Knowledge Unlearning on MCF
23.35SpeedBase
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| BaseModel=Llama-3.12026.05 | 23.35 | 27.7 | 22.5 | 7.48 | — | — | |
| ROMEModel=Llama-3.12026.05 | 23.17 | 27.3 | 22.7 | 7.48 | — | — | |
| BaseModel=Llama-3.12026.05 | 21.96 | — | — | 7.47 | 24.6 | 22.8 | |
| ROMEModel=Llama-3.12026.05 | 21.86 | — | — | 7.48 | 24.4 | 22.6 | |
| MEMITModel=Llama-3.12026.05 | 21.08 | — | — | 7.51 | 9.6 | 16.2 | |
| BaseModel=Llama-3.22026.05 | 20.59 | 22.1 | 19.6 | 12.88 | — | — | |
| ROMEModel=Llama-3.22026.05 | 20.47 | 21.9 | 19.55 | 12.89 | — | — | |
| AlphaEditModel=Llama-3.12026.05 | 19.74 | — | — | 7.49 | 0.2 | 7.8 | |
| BaseModel=Llama-3.22026.05 | 19.6 | — | — | 12.88 | 18.2 | 20.3 | |
| MEMITModel=Llama-3.12026.05 | 19.51 | 5.3 | 5.55 | 7.7 | — | — | |
| ROMEModel=Llama-3.22026.05 | 19.5 | — | — | 12.88 | 18.2 | 20.3 | |
| MEMITModel=Llama-3.22026.05 | 19.2 | — | — | 12.86 | 17 | 18.3 | |
| AlphaEditModel=Llama-3.22026.05 | 18.36 | — | — | 12.84 | 2.6 | 11.8 | |
| MEMITModel=Llama-3.22026.05 | 17.75 | 13.8 | 14.45 | 12.77 | — | — | |
| ZeroUnlearnModel=Llama-3.12026.05 | 16.82 | — | — | 7.77 | 0 | 4.6 | |
| AlphaEditModel=Llama-3.22026.05 | 15.68 | 1.4 | 10.2 | 12.78 | — | — | |
| AlphaEditModel=Llama-3.12026.05 | 15.46 | 0.5 | 2.85 | 8.01 | — | — | |
| ZeroUnlearnModel=Llama-3.22026.05 | 14.9 | — | — | 13.06 | 0.4 | 4.6 | |
| ROMEModel=Qwen-32026.05 | 13.41 | 14.6 | 13.15 | 11.38 | — | — | |
| ZeroUnlearn-GDModel=Llama-3.12026.05 | 13.31 | 0 | 2.3 | 7.94 | — | — | |
| BaseModel=Qwen-32026.05 | 13.3 | 14.9 | 12.85 | 11.21 | — | — | |
| ROMEModel=Qwen-32026.05 | 13.1 | 14.2 | 13.9 | 11.31 | — | — | |
| BaseModel=Qwen-32026.05 | 12.84 | 14.4 | 14.1 | 11.21 | — | — | |
| MEMITModel=Qwen-32026.05 | 12.82 | 0.8 | 3.8 | 11.23 | — | — | |
| AlphaEditModel=Qwen-32026.05 | 12.52 | 0.6 | 2.6 | 11.24 | — | — | |
| ZeroUnlearn-GDModel=Llama-3.22026.05 | 12.41 | 0 | 5.1 | 13.05 | — | — | |
| MEMITModel=Qwen-32026.05 | 11.35 | 0.7 | 4.5 | 11.54 | — | — | |
| AlphaEditModel=Qwen-32026.05 | 11.04 | 0.4 | 3.3 | 13.03 | — | — | |
| ZeroUnlearn-GDModel=Qwen-32026.05 | 9.52 | 0.4 | 3 | 11.63 | — | — | |
| ZeroUnlearnModel=Qwen-32026.05 | 9.38 | 0 | 2.2 | 11.19 | — | — | |
| FTModel=Qwen-32026.05 | 6.54 | 0 | 2.9 | 12.49 | — | — | |
| GAModel=Llama-3.22026.05 | 1.06 | — | — | 1,000 | 2 | 1.8 | |
| GAModel=Qwen-32026.05 | 0.58 | 0.8 | 0.5 | 1,000 | — | — | |
| GAModel=Llama-3.12026.05 | 0.26 | — | — | 1,000 | 1.2 | 0.9 | |
| GAModel=Qwen-32026.05 | 0 | 0 | 0 | 1,000 | — | — | |
| FTModel=Qwen-32026.05 | 0 | 0 | 0 | 714.52 | — | — | |
| GAModel=Llama-3.22026.05 | 0 | 0 | 0 | 1,000 | — | — | |
| FTModel=Llama-3.22026.05 | 0 | 0 | 0 | 63.7 | — | — | |
| FTModel=Llama-3.22026.05 | 0 | — | — | 18.25 | 0 | 0 | |
| FTModel=Llama-3.12026.05 | 0 | — | — | 10.23 | 0 | 0 | |
| GAModel=Llama-3.12026.05 | 0 | 0 | 0 | 1,000 | — | — | |
| FTModel=Llama-3.12026.05 | 0 | 0 | 0 | 275.3 | — | — |