Toxicity Mitigation on ToxTET
29.39ToxTET RateNone (original model)
Evaluation Results
| Method | Links | |
|---|---|---|
| None (original model)Base Model=Qwen 2.5 (7B)2025.12 | 29.39 | |
| None (original model)Base Model=Qwen 2.5 (1.5B)2025.12 | 25.5 | |
| ITIBase Model=Qwen 2.5 (7B)2025.12 | 18.76 | |
| None (original model)Base Model=Gemma 2 (2B)2025.12 | 18.53 | |
| ITIBase Model=Qwen 2.5 (1.5B)2025.12 | 16.5 | |
| LinEASBase Model=Qwen 2.5 (7B)2025.12 | 15.57 | |
| MERABase Model=Qwen 2.5 (7B)2025.12 | 15.28 | |
| LinEAS+E2E-DSAS (S)Base Model=Qwen 2.5 (7B)2025.12 | 14.83 | |
| CAABase Model=Qwen 2.5 (7B)2025.12 | 14.19 | |
| CASTBase Model=Gemma 2 (2B)2025.12 | 14.07 | |
| ITI+DSASBase Model=Qwen 2.5 (7B)2025.12 | 13.84 | |
| CAABase Model=Qwen 2.5 (1.5B)2025.12 | 13.67 | |
| MERABase Model=Qwen 2.5 (1.5B)2025.12 | 13.23 | |
| LinEAS+E2E-DSAS (R)Base Model=Qwen 2.5 (7B)2025.12 | 12.89 | |
| LinEAS+DSASBase Model=Qwen 2.5 (7B)2025.12 | 12.8 | |
| ITI+DSASBase Model=Qwen 2.5 (1.5B)2025.12 | 11.79 | |
| CASTBase Model=Qwen 2.5 (1.5B)2025.12 | 11.46 | |
| CASTBase Model=Qwen 2.5 (7B)2025.12 | 11.13 | |
| CAA+DSASBase Model=Qwen 2.5 (7B)2025.12 | 9.79 | |
| LinEASBase Model=Qwen 2.5 (1.5B)2025.12 | 9.78 | |
| CAA+DSASBase Model=Qwen 2.5 (1.5B)2025.12 | 8.64 | |
| ITIBase Model=Gemma 2 (2B)2025.12 | 6.97 | |
| LinEASBase Model=Gemma 2 (2B)2025.12 | 6.5 | |
| LinEAS+E2E-DSAS (S)Base Model=Qwen 2.5 (1.5B)2025.12 | 6.11 | |
| LinEAS+E2E-DSAS (R)Base Model=Qwen 2.5 (1.5B)2025.12 | 5.18 | |
| MERABase Model=Gemma 2 (2B)2025.12 | 4.61 | |
| CAABase Model=Gemma 2 (2B)2025.12 | 4.51 | |
| LinEAS+DSASBase Model=Qwen 2.5 (1.5B)2025.12 | 3.98 | |
| CAA+DSASBase Model=Gemma 2 (2B)2025.12 | 3.48 | |
| ITI+DSASBase Model=Gemma 2 (2B)2025.12 | 3.17 | |
| LinEAS+DSASBase Model=Gemma 2 (2B)2025.12 | 2.26 | |
| LinEAS+E2E-DSAS (R)Base Model=Gemma 2 (2B)2025.12 | 1.64 | |
| LinEAS+E2E-DSAS (S)Base Model=Gemma 2 (2B)2025.12 | 0.91 |