List your company

LLM Alignment on UltraFeedback (test)

21AlpacaEval 2 Win Rate (WR)PD (ours)
3.69448.187212.6817.1728Aug 11, 2025

Evaluation results

MethodLinks
PD (ours)Model=Llama3.1-8B, Category=SELT2025.082126.111.2418.6
RAFModel=Llama3.1-8B, Category=SELT2025.0819.6423.341.1816.2
DMPOModel=Llama3.1-8B, Category=FULL2025.0819.4224.731.2542.1
PD (rati.)Model=Llama3.1-8B, Category=SELT2025.0818.8525.381.2310.1
AVG.Model=Llama3.1-8B, Category=FULL2025.0816.6322.211.1533.6
ALLModel=Llama3.1-8B, Category=FULL2025.0815.1821.141.0833.6
RANDModel=Llama3.1-8B, Category=SELT2025.0814.7219.561.0210.1
OVA.Model=Llama3.1-8B, Category=FULL2025.0814.3519.96133.6
PDModel=Qwen2.5-7B, Strategy Group=SELT2025.0813.8122.17——
DMPOModel=Qwen2.5-7B, Strategy Group=FULL2025.0812.7718.42——
RAFModel=Qwen2.5-7B, Strategy Group=SELT2025.0812.0118.06——
PD (rati.)Model=Qwen2.5-7B, Strategy Group=SELT2025.0810.5317.03——
AVG.Model=Qwen2.5-7B, Strategy Group=FULL2025.089.7315.74——
RANDModel=Qwen2.5-7B, Strategy Group=SELT2025.089.3516.03——
ALLModel=Qwen2.5-7B, Strategy Group=FULL2025.089.3315.6——
OVA.Model=Qwen2.5-7B, Strategy Group=FULL2025.088.1116.47——
SFTModel=Llama3.1-8B, Category=INIT2025.087.08140.640
SFTModel=Qwen2.5-7B, Strategy Group=INIT2025.084.369.12——
ALLModel=Qwen2.5-7B, Strategy Group=FULL2025.08——1.15—
AVG.Model=Qwen2.5-7B, Strategy Group=FULL2025.08——1.2—
DMPOModel=Qwen2.5-7B, Strategy Group=FULL2025.08——1.33—
OVA.Model=Qwen2.5-7B, Strategy Group=FULL2025.08——1—
PDModel=Qwen2.5-7B, Strategy Group=SELT2025.08——1.3—
PD (rati.)Model=Qwen2.5-7B, Strategy Group=SELT2025.08——1.22—
RAFModel=Qwen2.5-7B, Strategy Group=SELT2025.08——1.28—
RANDModel=Qwen2.5-7B, Strategy Group=SELT2025.08——1.15—
SFTModel=Qwen2.5-7B, Strategy Group=INIT2025.08——0.77—