List your company

Sabotage Detection on TRACE

90.1Average PrecisionMonitor
52.45262.2267281.774Apr 13, 2026

Evaluation results

MethodLinks
MonitorTPC=50, Model=Qwen3.52026.0490.1
MeerkatTPC=10, Model=Qwen3.52026.0488.2
MeerkatTPC=100, Model=Qwen3.52026.0488.1
MeerkatTPC=50, Model=Qwen3.52026.0487.4
MonitorTPC=10, Model=GLM-52026.0486.1
MonitorTPC=100, Model=Qwen3.52026.0484.9
MonitorTPC=10, Model=Qwen3.52026.0484.3
MonitorTPC=25, Model=GLM-52026.0483
MonitorTPC=10, Model=GPT-5.4m2026.0482.2
Naive AgentTPC=50, Model=Qwen3.52026.0481.9
MonitorTPC=25, Model=Qwen3.52026.0480.8
MeerkatTPC=25, Model=Qwen3.52026.0477.2
MonitorTPC=50, Model=GLM-52026.0475.4
Naive AgentTPC=10, Model=Qwen3.52026.0475.1
Naive AgentTPC=25, Model=Qwen3.52026.0474.3
MeerkatTPC=10, Model=GPT-5.4m2026.0471.7
MeerkatTPC=10, Model=GLM-52026.0471.4
MeerkatTPC=100, Model=GLM-52026.0470
MonitorTPC=100, Model=GLM-52026.0470
MeerkatTPC=25, Model=GLM-52026.0468
MonitorTPC=50, Model=GPT-5.4m2026.0466.1
MeerkatTPC=50, Model=GLM-52026.0463.8
MonitorTPC=25, Model=GPT-5.4m2026.0463.4
Naive AgentTPC=100, Model=Qwen3.52026.0462.1
MeerkatTPC=50, Model=GPT-5.4m2026.0459.3
MonitorTPC=100, Model=GPT-5.4m2026.0456.4
MeerkatTPC=100, Model=GPT-5.4m2026.0454.6
MeerkatTPC=25, Model=GPT-5.4m2026.0453.9