Trace-level detection on ImpossibleBench
88.7APMeerkat
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MeerkatTPC=100, Backbone Model=Qwen3.52026.04 | 88.7 | 0.07 | |
| MeerkatTPC=50, Backbone Model=Qwen3.52026.04 | 84.5 | 0.056 | |
| MonitorTPC=10, Backbone Model=GPT-5.4m2026.04 | 83.2 | — | |
| MeerkatTPC=10, Backbone Model=Qwen3.52026.04 | 82 | 0.032 | |
| MeerkatTPC=10, Backbone Model=GPT-5.4m2026.04 | 82 | -0.012 | |
| BayesianTPC=100, Backbone Model=Qwen3.52026.04 | 81.7 | — | |
| BayesianTPC=100, Backbone Model=GPT-5.4m2026.04 | 81.7 | — | |
| BufferTPC=100, Backbone Model=GPT-5.4m2026.04 | 81.5 | — | |
| MonitorTPC=100, Backbone Model=Qwen3.52026.04 | 79.3 | — | |
| BayesianTPC=50, Backbone Model=Qwen3.52026.04 | 78.9 | — | |
| BayesianTPC=50, Backbone Model=GPT-5.4m2026.04 | 78.9 | — | |
| BayesianTPC=10, Backbone Model=Qwen3.52026.04 | 78.8 | — | |
| BayesianTPC=10, Backbone Model=GPT-5.4m2026.04 | 78.8 | — | |
| BufferTPC=100, Backbone Model=Qwen3.52026.04 | 78.1 | — | |
| MeerkatTPC=25, Backbone Model=GPT-5.4m2026.04 | 75.5 | 0.017 | |
| MeerkatTPC=50, Backbone Model=GPT-5.4m2026.04 | 75.3 | -0.036 | |
| MeerkatTPC=25, Backbone Model=Qwen3.52026.04 | 75 | 0.012 | |
| MonitorTPC=50, Backbone Model=Qwen3.52026.04 | 74.8 | — | |
| BayesianTPC=25, Backbone Model=Qwen3.52026.04 | 73.8 | — | |
| BayesianTPC=25, Backbone Model=GPT-5.4m2026.04 | 73.8 | — | |
| MeerkatTPC=100, Backbone Model=GPT-5.4m2026.04 | 73.1 | -0.085 | |
| MonitorTPC=25, Backbone Model=GPT-5.4m2026.04 | 72.6 | — | |
| MonitorTPC=100, Backbone Model=GPT-5.4m2026.04 | 70 | — | |
| BufferTPC=50, Backbone Model=GPT-5.4m2026.04 | 68.1 | — | |
| BufferTPC=10, Backbone Model=GPT-5.4m2026.04 | 67.4 | — | |
| MonitorTPC=50, Backbone Model=GPT-5.4m2026.04 | 65.9 | — | |
| BufferTPC=50, Backbone Model=Qwen3.52026.04 | 64.1 | — | |
| MonitorTPC=10, Backbone Model=Qwen3.52026.04 | 64 | — | |
| BufferTPC=10, Backbone Model=Qwen3.52026.04 | 63.6 | — | |
| BufferTPC=25, Backbone Model=Qwen3.52026.04 | 62 | — | |
| BufferTPC=25, Backbone Model=GPT-5.4m2026.04 | 58 | — | |
| MonitorTPC=25, Backbone Model=Qwen3.52026.04 | 57.8 | — | |
| Naive AgentTPC=50, Backbone Model=Qwen3.52026.04 | 57.3 | — | |
| Naive AgentTPC=10, Backbone Model=Qwen3.52026.04 | 53.2 | — | |
| Naive AgentTPC=100, Backbone Model=GPT-5.4m2026.04 | 49.7 | — | |
| Naive AgentTPC=25, Backbone Model=Qwen3.52026.04 | 46.2 | — | |
| Naive AgentTPC=100, Backbone Model=Qwen3.52026.04 | 41.9 | — | |
| Naive AgentTPC=50, Backbone Model=GPT-5.4m2026.04 | 26.4 | — | |
| Naive AgentTPC=10, Backbone Model=GPT-5.4m2026.04 | 10 | — | |
| Naive AgentTPC=25, Backbone Model=GPT-5.4m2026.04 | 4 | — |