← back to paper

Review history

arxiv: 2607.05462 · 2 revisions

BioSecBench-Refusal: A paired metric for performance and alignment in agentic biosecurity risk assessment

1. 2026-08-02 CONDITIONAL MODERATE v1.3.0-alltime-deepseek novelty 6.0
   115946 ms 8334 in 10308 out 2026-08-02T08:34:14.465066+00:00
2. 2026-07-11 CONDITIONAL HIGH v1.1.0-grok45 novelty 6.5
   24686 ms 14094 in 2872 out 2026-07-11T16:41:20.805624+00:00