dougalldeepmind/2026-08-20-odcv-feature-discovery-difficult-advice-716-5-pct-vs-numina-control
LLM-driven feature discovery over ODCV-Bench rollouts from TWO matched Qwen3.6-27B LoRA arms — 9,284 filtered instruction rows plus 716 rows that differ only in kind (constitution-grounded difficult advice vs NuminaMath chain-of-thought) — asking which reasoning and action properties separate the two models, and which go with the judged misalignment. field value experiment LLM-driven feature discovery over ODCV-Bench rollouts from TWO matched Qwen3.6-27B LoRA arms — 9… See the full description on the dataset page: https://huggingface.co/datasets/dougalldeepmind/2026-08-20-odcv-feature-discovery-difficult-advice-716-5-pct-vs-numina-control.
cards: point at the current names (naming law)
Remove a stale agreement file from an aborted detector run (98 cells, 23/25 calls blanked)
Upload folder using huggingface_hub
initial commit
