Human Preference Alignment Evaluation benchmark leaderboard · SOTA2 Research

Human Preference Alignment Evaluation

23.64PickScoreOP-GRPO

Evaluation Results

Method	Links
OP-GRPO2026.04		23.64	6.09	4.23	1.31	3.72
Flow-GRPO2026.04		23.32	6.01	4.18	1.26	3.66
OP-GRPO2026.04		23.27	5.92	4.17	1.22	3.59
Flow-GRPO2026.04		22.19	5.44	4.13	1.09	3.43
SD3.5-M2026.04		21.72	5.31	4.07	0.92	3.29