qwen3-4b-dpo-v2/README.md

base_model, datasets, language, license, library_name, pipeline_tag, tags

base_model	datasets	language	license	library_name	pipeline_tag	tags
takami2022/qwen3-4b-dpo-v1	u-10bei/dpo-dataset-qwen-cot	en	apache-2.0	transformers	text-generation	dpo unsloth qwen alignment

qwen3-4b-dpo-v2

DPO v1（0.7703）をベースにBeta=0.05で再学習したモデル。

Training Configuration

• Base model: takami2022/qwen3-4b-dpo-v1
• Method: DPO (Direct Preference Optimization)
• Epochs: 1
• Learning rate: 1e-07
• Beta: 0.05 (前回0.1→0.05に変更)
• Max sequence length: 1024
• LoRA Config: r=16, alpha=32 (merged into base)