Evaluation & BenchmarksEdwin H. Wintermute et al.

BioSecBench-Refusal released

BioSecBench-Refusal released — a benchmark that measures, as a pair, the ability to identify hazards in biology research tasks and the rate of over-refusal on legitimate dual-use research

Why it matters

Ties capability and refusal behavior into a single paired metric, formally bringing the over-refusal problem into biosecurity evaluation

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/html/2607.05462v1

Added in Week 35, 2026 · See that week

Same week

Thinking Machines Lab posts The Future Worth Building Is HumanOfficialWhat to Keep, What to Forget publishedPaperNYT and Daily News seek sanctions, alleging OpenAI hid evidencePressOllama raises $89 million; 8.9 million developersOfficial

Nearby on the same axis

imo-2026 project runs 7 frontier models on IMO 2026, publishes audit trailsOfficialEpoch AI releases a revised FrontierMathOfficialLiu et al. release ABC-BenchPaperSurge AI publishes LMArena is a cancer on AIOfficial