Machine Unlearning on WMDP-Cyber (MMLU, Forget, Relearn, Delta)
75.8MMLU ScoreMLP Breaking + CIR + MCU
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MLP Breaking + CIR + MCUModel=Qwen3-8B2026.05 | 75.8 | 29 | 36.4 | 7.4 | |
| MLP Breaking + CIRModel=Qwen3-8B2026.05 | 75.7 | 28 | 41.7 | 13.7 | |
| MLP Breaking + CIR + MCUModel=Gemma2-9B2026.05 | 69.9 | 22.4 | 24.3 | 1.9 | |
| MLP Breaking + CIRModel=Gemma2-9B2026.05 | 69.8 | 26.3 | 30.3 | 4 |