ResearchBenchmarksPreference Ranking on UltraFeedback Ambiguous (test)Follow73.4Ranking Accuracy (RA)SA-DPO71.52872.01472.572.986Apr 30, 2026Evaluation ResultsMethodMethodLinksRanking Accuracy (RA)SA-DPOBackbone=Llama-3-8B, F...Backbone=Llama-3-8B, Fine-tuning Protocol=LoRA2026.0473.4SimPOBackbone=Llama-3-8B, F...Backbone=Llama-3-8B, Fine-tuning Protocol=LoRA2026.0472.2DPOBackbone=Llama-3-8B, F...Backbone=Llama-3-8B, Fine-tuning Protocol=LoRA2026.0471.6