Toxicity
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Toxicity Unlearning | Toxicity | S-unlearning Score0.45 | 16 | |
| Steering | Toxicity | Steering Success64 | 11 | |
| Case Deletion Diagnostics | Toxicity binary subsample (test) | AUC-DEL2.08 | 10 | |
| Adversarial Robustness | Toxicity Perturbation-based | Perplexity9.52 | 9 | |
| Text Classification | Toxicity Nooverlap BERT-small | AUC-DEL Plus0.003 | 7 | |
| Text Classification | Toxicity BERT-small targeted Kaggle 2018 (test) | AUC-DEL+0.016 | 7 | |
| Toxicity Classification | Toxicity | Original Accuracy90.4 | 6 | |
| Label aggregation assessment | Toxicity (test) | Test Accuracy79 | 4 | |
| Toxicity Reduction | Toxicity | Final Toxicity0.21 | 3 |
Showing 9 of 9 rows