Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Preference Alignment on OpenRLHF Mixture

7.6Reward

AAD

0.55922.38714.2156.0429Sep 30, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.09
7.6-
2025.09
7.28-
2025.09
6.840.67
2025.09
6.26-
2025.09
6.180.72
2025.09
5.8857
2025.09
5.60.85
2025.09
5.45-
2025.09
5.340.83
2025.09
5.290.54
2025.09
5.2470
2025.09
4.930.89
2025.09
4.9173
2025.09
4.540.88
2025.09
4.480.69
2025.09
4.0751
2025.09
4.04-
2025.09
3.890.93
2025.09
3.6457
2025.09
3.5982
2025.09
3.590.9
2025.09
3.560.79
2025.09
3.42-
2025.09
3.1563
2025.09
3.080.55
2025.09
2.630.88
2025.09
2.420.68
2025.09
2.0672
2025.09
1.740.76
2025.09
0.830.83