Configurable Preference Tuning with Rubric-Guided Synthetic Data
Paper • 2506.11702 • Published • 1
Configurable Preference Tuning ⚙️📝
CPT uses rubric-guided synthetic data and DPO to enable LLMs to dynamically adjust behavior (e.g., writing style) at inference with system prompts
-
-
vicgalle/creative-rubrics-preferences
Viewer • Updated • 900 • 157 • 3 -
vicgalle/creative-rubrics
Viewer • Updated • 1.33k • 140 • 8 -
vicgalle/configurable-preference-mistral-nemo-12b
Text Generation • Updated -
vicgalle/configurable-preference-rocinante-12b
Text Generation • Updated • 6 -
vicgalle/configurable-preference-phi4
Text Generation • Updated -
vicgalle/configurable-preference-qwen3-4b
Text Generation • Updated