Share your thoughts, 1 month free Claude Pro on usSee more

LLM Preference Alignment on PrefEval

68.8AccPF

CDRA

Updated 1mo ago

Evaluation Results

Method	Links
CDRA 2025.10		68.8	62.5	21	27.7	37.5	15.2
GRPO 2025.10		67	51.8	27.3	53.6	17	1.8
SFT 2025.10		66.9	58	24.7	21.4	34.8	10.7
CoT 2025.10		62.5	61.6	20.3	54.5	18.8	0.9
TPO 2025.10		62.5	48.2	20.7	58.9	19.6	0
Few-shot 2025.10		56.3	38.4	23.3	38.4	2.7	0.9
Zero-shot 2025.10		37.5	10.7	28.7	8.9	4.5	0.9