LLM Preference Alignment on DeepPref benchmark leaderboard · SOTA2 Research

LLM Preference Alignment on DeepPref

84.7AccPFCDRA

Evaluation Results

Method	Links
CDRA2025.10		84.7	76.3	32.3	47	65	42.7
GRPO2025.10		83.7	70.3	30.7	46.3	58.7	34
SFT2025.10		83.3	75	34.7	46.7	63.7	40.3
CoT2025.10		59.7	49.3	50.3	39	25.3	0.7
TPO2025.10		55.3	36.3	56.3	29.7	15.7	0
Few-shot2025.10		49.7	32.7	61.3	29	10.7	0.3
Zero-shot2025.10		23	6.7	76	4.7	3	0