Knowledge Editing on MQuAKE-CF 4-hop v2
90H-AccCODE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CODEModel=Qwen-2.5-7B-Ins, Augmentation=true2026.05 | 90 | 83.5 | 3.6 | |
| CODEModel=Qwen-2.5-7B-Ins, Augmentation=false2026.05 | 88.9 | 75.5 | 4.4 | |
| CODEModel=LLaMA-3.1-8B-Ins, Augmentation=false2026.05 | 84.1 | 61 | 8.6 | |
| CaKEModel=Qwen-2.5-7B-Ins, Augmentation=true2026.05 | 83.9 | 87.5 | 32.1 | |
| CODEModel=LLaMA-3.1-8B-Ins, Augmentation=true2026.05 | 82.4 | 78 | 5.8 | |
| CaKEModel=LLaMA-3.1-8B-Ins, Augmentation=true2026.05 | 71.1 | 72.5 | 13.9 | |
| MEMITModel=Qwen-2.5-7B-Ins, Augmentation=false2026.05 | 46.3 | 18.5 | 33.5 | |
| EMMETModel=LLaMA-3.1-8B-Ins, Augmentation=false2026.05 | 45.9 | 7 | 41.9 | |
| EMMETModel=Qwen-2.5-7B-Ins, Augmentation=false2026.05 | 40.6 | 19 | 37.7 | |
| AlphaEditModel=LLaMA-3.1-8B-Ins, Augmentation=false2026.05 | 36.8 | 5 | 37.9 | |
| MEMITModel=LLaMA-3.1-8B-Ins, Augmentation=false2026.05 | 32.8 | 3.5 | 40.8 | |
| AdaLoRAModel=LLaMA-3.1-8B-Ins, Augmentation=false2026.05 | 26 | 10 | 57.6 | |
| AdaLoRAModel=Qwen-2.5-7B-Ins, Augmentation=false2026.05 | 21.6 | 8 | 42 | |
| AlphaEditModel=Qwen-2.5-7B-Ins, Augmentation=false2026.05 | 19.8 | 13.5 | 44.7 | |
| Pre-editedModel=LLaMA-3.1-8B-Ins, Augmentation=false2026.05 | 19.7 | 2.5 | — | |
| WISEModel=Qwen-2.5-7B-Ins, Augmentation=false2026.05 | 19.2 | 10.2 | 50 | |
| Pre-editedModel=Qwen-2.5-7B-Ins, Augmentation=false2026.05 | 18.5 | 5 | — | |
| WISEModel=LLaMA-3.1-8B-Ins, Augmentation=false2026.05 | 17.9 | 7.5 | 30.8 |