Instruction Following on Instruction-following dataset
78.57Clean F1Syns
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| SynsModel=Vicuna, Evaluation Protocol=After Clean SFT2026.05 | 78.57 | 21.35 | 94.37 | 85 | |
| ParaesthesiaModel=Llama 2, Evaluation Protocol=After Clean SFT2026.05 | 73.49 | 19 | 97.72 | 94.1 | |
| SleeperModel=Vicuna, Evaluation Protocol=After Clean SFT2026.05 | 73.37 | 17.55 | 93.23 | 75.1 | |
| SyndModel=Vicuna, Evaluation Protocol=After Clean SFT2026.05 | 73.32 | 17.85 | 94.55 | 87.25 | |
| SyndModel=Llama 2, Evaluation Protocol=After Clean SFT2026.05 | 73.29 | 17.65 | 94.8 | 85.45 | |
| VPIModel=Vicuna, Evaluation Protocol=After Clean SFT2026.05 | 73.23 | 17.25 | 83.8 | 59.55 | |
| BadNetsModel=Vicuna, Evaluation Protocol=After Clean SFT2026.05 | 73.09 | 17 | 46.06 | 0 | |
| ParaesthesiaModel=Vicuna, Evaluation Protocol=After Clean SFT2026.05 | 73.04 | 18.55 | 95.22 | 90.4 | |
| CBAModel=Vicuna, Evaluation Protocol=After Clean SFT2026.05 | 73.03 | 18.41 | 84.4 | 69.6 | |
| CBAModel=Llama 2, Evaluation Protocol=After Clean SFT2026.05 | 73.02 | 18.16 | 86.65 | 74.25 | |
| VPIModel=Llama 2, Evaluation Protocol=After Clean SFT2026.05 | 72.96 | 16.65 | 79.59 | 61.35 | |
| BadNetsModel=Llama 2, Evaluation Protocol=After Clean SFT2026.05 | 72.93 | 17.35 | 45.99 | 0 | |
| SleeperModel=Llama 2, Evaluation Protocol=After Clean SFT2026.05 | 72.88 | 17.15 | 94.8 | 80.85 | |
| SynsModel=Llama 2, Evaluation Protocol=After Clean SFT2026.05 | 72.82 | 17.45 | 48.02 | 0.55 |