List your company

Preference Optimization on HelpSteer2 (test)

0.8Avg Pref Score vs QWEN2.5-0.5BSTACKELBERGGDA
-0.0320.1840.40.616Dec 18, 2025

Evaluation results

MethodLinks
STACKELBERGGDARole=Follower2025.120.80.6560.5940.6050
STACKELBERGGDARole=Leader2025.120.7340.6130.50300.395
NASH-MD-PG2025.120.7210.60700.4970.406
RLOO2025.120.59300.3930.3870.344
QWEN2.5-0.5B2025.1200.4070.2790.2660.2