Unlearning on TOFU forget10 (forget)
0.954rLBase
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BaseBackbone=Dream-7B-Instruct2026.05 | 0.954 | 0.865 | |
| BaseBackbone=LLaDA-8B-Instruct2026.05 | 0.884 | 0.38 | |
| GDBackbone=LLaDA-8B-Instruct2026.05 | 0.533 | 0.061 | |
| DPOBackbone=Dream-7B-Instruct2026.05 | 0.517 | 0.341 | |
| SimNPOBackbone=LLaDA-8B-Instruct2026.05 | 0.485 | 0.036 | |
| DPOBackbone=LLaDA-8B-Instruct2026.05 | 0.479 | 0.168 | |
| GABackbone=Dream-7B-Instruct2026.05 | 0.456 | 0.079 | |
| GDBackbone=Dream-7B-Instruct2026.05 | 0.426 | 0.002 | |
| NPOBackbone=LLaDA-8B-Instruct2026.05 | 0.372 | 0.009 | |
| WGABackbone=Dream-7B-Instruct2026.05 | 0.37 | 0.051 | |
| NPOBackbone=Dream-7B-Instruct2026.05 | 0.361 | 0.031 | |
| GABackbone=LLaDA-8B-Instruct2026.05 | 0.348 | 0.02 | |
| SimNPOBackbone=Dream-7B-Instruct2026.05 | 0.297 | 0.002 | |
| MDUBackbone=Dream-7B-Instruct, τ=0.752026.05 | 0.284 | 0.001 | |
| MDUBackbone=Dream-7B-Instruct, τ=0.252026.05 | 0.277 | 0.001 | |
| MDUBackbone=Dream-7B-Instruct, τ=1.002026.05 | 0.199 | 0 | |
| MDUBackbone=Dream-7B-Instruct, τ=0.002026.05 | 0.191 | 0 | |
| MDUBackbone=Dream-7B-Instruct, τ=0.502026.05 | 0.158 | 0 | |
| MDUBackbone=LLaDA-8B-Instruct, τ=0.252026.05 | 0.135 | 0 | |
| WGABackbone=LLaDA-8B-Instruct2026.05 | 0.122 | 0.01 | |
| MDUBackbone=LLaDA-8B-Instruct, τ=0.502026.05 | 0.098 | 0.001 | |
| MDUBackbone=LLaDA-8B-Instruct, τ=0.752026.05 | 0.078 | 0.04 | |
| MDUBackbone=LLaDA-8B-Instruct, τ=0.002026.05 | 0.069 | 0 | |
| MDUBackbone=LLaDA-8B-Instruct, τ=1.002026.05 | 0.034 | 0.074 |