List your company

RLHF on HH-RLHF (held-out)

1.59Peak Gold RewardDRRO-RLHF
0.4980.78151.0651.3485Apr 30, 2026

Evaluation results

MethodLinks
DRRO-RLHFType=soft + dynamic2026.041.593.0335.74
GRPO2026.041.353.0644.83
DRRO-RLHFType=hard2026.041.292.8732.33
BSPO2026.041.213.0242.61
PPO2026.041.22.4318.25
InfoRM2026.041.143.0644.25
Ensemble-UWO2026.040.992.8434.51
Ensemble-Mean2026.040.942.7132.05
DRO-RLHF2026.040.541.7216.54