LAB-Bench ProtocolQA (SecureBio)
Science
SecureBio's own runs of LAB-Bench ProtocolQA. These are the same 108 public questions as lab-bench-protocolqa under a different protocol. Evaluated in Inspect with zero-shot prompting, usually over 10 runs. SecureBio doesn't state how refusals enter accuracy. Gemini 4 Argon prints 95.9% with 6.6% 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