List your company

Distributed Misuse Detection on DM-Bio 20x (Trace-level)

92.7Trace-level ROCAUCNaive Agent
43.8256.5169.281.89Apr 13, 2026

Evaluation results

MethodLinks
Naive AgentModel=Qwen3.52026.0492.7—
MeerkatModel=Qwen3.52026.0483.8-0.088
BayesianModel=GPT-5.4m2026.0483.7—
BayesianModel=Qwen3.52026.0475.8—
BufferModel=GPT-5.4m2026.0467.6—
Naive AgentModel=GPT-5.4m2026.0465.2—
MeerkatModel=GPT-5.4m2026.0460.6-0.231
BufferModel=Qwen3.52026.0460—
MonitorModel=Qwen3.52026.0445.7—