Toxicity Classification on Toxigen
60.41AccuracyMAT-STEER
Evaluation Results
| Method | Links | |
|---|---|---|
| MAT-STEERBackbone=Qwen2.5-7B2025.02 | 60.41 | |
| LITOBackbone=Qwen2.5-7B2025.02 | 58.29 | |
| ITIBackbone=Qwen2.5-7B2025.02 | 58.15 | |
| SFTBackbone=Qwen2.5-7B2025.02 | 57.76 | |
| MAT-STEERBackbone=Llama-3.1-8B2025.02 | 57.59 | |
| NL-ITIBackbone=Qwen2.5-7B2025.02 | 57.56 | |
| ICVBackbone=Qwen2.5-7B2025.02 | 56.76 | |
| MergeBackbone=Qwen2.5-7B2025.02 | 56.02 | |
| DPOBackbone=Llama-3.1-8B2025.02 | 55.94 | |
| ICLBackbone=Qwen2.5-7B2025.02 | 55.65 | |
| SFTBackbone=Llama-3.1-8B2025.02 | 55.51 | |
| DPOBackbone=Qwen2.5-7B2025.02 | 55.45 | |
| RAdaptBackbone=Llama-3.1-8B2025.02 | 55.02 | |
| RAdaptBackbone=Qwen2.5-7B2025.02 | 54.87 | |
| NL-ITIBackbone=Llama-3.1-8B2025.02 | 54.73 | |
| MergeBackbone=Llama-3.1-8B2025.02 | 54.65 | |
| LITOBackbone=Llama-3.1-8B2025.02 | 54.08 | |
| ICVBackbone=Llama-3.1-8B2025.02 | 53.61 | |
| ITIBackbone=Llama-3.1-8B2025.02 | 52.55 | |
| Qwen2.5-7BBackbone=Qwen2.5-7B, Status=Base Model2025.02 | 52.27 | |
| ICLBackbone=Llama-3.1-8B2025.02 | 51.26 | |
| Llama-3.1-8BBackbone=Llama-3.1-8B2025.02 | 48.1 |