Safety on T3
85.1T3 ScoreSource
Evaluation Results
| Method | Links | |
|---|---|---|
| SourceApproach=Source2025.12 | 85.1 | |
| SSU-WandaApproach=SSU-Wanda2025.12 | 85.1 | |
| SourceModel Scale=7B, Evaluation Protocol=Zero-shot2025.12 | 85.1 | |
| SSU-WandaModel Scale=7B, Evaluation Protocol=Zero-shot2025.12 | 85 | |
| SSU-RandModel Scale=7B, Evaluation Protocol=Zero-shot2025.12 | 83.5 | |
| S2FT (Down)Approach=S2FT (Down)2025.12 | 83.4 | |
| LoTAApproach=LoTA, Sparsity=90%2025.12 | 82.7 | |
| HFTModel Scale=7B, Evaluation Protocol=Zero-shot2025.12 | 82.6 | |
| LoTAApproach=LoTA, Sparsity=50%2025.12 | 82.4 | |
| AdaLoRAModel Scale=7B, Evaluation Protocol=Zero-shot2025.12 | 82.4 | |
| SSU-MagModel Scale=7B, Evaluation Protocol=Zero-shot2025.12 | 82.2 | |
| SourceModel Scale=13B, Evaluation Protocol=Zero-shot2025.12 | 82.1 | |
| SSU-WandaModel Scale=13B, Evaluation Protocol=Zero-shot2025.12 | 80.5 | |
| AdaLoRAModel Scale=13B, Evaluation Protocol=Zero-shot2025.12 | 79.9 | |
| FFTModel Scale=7B, Evaluation Protocol=Zero-shot2025.12 | 79.7 | |
| GMTModel Scale=7B, Evaluation Protocol=Zero-shot2025.12 | 79.5 | |
| HFTModel Scale=13B, Evaluation Protocol=Zero-shot2025.12 | 78.5 | |
| SSU-RandModel Scale=13B, Evaluation Protocol=Zero-shot2025.12 | 78.3 | |
| SSU-MagModel Scale=13B, Evaluation Protocol=Zero-shot2025.12 | 77.6 | |
| GMTModel Scale=13B, Evaluation Protocol=Zero-shot2025.12 | 75.5 | |
| FFTModel Scale=13B, Evaluation Protocol=Zero-shot2025.12 | 73.7 |