Machine Unlearning on MedMCQA QF=1000
90Forget AccuracySequential Retrain
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Sequential RetrainQF=10002026.02 | 90 | 88 | 88.67 | 0.509 | 0.982 | |
| Sequential RetrainQF=10002026.02 | 90 | 88 | 88.67 | 0.509 | 0.982 | |
| OriginQF=10002026.02 | 89 | 88 | 90 | 0.549 | 0.902 | |
| OriginQF=1000, Status=Pre-unlearning baseline2026.02 | 89 | 88 | 90 | 0.549 | 0.902 | |
| Adapter MergingQF=10002026.02 | 88.67 | 88.67 | 85.33 | 0.5087 | 0.9827 | |
| Adapter MergingQF=10002026.02 | 88.67 | 88.67 | 85.33 | 0.5087 | 0.9827 | |
| GAQF=10002026.02 | 87.33 | 88.67 | 87.67 | 0.508 | 0.984 | |
| GAQF=10002026.02 | 87.33 | 88.67 | 87.67 | 0.508 | 0.984 | |
| NPOQF=10002026.02 | 86 | 87 | 83 | 0.509 | 0.982 | |
| NPOQF=10002026.02 | 86 | 87 | 83 | 0.509 | 0.982 | |
| Sequential LoRAQF=10002026.02 | 76.67 | 87.67 | 89 | 0.508 | 0.984 | |
| Sequential LoRAQF=10002026.02 | 76.67 | 87.67 | 89 | 0.508 | 0.984 | |
| SBUQF=10002026.02 | 71.33 | 89.33 | 89 | 0.5 | 0.996 | |
| SBUQF=10002026.02 | 71.33 | 89.33 | 89 | 0.5 | 0.996 |