Toxicity Detection on Therapy transcripts (full evaluation set)
51.43RecallFinetuned Mistral
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Finetuned MistralTraining=Finetuned, Domain=Specific2026.05 | 51.43 | 58.21 | |
| Llama GuardType=Safety-focused model2026.05 | 34.29 | 50.89 | |
| Omni ModerationType=API-based moderation endpoint2026.05 | 8.6 | 52.6 | |
| Shield GemmaType=Safety-focused model2026.05 | 2.86 | 51.43 |