violetxi/qwen35-9b-wmrl-v4-R-only

wm-internalization v4 checkpoint — condition R-only, save final. Full-finetune of Qwen/Qwen3.5-9B on the Calderwood & Harkness synthetic law-firm corpus (world-internalization study, v4 lineage: 9B student, ~50k think-on seed pool). Grafted back into the hub composite layout (Qwen3_5ForConditionalGeneration) — servable with vLLM out of the box.

• training data: see train_summary.json in the training run directory
• graft: {"trained": "/scratch/11457/ziyxiang/wm-rl-runs/ckpts-v4/R-only/final", "ref": "/scratch/11457/ziyxiang/.cache/huggingface/hub/models--Qwen--Qwen3.5-9B/snapshots/c202236235762e1c871ad0ccb60c8ee5ba337b9a", "replaced": 427}
• uploaded: 2026-09-07T14:18:57+00:00 by hf_upload.py (PLAN4.md F-D policy)