNC"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "When did the original judge who was a music executive announce he was no longer going to be a judge? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "January 11, 2010",
"style": "rule"
}
|
{
"answer": "January 11, 2010",
"evidence_list": [],
"index": 352,
"need_tools_kwargs": true,
"question": "When did the original judge who was a music executive announce he was no longer going to be a judge?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "January 11, 2010"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "January 11, 2010"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is a cast member of the series that I'll Always Remember You is a part of? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Billy Ray Cyrus",
"style": "rule"
}
|
{
"answer": "Billy Ray Cyrus",
"evidence_list": [],
"index": 364,
"need_tools_kwargs": true,
"question": "Who is a cast member of the series that I'll Always Remember You is a part of?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Billy Ray Cyrus"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Billy Ray Cyrus"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "In which county is the town WPWT is licensed in? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Sullivan County",
"style": "rule"
}
|
{
"answer": "Sullivan County",
"evidence_list": [],
"index": 366,
"need_tools_kwargs": true,
"question": "In which county is the town WPWT is licensed in?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Sullivan County"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Sullivan County"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What year did the cast member of Tip Top win best actress? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "2001",
"style": "rule"
}
|
{
"answer": "2001",
"evidence_list": [],
"index": 371,
"need_tools_kwargs": true,
"question": "What year did the cast member of Tip Top win best actress?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "2001"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "2001"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "In what century did the person after who Saint-Nicolas church is named become the influence for red being worn by Santa Claus? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "4th",
"style": "rule"
}
|
{
"answer": "4th",
"evidence_list": [],
"index": 377,
"need_tools_kwargs": true,
"question": "In what century did the person after who Saint-Nicolas church is named become the influence for red being worn by Santa Claus?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "4th"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "4th"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "In what century did the person the Palmi Cathedral is named after become the influence for red being worn by Santa Claus? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "4th",
"style": "rule"
}
|
{
"answer": "4th",
"evidence_list": [],
"index": 379,
"need_tools_kwargs": true,
"question": "In what century did the person the Palmi Cathedral is named after become the influence for red being worn by Santa Claus?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "4th"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "4th"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "In what century did the person for whom Murska Sobota Cathedral was named become the influence for Santa Claus wearing red? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "4th",
"style": "rule"
}
|
{
"answer": "4th",
"evidence_list": [],
"index": 380,
"need_tools_kwargs": true,
"question": "In what century did the person for whom Murska Sobota Cathedral was named become the influence for Santa Claus wearing red?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "4th"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "4th"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "In what century did the person San Nicolò dei Mendicoli is named after become the influence for red being worn by Santa Claus? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "4th",
"style": "rule"
}
|
{
"answer": "4th",
"evidence_list": [],
"index": 383,
"need_tools_kwargs": true,
"question": "In what century did the person San Nicolò dei Mendicoli is named after become the influence for red being worn by Santa Claus?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "4th"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "4th"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What is the meaning of the name of the province where Larestan County is located? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Old Persian as Pars",
"style": "rule"
}
|
{
"answer": "Old Persian as Pars",
"evidence_list": [],
"index": 387,
"need_tools_kwargs": true,
"question": "What is the meaning of the name of the province where Larestan County is located?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Old Persian as Pars"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Old Persian as Pars"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What does the name of the state that Jam Bozorgi is located mean? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Old Persian as Pars",
"style": "rule"
}
|
{
"answer": "Old Persian as Pars",
"evidence_list": [],
"index": 391,
"need_tools_kwargs": true,
"question": "What does the name of the state that Jam Bozorgi is located mean?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Old Persian as Pars"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Old Persian as Pars"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What year was the developer of EPrints founded? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "1862",
"style": "rule"
}
|
{
"answer": "1862",
"evidence_list": [],
"index": 393,
"need_tools_kwargs": true,
"question": "What year was the developer of EPrints founded?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "1862"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "1862"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What was the political party of the person who regained control? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Union of Democrats for the Republic",
"style": "rule"
}
|
{
"answer": "Union of Democrats for the Republic",
"evidence_list": [],
"index": 397,
"need_tools_kwargs": true,
"question": "What was the political party of the person who regained control?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Union of Democrats for the Republic"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Union of Democrats for the Republic"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the spouse of the individual who appeared on the $20 bill in 1875? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Elizabeth Schuyler Hamilton",
"style": "rule"
}
|
{
"answer": "Elizabeth Schuyler Hamilton",
"evidence_list": [],
"index": 400,
"need_tools_kwargs": true,
"question": "Who is the spouse of the individual who appeared on the $20 bill in 1875?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Elizabeth Schuyler Hamilton"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Elizabeth Schuyler Hamilton"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the child of the 1958 Senate Majority Leader? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Lynda Bird Johnson Robb",
"style": "rule"
}
|
{
"answer": "Lynda Bird Johnson Robb",
"evidence_list": [],
"index": 408,
"need_tools_kwargs": true,
"question": "Who is the child of the 1958 Senate Majority Leader?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Lynda Bird Johnson Robb"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Lynda Bird Johnson Robb"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What year was the university Arthur Button attended founded? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "1862",
"style": "rule"
}
|
{
"answer": "1862",
"evidence_list": [],
"index": 412,
"need_tools_kwargs": true,
"question": "What year was the university Arthur Button attended founded?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "1862"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "1862"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What does the name of the province where Sahray-ye Bagh District is located mean? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Old Persian as Pars",
"style": "rule"
}
|
{
"answer": "Old Persian as Pars",
"evidence_list": [],
"index": 416,
"need_tools_kwargs": true,
"question": "What does the name of the province where Sahray-ye Bagh District is located mean?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Old Persian as Pars"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Old Persian as Pars"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who played the performer of When He Walks on You in Walk the Line? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Waylon Malloy Payne",
"style": "rule"
}
|
{
"answer": "Waylon Malloy Payne",
"evidence_list": [],
"index": 418,
"need_tools_kwargs": true,
"question": "Who played the performer of When He Walks on You in Walk the Line?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Waylon Malloy Payne"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Waylon Malloy Payne"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who played the writer of \"Lewis Boogie\" on walk the line? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Waylon Malloy Payne",
"style": "rule"
}
|
{
"answer": "Waylon Malloy Payne",
"evidence_list": [],
"index": 421,
"need_tools_kwargs": true,
"question": "Who played the writer of \"Lewis Boogie\" on walk the line?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Waylon Malloy Payne"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Waylon Malloy Payne"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Where in Plymouth is the owner of Torquay Lifeboat Station based? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Millbay Docks",
"style": "rule"
}
|
{
"answer": "Millbay Docks",
"evidence_list": [],
"index": 423,
"need_tools_kwargs": true,
"question": "Where in Plymouth is the owner of Torquay Lifeboat Station based?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Millbay Docks"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Millbay Docks"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who authored The Road to City A where City A is the birthplace of Ryan Junge? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Robert Ludlum",
"style": "rule"
}
|
{
"answer": "Robert Ludlum",
"evidence_list": [],
"index": 425,
"need_tools_kwargs": true,
"question": "Who authored The Road to City A where City A is the birthplace of Ryan Junge?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Robert Ludlum"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Robert Ludlum"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What is the city where Todor Kableshkov lived when he died the capital of? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Gabrovo Province",
"style": "rule"
}
|
{
"answer": "Gabrovo Province",
"evidence_list": [],
"index": 426,
"need_tools_kwargs": true,
"question": "What is the city where Todor Kableshkov lived when he died the capital of?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Gabrovo Province"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Gabrovo Province"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What was the record label of the songwriter for I've Done Everything For You? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Capitol Records",
"style": "rule"
}
|
{
"answer": "Capitol Records",
"evidence_list": [],
"index": 429,
"need_tools_kwargs": true,
"question": "What was the record label of the songwriter for I've Done Everything For You?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Capitol Records"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Capitol Records"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Where in Plymouth is the operator of Oakley class lifeboat based? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Millbay Docks",
"style": "rule"
}
|
{
"answer": "Millbay Docks",
"evidence_list": [],
"index": 430,
"need_tools_kwargs": true,
"question": "Where in Plymouth is the operator of Oakley class lifeboat based?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Millbay Docks"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Millbay Docks"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who wrote a novel with a title named after a road to the city where Abbie Cornett was born? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Robert Ludlum",
"style": "rule"
}
|
{
"answer": "Robert Ludlum",
"evidence_list": [],
"index": 435,
"need_tools_kwargs": true,
"question": "Who wrote a novel with a title named after a road to the city where Abbie Cornett was born?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Robert Ludlum"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Robert Ludlum"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "When did the performer of The Promise of a New Day quit as a judge? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "before season nine",
"style": "rule"
}
|
{
"answer": "before season nine",
"evidence_list": [],
"index": 438,
"need_tools_kwargs": true,
"question": "When did the performer of The Promise of a New Day quit as a judge?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "before season nine"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "before season nine"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the spouse of the person who suggested that the ratio of silver to gold should be fixed? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Elizabeth Schuyler Hamilton",
"style": "rule"
}
|
{
"answer": "Elizabeth Schuyler Hamilton",
"evidence_list": [],
"index": 440,
"need_tools_kwargs": true,
"question": "Who is the spouse of the person who suggested that the ratio of silver to gold should be fixed?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Elizabeth Schuyler Hamilton"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Elizabeth Schuyler Hamilton"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Where in Plymouth is the operator of Barnett class lifeboat based? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Millbay Docks",
"style": "rule"
}
|
{
"answer": "Millbay Docks",
"evidence_list": [],
"index": 441,
"need_tools_kwargs": true,
"question": "Where in Plymouth is the operator of Barnett class lifeboat based?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Millbay Docks"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Millbay Docks"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What is the meaning of the name of the administrative territorial entity that contains Sivand? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Old Persian as Pars",
"style": "rule"
}
|
{
"answer": "Old Persian as Pars",
"evidence_list": [],
"index": 443,
"need_tools_kwargs": true,
"question": "What is the meaning of the name of the administrative territorial entity that contains Sivand?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Old Persian as Pars"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Old Persian as Pars"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Where in Plymouth is the past operator of the Clyde-class lifeboats? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Millbay Docks",
"style": "rule"
}
|
{
"answer": "Millbay Docks",
"evidence_list": [],
"index": 448,
"need_tools_kwargs": true,
"question": "Where in Plymouth is the past operator of the Clyde-class lifeboats?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Millbay Docks"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Millbay Docks"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who did the hitman from The Hitman's Bodyguard play in Star Wars? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Mace Windu",
"style": "rule"
}
|
{
"answer": "Mace Windu",
"evidence_list": [],
"index": 449,
"need_tools_kwargs": true,
"question": "Who did the hitman from The Hitman's Bodyguard play in Star Wars?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Mace Windu"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Mace Windu"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Where in Plymouth is the operator of RNLB Aguila Ren based? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Millbay Docks",
"style": "rule"
}
|
{
"answer": "Millbay Docks",
"evidence_list": [],
"index": 450,
"need_tools_kwargs": true,
"question": "Where in Plymouth is the operator of RNLB Aguila Ren based?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Millbay Docks"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Millbay Docks"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What does the name of the region encompassing Kavar County mean? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Old Persian as Pars",
"style": "rule"
}
|
{
"answer": "Old Persian as Pars",
"evidence_list": [],
"index": 468,
"need_tools_kwargs": true,
"question": "What does the name of the region encompassing Kavar County mean?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Old Persian as Pars"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Old Persian as Pars"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "When did the original judge who left prior to the start of the season quit as a judge? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "before season nine",
"style": "rule"
}
|
{
"answer": "before season nine",
"evidence_list": [],
"index": 474,
"need_tools_kwargs": true,
"question": "When did the original judge who left prior to the start of the season quit as a judge?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "before season nine"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "before season nine"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What is the main tool against organized crime of the agency that pioneered DNA testing? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "RICO",
"style": "rule"
}
|
{
"answer": "RICO",
"evidence_list": [],
"index": 475,
"need_tools_kwargs": true,
"question": "What is the main tool against organized crime of the agency that pioneered DNA testing?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "RICO"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "RICO"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who paid £15 million to transfer the best scorer in premier league to their club? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Newcastle United",
"style": "rule"
}
|
{
"answer": "Newcastle United",
"evidence_list": [],
"index": 478,
"need_tools_kwargs": true,
"question": "Who paid £15 million to transfer the best scorer in premier league to their club?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Newcastle United"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Newcastle United"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Where did the husband of Alice Hoschede have his water lily pond? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "his flower garden at his home in Giverny",
"style": "rule"
}
|
{
"answer": "his flower garden at his home in Giverny",
"evidence_list": [],
"index": 481,
"need_tools_kwargs": true,
"question": "Where did the husband of Alice Hoschede have his water lily pond?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "his flower garden at his home in Giverny"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "his flower garden at his home in Giverny"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What is the location of the water lily pond by the creator of Boating on the River Epte? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "his flower garden at his home in Giverny",
"style": "rule"
}
|
{
"answer": "his flower garden at his home in Giverny",
"evidence_list": [],
"index": 484,
"need_tools_kwargs": true,
"question": "What is the location of the water lily pond by the creator of Boating on the River Epte?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "his flower garden at his home in Giverny"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "his flower garden at his home in Giverny"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "On Walk the Line, who played the performer of She Even Woke Me Up to Say Goodbye? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Waylon Malloy Payne",
"style": "rule"
}
|
{
"answer": "Waylon Malloy Payne",
"evidence_list": [],
"index": 489,
"need_tools_kwargs": true,
"question": "On Walk the Line, who played the performer of She Even Woke Me Up to Say Goodbye?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Waylon Malloy Payne"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Waylon Malloy Payne"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the spouse of the Democratic candidate who lost his own state in the 2000 general election? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Tipper Gore",
"style": "rule"
}
|
{
"answer": "Tipper Gore",
"evidence_list": [],
"index": 490,
"need_tools_kwargs": true,
"question": "Who is the spouse of the Democratic candidate who lost his own state in the 2000 general election?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Tipper Gore"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Tipper Gore"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is married to the former US senator from Tennessee who ran for president in 2000? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Tipper Gore",
"style": "rule"
}
|
{
"answer": "Tipper Gore",
"evidence_list": [],
"index": 492,
"need_tools_kwargs": true,
"question": "Who is married to the former US senator from Tennessee who ran for president in 2000?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Tipper Gore"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Tipper Gore"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the child of the United States Vice President which Donald Moffat played? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Lynda Bird Johnson Robb",
"style": "rule"
}
|
{
"answer": "Lynda Bird Johnson Robb",
"evidence_list": [],
"index": 501,
"need_tools_kwargs": true,
"question": "Who is the child of the United States Vice President which Donald Moffat played?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Lynda Bird Johnson Robb"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Lynda Bird Johnson Robb"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the mother of the last Democratic presidential candidate to win any counties in Oklahoma? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Pauline LaFon Gore",
"style": "rule"
}
|
{
"answer": "Pauline LaFon Gore",
"evidence_list": [],
"index": 502,
"need_tools_kwargs": true,
"question": "Who is the mother of the last Democratic presidential candidate to win any counties in Oklahoma?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Pauline LaFon Gore"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Pauline LaFon Gore"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "In what language did the most famous student of Bell speak? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "English language",
"style": "rule"
}
|
{
"answer": "English language",
"evidence_list": [],
"index": 505,
"need_tools_kwargs": true,
"question": "In what language did the most famous student of Bell speak?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "English language"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "English language"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who played the performer of \"One Minute Past Eternity\" on Walk the Line? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Waylon Malloy Payne",
"style": "rule"
}
|
{
"answer": "Waylon Malloy Payne",
"evidence_list": [],
"index": 506,
"need_tools_kwargs": true,
"question": "Who played the performer of \"One Minute Past Eternity\" on Walk the Line?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Waylon Malloy Payne"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Waylon Malloy Payne"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who paid £15 million to transfer the player who scored the most goal in the Premier League to their club? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Newcastle United",
"style": "rule"
}
|
{
"answer": "Newcastle United",
"evidence_list": [],
"index": 514,
"need_tools_kwargs": true,
"question": "Who paid £15 million to transfer the player who scored the most goal in the Premier League to their club?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Newcastle United"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Newcastle United"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who played the Would You Take Another Chance on Me performer in Walk the Line? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Waylon Malloy Payne",
"style": "rule"
}
|
{
"answer": "Waylon Malloy Payne",
"evidence_list": [],
"index": 516,
"need_tools_kwargs": true,
"question": "Who played the Would You Take Another Chance on Me performer in Walk the Line?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Waylon Malloy Payne"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Waylon Malloy Payne"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "The 2000 presidential candidate from Tennessee was birthed by whom? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Pauline LaFon Gore",
"style": "rule"
}
|
{
"answer": "Pauline LaFon Gore",
"evidence_list": [],
"index": 517,
"need_tools_kwargs": true,
"question": "The 2000 presidential candidate from Tennessee was birthed by whom?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Pauline LaFon Gore"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Pauline LaFon Gore"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who paid £15 million to transfer the player who has scored the most Premier League goals to their club? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Newcastle United",
"style": "rule"
}
|
{
"answer": "Newcastle United",
"evidence_list": [],
"index": 518,
"need_tools_kwargs": true,
"question": "Who paid £15 million to transfer the player who has scored the most Premier League goals to their club?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Newcastle United"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Newcastle United"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Along with the singer of Nude as the News, what indie/folk musician is based in Miami? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Iron & Wine",
"style": "rule"
}
|
{
"answer": "Iron & Wine",
"evidence_list": [],
"index": 527,
"need_tools_kwargs": true,
"question": "Along with the singer of Nude as the News, what indie/folk musician is based in Miami?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Iron & Wine"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Iron & Wine"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What county contains the town that WWOT is licensed in? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Blair County",
"style": "rule"
}
|
{
"answer": "Blair County",
"evidence_list": [],
"index": 529,
"need_tools_kwargs": true,
"question": "What county contains the town that WWOT is licensed in?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Blair County"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Blair County"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the spouse of the director of La pequeña señora de Pérez? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Susana Freyre",
"style": "rule"
}
|
{
"answer": "Susana Freyre",
"evidence_list": [],
"index": 531,
"need_tools_kwargs": true,
"question": "Who is the spouse of the director of La pequeña señora de Pérez?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Susana Freyre"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Susana Freyre"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What was the release date of the iphone 6s plus, produced by the company that lost the lawsuit vs. Microsoft and Hewlett-Packard? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "September 25, 2015",
"style": "rule"
}
|
{
"answer": "September 25, 2015",
"evidence_list": [],
"index": 532,
"need_tools_kwargs": true,
"question": "What was the release date of the iphone 6s plus, produced by the company that lost the lawsuit vs. Microsoft and Hewlett-Packard?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "September 25, 2015"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "September 25, 2015"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who paid £15 million to transfer the highest goal scorer in Premier League to their club? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Newcastle United",
"style": "rule"
}
|
{
"answer": "Newcastle United",
"evidence_list": [],
"index": 533,
"need_tools_kwargs": true,
"question": "Who paid £15 million to transfer the highest goal scorer in Premier League to their club?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Newcastle United"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Newcastle United"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who paid £15 million to transfer the player who scored the most hat tricks in the Premier League to their club? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Newcastle United",
"style": "rule"
}
|
{
"answer": "Newcastle United",
"evidence_list": [],
"index": 538,
"need_tools_kwargs": true,
"question": "Who paid £15 million to transfer the player who scored the most hat tricks in the Premier League to their club?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Newcastle United"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Newcastle United"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who was the spouse of the director of La muerte camina en la lluvia? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Susana Freyre",
"style": "rule"
}
|
{
"answer": "Susana Freyre",
"evidence_list": [],
"index": 541,
"need_tools_kwargs": true,
"question": "Who was the spouse of the director of La muerte camina en la lluvia?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Susana Freyre"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Susana Freyre"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "When did the relationship of the person a lot of modern pop artists grew up listening to with Carlos Leon end? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "May 1997",
"style": "rule"
}
|
{
"answer": "May 1997",
"evidence_list": [],
"index": 546,
"need_tools_kwargs": true,
"question": "When did the relationship of the person a lot of modern pop artists grew up listening to with Carlos Leon end?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "May 1997"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "May 1997"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who paid £15 million to transfer the highest goal scorer in the English Premier League to their club? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Newcastle United",
"style": "rule"
}
|
{
"answer": "Newcastle United",
"evidence_list": [],
"index": 548,
"need_tools_kwargs": true,
"question": "Who paid £15 million to transfer the highest goal scorer in the English Premier League to their club?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Newcastle United"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Newcastle United"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Where the water lily pond of the painter of Le Bassin Aux Nymphéas? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "his flower garden at his home in Giverny",
"style": "rule"
}
|
{
"answer": "his flower garden at his home in Giverny",
"evidence_list": [],
"index": 550,
"need_tools_kwargs": true,
"question": "Where the water lily pond of the painter of Le Bassin Aux Nymphéas?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "his flower garden at his home in Giverny"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "his flower garden at his home in Giverny"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "How did the agency which pioneered DNA testing respond to requests to release information beyond the initial press release in the case of Filiberto Ojeda Ríos? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "The FBI refused",
"style": "rule"
}
|
{
"answer": "The FBI refused",
"evidence_list": [],
"index": 553,
"need_tools_kwargs": true,
"question": "How did the agency which pioneered DNA testing respond to requests to release information beyond the initial press release in the case of Filiberto Ojeda Ríos?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "The FBI refused"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "The FBI refused"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Kante is found in what district of the country with violent police confrontations? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Darchula District",
"style": "rule"
}
|
{
"answer": "Darchula District",
"evidence_list": [],
"index": 555,
"need_tools_kwargs": true,
"question": "Kante is found in what district of the country with violent police confrontations?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Darchula District"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Darchula District"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "When they were young many modern pop stars listened to the artist described as an almost scared feminist icon by who? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Professor Sut Jhally",
"style": "rule"
}
|
{
"answer": "Professor Sut Jhally",
"evidence_list": [],
"index": 557,
"need_tools_kwargs": true,
"question": "When they were young many modern pop stars listened to the artist described as an almost scared feminist icon by who?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Professor Sut Jhally"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Professor Sut Jhally"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "How many countries saw the debut of the eleventh album of the artist who a lot of modern pop artists grew up listening to? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "37",
"style": "rule"
}
|
{
"answer": "37",
"evidence_list": [],
"index": 561,
"need_tools_kwargs": true,
"question": "How many countries saw the debut of the eleventh album of the artist who a lot of modern pop artists grew up listening to?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "37"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "37"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "what percentage of revenue for the company that lost a lawsuit against microsoft and Hewlett-Packard was from desktop computers? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "19.22%",
"style": "rule"
}
|
{
"answer": "19.22%",
"evidence_list": [],
"index": 562,
"need_tools_kwargs": true,
"question": "what percentage of revenue for the company that lost a lawsuit against microsoft and Hewlett-Packard was from desktop computers?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "19.22%"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "19.22%"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "The programming language with the HAVING function, is named after whom? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Michael Widenius's daughter",
"style": "rule"
}
|
{
"answer": "Michael Widenius's daughter",
"evidence_list": [],
"index": 564,
"need_tools_kwargs": true,
"question": "The programming language with the HAVING function, is named after whom?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Michael Widenius's daughter"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Michael Widenius's daughter"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "In what year was the performer of Dignity founded? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "1985",
"style": "rule"
}
|
{
"answer": "1985",
"evidence_list": [],
"index": 571,
"need_tools_kwargs": true,
"question": "In what year was the performer of Dignity founded?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "1985"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "1985"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "In what year did the performers of Everytime You Sleep form their band? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "1985",
"style": "rule"
}
|
{
"answer": "1985",
"evidence_list": [],
"index": 580,
"need_tools_kwargs": true,
"question": "In what year did the performers of Everytime You Sleep form their band?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "1985"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "1985"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the programming language that includes the UPDATE statement partially named after? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Michael Widenius's daughter",
"style": "rule"
}
|
{
"answer": "Michael Widenius's daughter",
"evidence_list": [],
"index": 581,
"need_tools_kwargs": true,
"question": "Who is the programming language that includes the UPDATE statement partially named after?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Michael Widenius's daughter"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Michael Widenius's daughter"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "what was Michelle Douglas's alma mater called at first? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Carleton College",
"style": "rule"
}
|
{
"answer": "Carleton College",
"evidence_list": [],
"index": 588,
"need_tools_kwargs": true,
"question": "what was Michelle Douglas's alma mater called at first?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Carleton College"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Carleton College"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who did the person who played the mother on This Is Us play in The Princess Diaries? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Lana Thomas",
"style": "rule"
}
|
{
"answer": "Lana Thomas",
"evidence_list": [],
"index": 591,
"need_tools_kwargs": true,
"question": "Who did the person who played the mother on This Is Us play in The Princess Diaries?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Lana Thomas"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Lana Thomas"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the brother of the developer of Buffy the Animated Series? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Jed Whedon",
"style": "rule"
}
|
{
"answer": "Jed Whedon",
"evidence_list": [],
"index": 592,
"need_tools_kwargs": true,
"question": "Who is the brother of the developer of Buffy the Animated Series?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Jed Whedon"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Jed Whedon"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Where did the actor who produced Reds attend school? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Northwestern's School of Communication",
"style": "rule"
}
|
{
"answer": "Northwestern's School of Communication",
"evidence_list": [],
"index": 593,
"need_tools_kwargs": true,
"question": "Where did the actor who produced Reds attend school?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Northwestern's School of Communication"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Northwestern's School of Communication"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who was played by the Nothing That You Are performer in Princess Diaries? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Lana Thomas",
"style": "rule"
}
|
{
"answer": "Lana Thomas",
"evidence_list": [],
"index": 595,
"need_tools_kwargs": true,
"question": "Who was played by the Nothing That You Are performer in Princess Diaries?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Lana Thomas"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Lana Thomas"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who sings \"meet me in Montana\" with the performer of \"All in Love\" You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Dan Seals",
"style": "rule"
}
|
{
"answer": "Dan Seals",
"evidence_list": [],
"index": 598,
"need_tools_kwargs": true,
"question": "Who sings \"meet me in Montana\" with the performer of \"All in Love\"",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Dan Seals"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Dan Seals"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "What school did the director of Reds attend? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Northwestern's School of Communication",
"style": "rule"
}
|
{
"answer": "Northwestern's School of Communication",
"evidence_list": [],
"index": 600,
"need_tools_kwargs": true,
"question": "What school did the director of Reds attend?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Northwestern's School of Communication"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Northwestern's School of Communication"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who did the singer of Candy play in Princes Diaries? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Lana Thomas",
"style": "rule"
}
|
{
"answer": "Lana Thomas",
"evidence_list": [],
"index": 602,
"need_tools_kwargs": true,
"question": "Who did the singer of Candy play in Princes Diaries?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Lana Thomas"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Lana Thomas"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who was married to the author of The Night of Enitharmon's Joy? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Catherine Blake",
"style": "rule"
}
|
{
"answer": "Catherine Blake",
"evidence_list": [],
"index": 605,
"need_tools_kwargs": true,
"question": "Who was married to the author of The Night of Enitharmon's Joy?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Catherine Blake"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Catherine Blake"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the sibling of the creator of The Disquieting Muses? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Alberto Savinio",
"style": "rule"
}
|
{
"answer": "Alberto Savinio",
"evidence_list": [],
"index": 606,
"need_tools_kwargs": true,
"question": "Who is the sibling of the creator of The Disquieting Muses?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Alberto Savinio"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Alberto Savinio"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who sings with performer of Today on Every Storm? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Co-writer Hillary Lindsey",
"style": "rule"
}
|
{
"answer": "Co-writer Hillary Lindsey",
"evidence_list": [],
"index": 608,
"need_tools_kwargs": true,
"question": "Who sings with performer of Today on Every Storm?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Co-writer Hillary Lindsey"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Co-writer Hillary Lindsey"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who sings with the performer of It Would Be You on Every Storm? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Co-writer Hillary Lindsey",
"style": "rule"
}
|
{
"answer": "Co-writer Hillary Lindsey",
"evidence_list": [],
"index": 609,
"need_tools_kwargs": true,
"question": "Who sings with the performer of It Would Be You on Every Storm?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Co-writer Hillary Lindsey"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Co-writer Hillary Lindsey"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who sang Meet Me in Montana with the performer of I Only wanted You? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Dan Seals",
"style": "rule"
}
|
{
"answer": "Dan Seals",
"evidence_list": [],
"index": 610,
"need_tools_kwargs": true,
"question": "Who sang Meet Me in Montana with the performer of I Only wanted You?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Dan Seals"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Dan Seals"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "In what year was the performer of A New House formed? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "1985",
"style": "rule"
}
|
{
"answer": "1985",
"evidence_list": [],
"index": 611,
"need_tools_kwargs": true,
"question": "In what year was the performer of A New House formed?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "1985"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "1985"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "who was the spouse of the seventh president of the U.S.? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Rachel Jackson",
"style": "rule"
}
|
{
"answer": "Rachel Jackson",
"evidence_list": [],
"index": 612,
"need_tools_kwargs": true,
"question": "who was the spouse of the seventh president of the U.S.?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Rachel Jackson"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Rachel Jackson"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who did So Real's performer play in Princess Diaries? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Lana Thomas",
"style": "rule"
}
|
{
"answer": "Lana Thomas",
"evidence_list": [],
"index": 614,
"need_tools_kwargs": true,
"question": "Who did So Real's performer play in Princess Diaries?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Lana Thomas"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Lana Thomas"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who did Cry's performer play in princess diaries? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Lana Thomas",
"style": "rule"
}
|
{
"answer": "Lana Thomas",
"evidence_list": [],
"index": 616,
"need_tools_kwargs": true,
"question": "Who did Cry's performer play in princess diaries?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Lana Thomas"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Lana Thomas"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the spouse of the cast member of La Boum? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Christopher Lambert",
"style": "rule"
}
|
{
"answer": "Christopher Lambert",
"evidence_list": [],
"index": 617,
"need_tools_kwargs": true,
"question": "Who is the spouse of the cast member of La Boum?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Christopher Lambert"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Christopher Lambert"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the king singing to in the musical named for the person on the $10 bill? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "the rebelling American colonists",
"style": "rule"
}
|
{
"answer": "the rebelling American colonists",
"evidence_list": [],
"index": 619,
"need_tools_kwargs": true,
"question": "Who is the king singing to in the musical named for the person on the $10 bill?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "the rebelling American colonists"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "the rebelling American colonists"
}
}
}
}
|
musique_train
|
musique_train
|
[
{
"content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n",
"role": "system"
},
{
"content": "Who is the spouse of Ernest the Rebel's director? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.",
"role": "user"
}
] |
qa
|
{
"ground_truth": "Martine Carol",
"style": "rule"
}
|
{
"answer": "Martine Carol",
"evidence_list": [],
"index": 620,
"need_tools_kwargs": true,
"question": "Who is the spouse of Ernest the Rebel's director?",
"question_type": "inference_query",
"split": "test",
"tools_kwargs": {
"google_search": {
"create_kwargs": {
"ground_truth": "Martine Carol"
}
},
"scrape": {
"create_kwargs": {
"ground_truth": "Martine Carol"
}
}
}
}
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.