List your company

Case-level safety detection on TRACE

0.901ROC-AUCMonitor
0.556760.646130.73550.82487Apr 13, 2026

Evaluation results

MethodLinks
MonitorTPC=50, Model=Qwen3.52026.040.901—
MeerkatTPC=100, Model=Qwen3.52026.040.870.023
MeerkatTPC=50, Model=Qwen3.52026.040.86-0.041
MeerkatTPC=10, Model=Qwen3.52026.040.8520.017
MonitorTPC=100, Model=Qwen3.52026.040.847—
MonitorTPC=10, Model=GLM-52026.040.838—
MonitorTPC=10, Model=Qwen3.52026.040.835—
MonitorTPC=50, Model=GLM-52026.040.821—
Naive AgentTPC=50, Model=Qwen3.52026.040.818—
MeerkatTPC=25, Model=Qwen3.52026.040.7980.008
MonitorTPC=25, Model=GLM-52026.040.797—
MonitorTPC=25, Model=Qwen3.52026.040.79—
MonitorTPC=10, Model=GPT-5.4m2026.040.782—
MeerkatTPC=100, Model=GLM-52026.040.7620
MonitorTPC=100, Model=GLM-52026.040.762—
MeerkatTPC=10, Model=GLM-52026.040.755-0.083
Naive AgentTPC=25, Model=Qwen3.52026.040.742—
MeerkatTPC=25, Model=GLM-52026.040.742-0.055
MeerkatTPC=10, Model=GPT-5.4m2026.040.73-0.051
Naive AgentTPC=10, Model=Qwen3.52026.040.729—
MeerkatTPC=50, Model=GLM-52026.040.716-0.105
MonitorTPC=50, Model=GPT-5.4m2026.040.699—
MonitorTPC=25, Model=GPT-5.4m2026.040.692—
Naive AgentTPC=100, Model=Qwen3.52026.040.658—
MeerkatTPC=50, Model=GPT-5.4m2026.040.633-0.066
MonitorTPC=100, Model=GPT-5.4m2026.040.606—
MeerkatTPC=100, Model=GPT-5.4m2026.040.582-0.025
MeerkatTPC=25, Model=GPT-5.4m2026.040.57-0.122