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

WebGPT

Benchmarks

Task NameDataset NameSOTA ResultTrend
Question AnsweringWebGPT
Average Score76.42
18
Preference AlignmentWebGPT (test)
Accuracy61.24
11
Direct Preference OptimizationWebGPT
Accuracy58.92
11
Reward ModelingWebGPT
Accuracy58.4
8
Preference ClassificationWebGPT comparisons (test)
Accuracy60.8
7
Showing 5 of 5 rows