Machine Unlearning on internal e-commerce benchmark large-scale seller 1065 items (Forget Set)
48.9ROUGEBaseline
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BaselineBackbone=Llama 3.1 8B2025.05 | 48.9 | 0.54 | |
| ME+GDBackbone=Llama 3.1 8B2025.05 | 48.9 | 0.54 | |
| Undial+KLBackbone=Llama 3.1 8B, variant=12025.05 | 45.6 | 0.66 | |
| GA+KLBackbone=Llama 3.1 8B2025.05 | 44 | 0.41 | |
| NPO+KLBackbone=Llama 3.1 8B2025.05 | 41 | 1 | |
| Undial+KLBackbone=Llama 3.1 8B, variant=22025.05 | 39.2 | 0.78 | |
| Unilogit+KLBackbone=Llama 3.1 8B, variant=12025.05 | 33.8 | 1.06 | |
| Unilogit+KLBackbone=Llama 3.1 8B, variant=22025.05 | 31.5 | 3.86 |