Knowledge Editing on Bi-ZsRE 1600 edits (test)
86.49EfficacyORE
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| OREModel=Qwen2.5-7B2026.06 | 86.49 | 83.45 | 31.04 | 66.99 | |
| OREModel=LLaMA-3-8B2026.06 | 85.1 | 81.84 | 29.14 | 65.36 | |
| AlphaEditModel=Qwen2.5-7B2026.06 | 82.28 | 78.49 | 39.81 | 66.86 | |
| MEMITModel=Qwen2.5-7B2026.06 | 79.45 | 76.52 | 40.77 | 65.58 | |
| MEMITModel=LLaMA-3-8B2026.06 | 76.93 | 74.55 | 31.4 | 60.96 | |
| RECTModel=Qwen2.5-7B2026.06 | 76.36 | 72.23 | 41 | 63.2 | |
| AlphaEditModel=LLaMA-3-8B2026.06 | 76.14 | 73.3 | 31.23 | 60.22 | |
| RECTModel=LLaMA-3-8B2026.06 | 67.78 | 65.06 | 31.39 | 54.74 | |
| PRUNEModel=Qwen2.5-7B2026.06 | 67.39 | 64.91 | 43.07 | 58.46 | |
| ReFTModel=LLaMA-3-8B2026.06 | 53.67 | 52.29 | 24.49 | 43.48 | |
| PRUNEModel=LLaMA-3-8B2026.06 | 53.09 | 50.59 | 23.12 | 42.27 | |
| ReFTModel=Qwen2.5-7B2026.06 | 52.89 | 50.37 | 35.27 | 46.18 | |
| NSEModel=Qwen2.5-7B2026.06 | 52.54 | 51.27 | 38.79 | 47.53 | |
| NSEModel=LLaMA-3-8B2026.06 | 49.46 | 48.51 | 30.76 | 42.91 | |
| FTModel=LLaMA-3-8B2026.06 | 48.52 | 45.06 | 22.47 | 38.68 | |
| FTModel=Qwen2.5-7B2026.06 | 47.25 | 42.39 | 33.32 | 40.99 | |
| ROMEModel=LLaMA-3-8B2026.06 | 42.08 | 40.8 | 30.18 | 37.69 | |
| ROMEModel=Qwen2.5-7B2026.06 | 37.02 | 36.66 | 39.72 | 37.8 |