LAB-Bench CloningScenarios (SecureBio)
Science
SecureBio's own runs of LAB-Bench CloningScenarios without tool use. These are the same 33 public questions as lab-bench-cloningscenarios under a different protocol. Evaluated in Inspect with zero-shot prompting, usually over 10 runs. SecureBio doesn't state how refusals enter accuracy. Claude Opus 4.8 prints 95.9% with 46.7% refused, more than it could score if refusals counted as wrong, so accuracy appears to cover answered questions only and each model's denominator differs. These scores aren't comparable with other runs of the test, which is why this is a diagnostic.
Benchmark source: Securebio