Toxicity Detection on Reconstructed Embedding Space Toxicity (100 sampled sentences per label)
100AccuracyLabel Forensics
Evaluation results
| Method | Links | |
|---|---|---|
| Label ForensicsLabel=0 (Non-Toxic)2025.12 | 100 | |
| Label ForensicsLabel=Avg.2025.12 | 95.5 | |
| Label ForensicsLabel=1 (Toxic)2025.12 | 91 |