Machine Unlearning on TOFU 5% forget set
0FQGT
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GTPre-trained model=Llama-3.2 8B2026.01 | 0 | 0.632 | 0.074 | 0.992 | 0.918 | 0.886 | |
| PerTA-gradPre-trained model=Llama-3.2 8B2026.01 | -4.529 | 0.659 | 0.164 | 0.882 | 0.718 | 0.895 | |
| NPO+Pre-trained model=Llama-3.2 8B2026.01 | -7.888 | 0.569 | 0.16 | 0.521 | 0.361 | 0.914 | |
| GDPre-trained model=Llama-3.2 8B2026.01 | -10.225 | 0.509 | 0.158 | 0.397 | 0.239 | 0.811 | |
| NPOPre-trained model=Llama-3.2 8B2026.01 | -11.183 | 0.131 | 0.033 | 0.037 | 0.004 | 0.141 | |
| FullPre-trained model=Llama-3.2 8B2026.01 | -12.184 | 0.628 | 0.972 | 0.992 | 0.02 | 0.852 | |
| GAPre-trained model=Llama-3.2 8B2026.01 | -118.712 | 0 | 0.033 | 0.035 | 0.002 | 0.038 |