CK0607/qwen3.5-9b-blogprovider-rl-cheatsheet

Qwen3.5-9B — Blog-Provider-ID — RL (GRPO, cheatsheet)

Method: GRPO RL with a train-derived style cheatsheet in context, trio curriculum. Representative best RL checkpoint (step_40).

Result: cheatsheet-free accuracy plateaued ~0.40; RL polished but did not exceed the SFT ceiling. Part of the RL-vs-SFT analysis (reasoning-channel RL is the mismatched tool for this densely-separable task).

• Base model: Qwen/Qwen3.5-9B (thinking OFF)
• Task: 3-way AI-provider classification — given a blog/essay, identify whether it was written by CLAUDE, CHATGPT, or GEMINI. Output format: <reason_why>...</reason_why><answer>LABEL\nConfidence: ...</answer>.
• Eval: val (in-distribution topics, n=414) and val_ood (held-out topics, n=471), zero eval leakage.
• Provenance: prime-rl; code at https://github.com/ChinmayK0607/prime-rl/tree/blog-author-id-experiments