SOTA Human Preference Alignment benchmarks and papers with code

Benchmarks

Dataset Name	SOTA Method	Metric
PKU-SafeRLHF	EXO-BT	BLEU0.324	31	26d ago
MM-AlignBench 1.0 (test)		Win Rate84.9	18	1mo ago
Out-of-Domain (test)		Agreement81.8	15	23d ago
HH (test)	TRE-P	Reward3.8764	14	1mo ago
REACT-Video	REACT	Acc (Tie, Overall)61	12	1mo ago
HPD v2	BranchGRPO	HPS-v2.10.379	10	1mo ago
HPD v2 (test)	DanceGRPO	HPSv2.10.371	7	1mo ago
Human Preference Alignment Out-of-Domain (test)	TAFS-GRPO	HPS-v2.135.3	7	1mo ago
Human Preference Alignment In-Domain (test)	TAFS-GRPO	Pick Score22.46	7	1mo ago
Multi-Challenge	Qwen3-30A3-2507	Avg@349.4	6	1mo ago
ArenaHard V2	Qwen3-30A3-2507	Avg@3 Score60	6	1mo ago
Human Preference Alignment	OP-GRPO	PickScore23.64	5	12d ago
HPDv2	TreeGRPO	HPS-v2.10.3735	5	1mo ago
PickScore	VGPO	PickScore (Task)23.55	5	1mo ago
VideoGen-RewardBench (test)	VideoReward	VQ Acc (w/ Tie)66	5	1mo ago
User study dataset HH-RLHF and PKU-SafeRLHF prompts (test)	DPO-HPS	Quality Score3.93	4	26d ago
PickScore	SuperFlow	PickScore86.851	4	1mo ago
DrawBench Task-specific (test)	DenseGRPO	PickScore (Task Metric)24.64	4	1mo ago
PKU-Safety	DPO-HPS	Win Rate67.1	3	26d ago

Showing 19 of 19 rows