Pinned Loading

1. Psy-Qwen-SFT Psy-Qwen-SFT Public

   AI Psychological Counselor (Qwen 3.5B-0.8B Fine-Tuning Project)

   Shell 1

2. Psy-Qwen-DPO Psy-Qwen-DPO Public

   DPO alignment of a Chinese psychological counselor LLM (Qwen3.5-0.8B). 74.9% win rate vs SFT baseline on 202 held-out prompts, evaluated by DeepSeek V4-Flash with 2-way position-bias mitigation.

   Python