List your company

Dialog on Social & Entertainment Out-of-Domain

87.13AccuracyBaseline
23.138839.751956.36572.9781Feb 5, 2026

Evaluation results

MethodLinks
BaselineTraining Steps=10002026.0287.13
DFPOTraining Steps=10002026.0286.4
Reinforce++Training Steps=10002026.0285.4
PPOTraining Steps=10002026.0259.7
Dr.GRPOTraining Steps=10002026.0250.8
GRPOTraining Steps=10002026.0225.6