List your company

Risk Source Prediction on ATBench

52AccuracyAgentDoG
2.28815.19428.141.006May 26, 2026

Evaluation results

MethodLinks
AgentDoGBase Model=Qwen3-4B2026.0552
AgentDoGBase Model=Llama3.1-8B-Instruct2026.0551.6
TRACESBase Model=Llama3.1-8B-Instruct2026.0550
TRACESBase Model=Qwen3-4B2026.0548.8
GPT-5.2Base Model=Original2026.0541.6
Gemini-3-FlashBase Model=Original2026.0538
Gemini-3-ProBase Model=Original2026.0536.8
QWQ-32BBase Model=Original2026.0523.2
Llama3.1-8BBase Model=Original2026.056.2
Qwen3-4BBase Model=Original2026.054.2