tintin1027/dpo_data_0907 download history

tintin1027/dpo_data_0907 is a text generation dataset on the Hugging Face Hub. In the last 30 days it was downloaded 195 times (65 in the last 7 days), and 195 times in total. It ranks #68,932 among datasets by monthly downloads.

PrefGap DPO pairs (0907) Preference pairs for DPO on Qwen3-4B, written from nested n=50 / n=100 / n=200 profiles on held-out HelpSteer2 prompts. Pipeline: profile-inferred baseline → isolated rejected / chosen forks → eight-gate semantic verifier → one chosen self-refine. Analyst/synt

Open tintin1027/dpo_data_0907 on Hugging Face

Sister project: Paper Pulse, the upvote history of every Hugging Face Daily Paper.