List your company

Discriminative Performance on Ultrafeedback 61.1k (test)

73.05AccuracyFA-DPO
46.654853.507460.3667.2126Nov 30, 2025

Evaluation results

MethodLinks
FA-DPOFlipping ratio (η)=0%, Model Backbone=Pythia-1B2025.1173.05
FA-DPOFlipping ratio (η)=30%, Model Backbone=Pythia-1B2025.1170.97
FA-DPOFlipping ratio (η)=40%, Model Backbone=Pythia-1B2025.1170.77
ROPOFlipping ratio (η)=0%, Model Backbone=Pythia-1B2025.1170.75
rDPOFlipping ratio (η)=0%, Model Backbone=Pythia-1B2025.1170.13
FA-DPOFlipping ratio (η)=20%, Model Backbone=Pythia-1B2025.1169.77
DPOFlipping ratio (η)=0%, Model Backbone=Pythia-1B2025.1168.22
cDPOFlipping ratio (η)=0%, Model Backbone=Pythia-1B2025.1167.2
FA-DPOFlipping ratio (η)=10%, Model Backbone=Pythia-1B2025.1167.2
rDPOFlipping ratio (η)=10%, Model Backbone=Pythia-1B2025.1165.9
ROPOFlipping ratio (η)=10%, Model Backbone=Pythia-1B2025.1164.93
SIMPOFlipping ratio (η)=10%, Model Backbone=Pythia-1B2025.1163.73
SIMPOFlipping ratio (η)=0%, Model Backbone=Pythia-1B2025.1163.64
cDPOFlipping ratio (η)=10%, Model Backbone=Pythia-1B2025.1162.57
DPOFlipping ratio (η)=10%, Model Backbone=Pythia-1B2025.1161.77
rDPOFlipping ratio (η)=20%, Model Backbone=Pythia-1B2025.1161.7
cDPOFlipping ratio (η)=20%, Model Backbone=Pythia-1B2025.1159
ROPOFlipping ratio (η)=20%, Model Backbone=Pythia-1B2025.1158.93
DPOFlipping ratio (η)=20%, Model Backbone=Pythia-1B2025.1158.63
SIMPOFlipping ratio (η)=20%, Model Backbone=Pythia-1B2025.1157.81
rDPOFlipping ratio (η)=30%, Model Backbone=Pythia-1B2025.1156.87
cDPOFlipping ratio (η)=30%, Model Backbone=Pythia-1B2025.1156.67
SIMPOFlipping ratio (η)=30%, Model Backbone=Pythia-1B2025.1155.53
DPOFlipping ratio (η)=30%, Model Backbone=Pythia-1B2025.1155.43
ROPOFlipping ratio (η)=30%, Model Backbone=Pythia-1B2025.1155.23
ROPOFlipping ratio (η)=40%, Model Backbone=Pythia-1B2025.1154.07
cDPOFlipping ratio (η)=40%, Model Backbone=Pythia-1B2025.1153.93
DPOFlipping ratio (η)=40%, Model Backbone=Pythia-1B2025.1151.87
SIMPOFlipping ratio (η)=40%, Model Backbone=Pythia-1B2025.1151.77
rDPOFlipping ratio (η)=40%, Model Backbone=Pythia-1B2025.1147.67