List your company

User Simulation Behavioral Alignment on tau2-Bench Airline (test)

90.3HLHumans
29.87645.56361.2576.937May 13, 2026

Evaluation results

MethodLinks
HumansUser Simulator=GPT-5.4-Mini2026.0590.358.474.491.398.184.877.888
PPol: EvolvedUser Simulator=GPT-5.4-Mini2026.0560.454.557.468.787.766.966.772.5
PPol: InitialUser Simulator=GPT-5.4-Mini2026.0545.716.43159.365.657.654.459.2
DP PersonasUser Simulator=GPT-5.4-Mini2026.0535.819.627.744.559.462.642.952.4
Base-simulatorUser Simulator=GPT-5.4-Mini2026.0532.216.324.347.262.361.146.854.3