List your company

LLM Alignment on Taobao Live proprietary fine-grained preference dataset

1.53Win ScorePD (ours)
0.4380.72151.0051.2885Aug 11, 2025

Evaluation results

MethodLinks
PD (ours)Budget (λ)=50%2025.081.53
PD (ours)Budget (λ)=40%2025.081.45
PD (ours)Budget (λ)=30%2025.081.43
RANDBudget (λ)=40%2025.081.37
PD (mid)Budget (λ)=50%2025.081.37
ALLBudget (λ)=100%2025.081.33
PD (mid)Budget (λ)=30%2025.081.33
RANDBudget (λ)=50%2025.081.32
PD (mid)Budget (λ)=40%2025.081.32
RANDBudget (λ)=30%2025.081.18
PD (high)Budget (λ)=50%2025.080.93
PD (high)Budget (λ)=40%2025.080.57
PD (high)Budget (λ)=30%2025.080.48