Distributed Misuse Detection on DM-Cyber 100x (Trace-level)
85ROCAUC (Trace-level)Naive Agent
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Naive AgentModel=Qwen3.52026.04 | 85 | — | |
| MeerkatModel=GPT-5.4m2026.04 | 84.8 | 0.104 | |
| BufferModel=GPT-5.4m2026.04 | 74.4 | — | |
| BayesianModel=GPT-5.4m2026.04 | 73.3 | — | |
| BayesianModel=Qwen3.52026.04 | 69.7 | — | |
| Naive AgentModel=GPT-5.4m2026.04 | 64.5 | — | |
| BufferModel=Qwen3.52026.04 | 57.6 | — | |
| MeerkatModel=Qwen3.52026.04 | 53.9 | -0.311 | |
| MonitorModel=Qwen3.52026.04 | 43.9 | — |