ResearchBenchmarksHuman Preference Alignment on HH-RLHF and PKU-SafeRLHF (test)Follow3.93Quality ScoreDPO-HPS3.6183.6993.783.861Feb 20, 2025Evaluation ResultsMethodMethodLinksQuality ScoreDPO-HPSSetting=fine-tuningSetting=fine-tuning2025.023.93DPO-BTSetting=fine-tuningSetting=fine-tuning2025.023.82DPO-PLSetting=fine-tuningSetting=fine-tuning2025.023.69SFTSetting=fine-tuningSetting=fine-tuning2025.023.63