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

HHH-Alignment

Benchmarks

Task NameDataset NameSOTA ResultTrend
Reward ModelingHHH-Alignment Reversed
Accuracy86.2
9
Reward ModelingHHH-Alignment Standard
Accuracy91.8
9
Reward ModelingHHH-Alignment (OOD)
Accuracy79.8
8
Reward ModelingHHH-Alignment OOD (test)
Score78.7
8
Reward ModelingHHH Alignment
Accuracy87.8
4
Showing 5 of 5 rows