Toxicity
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Toxicity
0.45S-unlearning Score
16
Toxicity
64Steering Success
11
Toxicity binary subsample (test)
2.08AUC-DEL
10
Toxicity Perturbation-based
9.52Perplexity
9
Toxicity Nooverlap BERT-small
0.003AUC-DEL Plus
7
Toxicity BERT-small targeted Kaggle 2018 (test)
0.016AUC-DEL+
7
Toxicity
90.4Original Accuracy
6
Toxicity (test)
79Test Accuracy
4
Toxicity
0.21Final Toxicity
3