List your company

RLHF Alignment on UltraFeedback In-domain v1 (test)

81Win RateGEB-π
46.191255.228164.26573.3019Sep 27, 2025

Evaluation results

MethodLinks
GEB-πBackbone=Mistral-Instruct-v0.3, Divergence=KL, Alpha=12025.09810.7542
GEB-1/πBackbone=Mistral-Instruct-v0.3, Divergence=KL, Alpha=12025.09800.7554
XPOBackbone=Mistral-Instruct-v0.3, Divergence=KL, Alpha=12025.0979.710.7492
GEB-arctanh(π − 1)Backbone=Mistral-Instruct-v0.3, Divergence=KL, Alpha=12025.0979.710.7559
VPOBackbone=Mistral-Instruct-v0.3, Divergence=KL, Alpha=12025.0978.570.7426
f-DPOBackbone=Mistral-Instruct-v0.3, Divergence=KL, Alpha=12025.0978.420.748
FEBBackbone=Mistral-Instruct-v0.3, Divergence=KL, Alpha=12025.0978.420.748
SELMBackbone=Mistral-Instruct-v0.3, Divergence=KL, Alpha=12025.0977.560.753
GEB-arctanh(π − 1)Backbone=Mistral-Instruct-v0.3, Divergence=Hellinger, Alpha=0.52025.0975.690.6614
GEB-πBackbone=Mistral-Instruct-v0.3, Divergence=Hellinger, Alpha=0.52025.0975.480.6641
GEB-arctanh(π − 1)Backbone=LLaMA-3-8B-SFT, Divergence=KL, Alpha=12025.0974.980.808
GEB-1/πBackbone=LLaMA-3-8B-SFT, Divergence=KL, Alpha=12025.0974.760.8102
GEB-πBackbone=LLaMA-3-8B-SFT, Divergence=KL, Alpha=12025.0974.340.8156
SELMBackbone=LLaMA-3-8B-SFT, Divergence=KL, Alpha=12025.0974.190.8126
GEB-1/πBackbone=Mistral-Instruct-v0.3, Divergence=Hellinger, Alpha=0.52025.0973.970.6541
GEB-arctanh(π − 1)Backbone=LLaMA-3-8B-SFT, Divergence=Hellinger, Alpha=0.52025.0973.260.7877
f-DPOBackbone=LLaMA-3-8B-SFT, Divergence=KL, Alpha=12025.0973.110.805
FEBBackbone=LLaMA-3-8B-SFT, Divergence=KL, Alpha=12025.0973.110.805
f-DPOBackbone=Mistral-Instruct-v0.3, Divergence=Hellinger, Alpha=0.52025.0972.690.6536
XPOBackbone=LLaMA-3-8B-SFT, Divergence=KL, Alpha=12025.0972.40.8119
GEB-arctanh(π − 1)Backbone=LLaMA-3-8B-SFT, Divergence=Average2025.0972.380.7842
GEB-1/πBackbone=LLaMA-3-8B-SFT, Divergence=Hellinger, Alpha=0.52025.0972.250.7859
GEB-1/πBackbone=LLaMA-3-8B-SFT, Divergence=Average2025.0971.730.7851
GEB-πBackbone=LLaMA-3-8B-SFT, Divergence=Hellinger, Alpha=0.52025.0971.680.784
VPOBackbone=LLaMA-3-8B-SFT, Divergence=KL, Alpha=12025.0971.610.7971
FEBBackbone=Mistral-Instruct-v0.3, Divergence=Hellinger, Alpha=0.52025.0971.540.6525
GEB-πBackbone=LLaMA-3-8B-SFT, Divergence=Average2025.0971.230.7892
f-DPOBackbone=LLaMA-3-8B-SFT, Divergence=Hellinger, Alpha=0.52025.0971.110.7859
f-DPOBackbone=LLaMA-3-8B-SFT, Divergence=Average2025.0970.530.7829
FEBBackbone=LLaMA-3-8B-SFT, Divergence=Average2025.0969.740.7751
GEB-πBackbone=Mistral-Instruct-v0.3, Divergence=Average2025.0969.390.672
GEB-arctanh(π − 1)Backbone=Mistral-Instruct-v0.3, Divergence=Average2025.0969.390.6721
GEB-arctanh(π − 1)Backbone=LLaMA-3-8B-SFT, Divergence=f-KL, Alpha=02025.0968.890.7569
GEB-1/πBackbone=Mistral-Instruct-v0.3, Divergence=Average2025.0968.740.6715
FEBBackbone=LLaMA-3-8B-SFT, Divergence=Hellinger, Alpha=0.52025.0968.170.7591
GEB-1/πBackbone=LLaMA-3-8B-SFT, Divergence=f-KL, Alpha=02025.0968.170.7591
FEBBackbone=LLaMA-3-8B-SFT, Divergence=f-KL, Alpha=02025.0967.950.7611
GEB-πBackbone=LLaMA-3-8B-SFT, Divergence=f-KL, Alpha=02025.0967.670.7681
f-DPOBackbone=Mistral-Instruct-v0.3, Divergence=Average2025.0967.40.6645
f-DPOBackbone=LLaMA-3-8B-SFT, Divergence=f-KL, Alpha=02025.0967.380.7579
FEBBackbone=Mistral-Instruct-v0.3, Divergence=Average2025.0965.830.6644
GEB-arctanh(π − 1)Backbone=Mistral-Instruct-v0.3, Divergence=f-KL, Alpha=02025.0952.760.5989
GEB-1/πBackbone=Mistral-Instruct-v0.3, Divergence=f-KL, Alpha=02025.0952.260.6051
GEB-πBackbone=Mistral-Instruct-v0.3, Divergence=f-KL, Alpha=02025.0951.680.5976
f-DPOBackbone=Mistral-Instruct-v0.3, Divergence=f-KL, Alpha=02025.0951.110.5918
FEBBackbone=Mistral-Instruct-v0.3, Divergence=f-KL, Alpha=02025.0947.530.5928