Distributed Misuse Detection on DM-Bio 100x
14.2Trace-level APMeerkat
Evaluation Results
| Method | Links | |
|---|---|---|
| MeerkatModel=Qwen2.5-72B2026.04 | 14.2 | |
| Naive AgentModel=Qwen2.5-72B2026.04 | 13.8 | |
| BayesianModel=Qwen2.5-72B2026.04 | 4 | |
| BufferModel=Qwen2.5-72B2026.04 | 2.6 | |
| BayesianModel=GPT-4o2026.04 | 2 | |
| MeerkatModel=GPT-4o2026.04 | 1.5 | |
| MonitorModel=Qwen2.5-72B2026.04 | 1.2 | |
| BufferModel=GPT-4o2026.04 | 1.1 | |
| Naive AgentModel=GPT-4o2026.04 | 1 |