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

BAD

Benchmarks

Task NameDataset NameSOTA ResultTrend
Toxicity DetectionBAD
F1 Score80.8
11
Language DetoxificationBAD (test)
Toxicity Reduction37
10
Open domain dialogueBAD
RSR53.7
9
Red Teaming against BB-3BBAD
RSR66.4
9
Language DetoxificationBAD (val)
Toxicity Proportion11
7
Red TeamingBAD Against Friend Chat (test)
RSR64.2
7
Red TeamingBAD Against Marv (test)
RSR88.1
7
Showing 7 of 7 rows