List your company

Error Propagation Analysis on AgentProp-Bench semantic-wrong 1,350 traces (P2)

74S1 (%)GPT-4o
2.2420.8739.558.13Apr 17, 2026

Evaluation results

MethodLinks
GPT-4on=50, nS2=112026.047422740.30.820.68
o3-minin=50, nS2=82026.047416700.220.620.76
GPT-4.1-nanon=200, nS2=352026.0462.517.584.50.280.970.83
DeepSeek-V3n=200, nS2=142026.0452762.50.130.430.57
GPT-4.1-minin=200, nS2=212026.0450.510.565.50.210.670.66
Gemini-2.5-Fn=50, nS2=32026.04506760.120.670.84
GPT-4o-minin=200, nS2=452026.044022.5710.560.840.84
GPT-3.5-Tn=200, nS2=192026.0437.59.567.50.250.740.76
Gemini-2.0-Fn=200, nS2=22026.0451670.20.50.9