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

Human Preference Alignment

Benchmarks

Task NameDataset NameSOTA ResultTrend
Human preference alignmentHuman Preference Alignment Out-of-Domain (test)
HPS-v2.135.3
7
Human preference alignmentHuman Preference Alignment In-Domain (test)
Pick Score22.46
7
Human Preference AlignmentHuman Preference Alignment
PickScore23.64
5
Human Preference AlignmentHuman Preference Alignment
Qwen-VL Score4.05
2
Showing 4 of 4 rows