Model Editing on UnKE
94.09EfficacyUltraEdit
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| UltraEditBase Model=LLaMA-3-8B-Instruct, Edited Instances=10002025.05 | 94.09 | 92.68 | 73.64 | 85.84 | 50.92 | |
| UltraEditBase Model=LLaMA-3-8B-Instruct, Edited Instances=5002025.05 | 93.05 | 91.43 | 65.62 | 84.29 | 45.33 | |
| RLEditBase Model=Qwen2.5-7B-Instruct, Edited Instances=5002025.05 | 92.65 | 91.59 | 63.17 | 84.03 | 51.68 | |
| RLEditBase Model=LLaMA-3-8B-Instruct, Edited Instances=5002025.05 | 91.91 | 91.03 | 62.23 | 85.76 | 56.78 | |
| UltraEditBase Model=Qwen2.5-7B-Instruct, Edited Instances=10002025.05 | 88.84 | 87.01 | 72.17 | 82.02 | 36.31 | |
| UltraEditBase Model=Qwen2.5-7B-Instruct, Edited Instances=5002025.05 | 86.39 | 84.33 | 61.11 | 81.14 | 32.22 | |
| WISEBase Model=LLaMA-3-8B-Instruct, Edited Instances=5002025.05 | 83.95 | 82.93 | 58.08 | 80.43 | 32.02 | |
| AlphaEditBase Model=LLaMA-3-8B-Instruct, Edited Instances=5002025.05 | 71.07 | 68.31 | 69.17 | 81.39 | 37.37 | |
| FTBase Model=LLaMA-3-8B-Instruct, Edited Instances=5002025.05 | 67.09 | 66.39 | 67.31 | 64.65 | 6.89 | |
| FTBase Model=Qwen2.5-7B-Instruct, Edited Instances=5002025.05 | 45.72 | 45.44 | 32.55 | 47.67 | 12.12 | |
| AlphaEditBase Model=Qwen2.5-7B-Instruct, Edited Instances=5002025.05 | 39.98 | 37.78 | 25.14 | 64.72 | 23.36 |