Distributed Misuse Detection on DM-Cyber 100x
45.6Trace-level APNaive Agent
Evaluation Results
| Method | Links | |
|---|---|---|
| Naive AgentModel=Qwen2.5-72B2026.04 | 45.6 | |
| MeerkatModel=GPT-4o2026.04 | 29.7 | |
| Naive AgentModel=GPT-4o2026.04 | 13.6 | |
| MeerkatModel=Qwen2.5-72B2026.04 | 9.8 | |
| BayesianModel=Qwen2.5-72B2026.04 | 8.1 | |
| BayesianModel=GPT-4o2026.04 | 5.3 | |
| BufferModel=Qwen2.5-72B2026.04 | 2.2 | |
| BufferModel=GPT-4o2026.04 | 2.1 | |
| MonitorModel=Qwen2.5-72B2026.04 | 1.1 |