List your company

Dialog on Real-world Dialog Domains Aggregate

87.31Average AccuracyDFPO
25.440441.502757.56573.6273Feb 5, 2026

Evaluation results

MethodLinks
DFPOTraining Steps=10002026.0287.31
BaselineTraining Steps=10002026.0286.26
Reinforce++Training Steps=10002026.0281.15
PPOTraining Steps=10002026.0261.55
Dr.GRPOTraining Steps=10002026.0248.42
GRPOTraining Steps=10002026.0227.82