Sequential Knowledge Editing on CounterFact top
93.87EfficacyDeltaEdit
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DeltaEditModel=Llama3-8B2025.05 | 93.87 | 56.52 | 34.2 | |
| DeltaEditModel=GPT2-XL2025.05 | 93.8 | 54.37 | 53.28 | |
| AlphaEditModel=GPT2-XL2025.05 | 85.93 | 46.23 | 53.18 | |
| AlphaEditModel=Llama3-8B2025.05 | 77.07 | 53.92 | 17.59 | |
| RECTModel=GPT2-XL2025.05 | 56.77 | 31.13 | 18.11 | |
| FTModel=GPT2-XL2025.05 | 27.17 | 8.3 | 2.22 | |
| MEMITModel=GPT2-XL2025.05 | 25.73 | 16.85 | 12.71 | |
| FTModel=Llama3-8B2025.05 | 9.6 | 3.63 | 0.23 | |
| PRUNEModel=GPT2-XL2025.05 | 7.93 | 6.45 | 8.91 | |
| ROMEModel=GPT2-XL2025.05 | 0.53 | 0.38 | 0.41 | |
| ROMEModel=Llama3-8B2025.05 | 0 | 0 | 0 | |
| MEMITModel=Llama3-8B2025.05 | 0 | 0 | 0 | |
| PRUNEModel=Llama3-8B2025.05 | 0 | 0 | 0 | |
| RECTModel=Llama3-8B2025.05 | 0 | 0 | 0 |