🚀 Qwen3-4B-I-1509

🧾 Model Overview

• 🏗️ Base Model: Qwen3-4B-Instruct-2507
• 🎯 Training Method: Reinforcement Learning (GRPO) with multiple reward functions

This model (Qwen3-4B-I-1509) is finetuned for 🔧 tool-use and 📞 function call generation.

---

🏆 Reward Functions

The model was trained with multi-signal rewards:

1. 📝 Rule-based Reward
   ✔️ Checks correctness of function call name and arguments.
   ➕ Partial credit for matching subsets of arguments.

2. 🔒 Self-Certainty Reward
   ⚡ Encourages confident predictions.

3. 🔧 Tool-Call Reward
   ✅ Validates structural correctness.

---

⚙️ Training Configuration

• ⚡ Optimizer: AdamW
• 📉 Learning Rate: 5e-6 with cosine decay (min_lr_rate=0.1)
• ⏳ Scheduler: cosine_with_min_lr
• 🔄 Generations per Prompt: 4

---

📊 Eval Result:

Important notes:

• Why it lower than technical report?

  There have a limit of hardware so have to reduce some max tokens when evaluation for both 2 models

• Fair evaluate ?

  I use the same configuration for all the models I review for larger or with a same size model.

Tau-Bench

🧠 Model	✈️ Airline	🛍️ Retail
Qwen3-4B-I-1509	0.2800	0.2783
Base Model	0.3000	0.2261

ACEBench

Model	Overall Accuracy
Qwen3-4B-I-1509	0.677
Qwen3-4B-Instruct-2507 (base)	0.635
Salesforce/Llama-xLAM-2-8b-fc-r	0.5792

curently upadate more

---

Contribute:

I would be happy to receive a contribution to this model and get feedback about performance, quality of model

Support me at:

📖 Citation

If you use this model in your research or application, please cite:

@misc{qwen3-4b-i-1509,
  title        = {Qwen3-4B-I-1509: Fine-tuned Qwen3-4B-Instruct with GRPO for Tool-Use and Function Calling},
  author       = {Beyoru},
  year         = {2025},
  howpublished = {\url{https://huggingface.co/beyoru/Qwen3-4B-I-1509}}
}