List your company

Distributed Misuse Detection on DM-Cyber 100x (Trace-level)

85ROCAUC (Trace-level)Naive Agent
42.25653.35364.4575.547Apr 13, 2026

Evaluation results

MethodLinks
Naive AgentModel=Qwen3.52026.0485—
MeerkatModel=GPT-5.4m2026.0484.80.104
BufferModel=GPT-5.4m2026.0474.4—
BayesianModel=GPT-5.4m2026.0473.3—
BayesianModel=Qwen3.52026.0469.7—
Naive AgentModel=GPT-5.4m2026.0464.5—
BufferModel=Qwen3.52026.0457.6—
MeerkatModel=Qwen3.52026.0453.9-0.311
MonitorModel=Qwen3.52026.0443.9—