Toxicity Classification on ToxiFrench Sbench
99Class 0 Precisiongpt-4o
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| gpt-4o2025.08 | 99 | 69.6 | 81.7 | 76.6 | 99.3 | 86.4 | 84.4 | |
| deepseek-V32025.08 | 98.3 | 68.3 | 80.6 | 75.7 | 98.8 | 85.8 | 83.6 | |
| deepseek-R12025.08 | 96.7 | 70.9 | 81.8 | 77 | 97.6 | 86.1 | 84.2 | |
| gpt-4o-mini2025.08 | 96.2 | 77.1 | 85.6 | 80.9 | 97 | 88.2 | 87 | |
| mistral large2025.08 | 95.3 | 75.2 | 84.1 | 79.5 | 96.3 | 87.1 | 85.7 | |
| gemini-2.5-flash2025.08 | 95.2 | 62.5 | 75.5 | 72.1 | 96.8 | 82.7 | 79.7 | |
| gemini-1.5-pro2025.08 | 94.2 | 75.5 | 83.8 | 79.6 | 95.4 | 86.8 | 85.4 | |
| Qwen-2.5-7Bmode=Instruct2025.08 | 91.7 | 70.2 | 79.5 | 75.8 | 93.7 | 83.8 | 81.9 | |
| o4-mini2025.08 | 86.2 | 76.7 | 81.2 | 79 | 87.8 | 83.1 | 82.2 | |
| Qwen-2.5-3Bmode=Instruct2025.08 | 82.9 | 81 | 81.9 | 81.4 | 83.3 | 82.3 | 82.1 | |
| mistral moderation2025.08 | 80.9 | 83.7 | 82.3 | 83.1 | 80.3 | 81.7 | 82 | |
| CamemBERTversion=Ours2025.08 | 77.9 | 85.4 | 81.5 | 83.9 | 75.8 | 79.6 | 80.6 | |
| openai omni2025.08 | 76 | 78.1 | 77 | 77.5 | 75.4 | 76.4 | 76.7 | |
| Mistral-7B-Instruct2025.08 | 73.5 | 72.9 | 73.2 | 73.1 | 73.8 | 73.5 | 73.3 | |
| CamemBERTversion=Ext. I2025.08 | 72.6 | 75.1 | 73.8 | 74.2 | 71.6 | 72.9 | 73.3 | |
| Qwen-3-4Bmode=Instruct2025.08 | 72.1 | 88.9 | 79.6 | 85.5 | 65.6 | 74.2 | 77.2 | |
| CamemBERTversion=Ext. II2025.08 | 66.7 | 62.4 | 64.5 | 64.7 | 68.9 | 66.7 | 65.6 | |
| perspective2025.08 | 63.8 | 90.3 | 74.8 | 83.5 | 48.8 | 61.6 | 69.6 | |
| llamaguard2025.08 | 61 | 97.1 | 74.9 | 92.9 | 37.9 | 53.8 | 67.5 |