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

Human Preference Alignment on PKU-SafeRLHF

0.324BLEU

EXO-BT

0.29280.30090.3090.3171Feb 20, 2025
Updated 27d ago

Evaluation Results

MethodLinks
2025.02
0.3240.421-5.553-6.164
2025.02
0.3140.425-5.495-6.031
2025.02
0.310.4075.7255.116
2025.02
0.310.4072.9033.495
2025.02
0.310.405-23.07-23.678
2025.02
0.3090.406-8.16-8.197
2025.02
0.3090.407-0.134-0.843
2025.02
0.3090.407-0.101-0.81
2025.02
0.3090.407-0.373-0.982
2025.02
0.3090.401-66.337-66.946
2025.02
0.3080.4071.3461.738
2025.02
0.3080.407-0.102-0.711
2025.02
0.3080.406-21.607-22.215
2025.02
0.3070.4070.144-0.464
2025.02
0.3060.4072.712.102
2025.02
0.3060.4062.2941.686
2025.02
0.3060.4054.3933.786
2025.02
0.3060.409-8.23-8.267
2025.02
0.3060.417-5.441-6.167
2025.02
0.3060.407-5.359-5.851
2025.02
0.3050.4095.3864.779
2025.02
0.3050.412-6.852-5.961
2025.02
0.3050.41-5.135-5.644
2025.02
0.3040.411-5.109-5.318
2025.02
0.3030.4090.8450.237
2025.02
0.3030.409-5.455-6.128
2025.02
0.30.411-70.91-71.518
2025.02
0.2980.435-5.273-5.881
2025.02
0.2970.413-67.474-68.082
2025.02
0.2970.433-13.442-14.05
2025.02
0.2940.406--