latentMO/cot-red-handed-runs

CoT Red-Handed runs: base gpt-oss-120b and Qwen3.8-27B Inspect logs from the MATS 12 sprint (latentMO) runs of the CoT Red-Handed eval (arXiv:2505.23575): TEST split, paired side tasks, baseline attack policy, Tinker sampling, 855 samples per model, single-turn. logs/full-test-gptoss120b/, logs/full-test-qwen3.8-27b/: the side-task runs (.eval files; per-sample score metadata has main_only / side_only / main_and_side). logs/benign-test-*/: the same main tasks with no side task.… See the full description on the dataset page: https://huggingface.co/datasets/latentMO/cot-red-handed-runs.

<iframe src="https://silytics.si/embed/item/cd13ac60-943d-487b-8347-6bb2e42c1c3e" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 24 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.