Knowledge Editing on Bi-ZsRE Edit in Chinese (test)
93.49Reliability (En Test)X-KDE
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| X-KDEBackbone=Llama2-chat-7B2025.02 | 93.49 | 92.22 | 90.56 | 65.55 | 100 | 99.11 | 92.85 | 74.14 | 88.49 | |
| IKEBackbone=Llama2-chat-7B2025.02 | 65.87 | 65.74 | 69.41 | 63.06 | 99.86 | 99.73 | 64.86 | 72.39 | 80.79 | |
| LTEBackbone=Llama2-chat-7B2025.02 | 64.63 | 62.56 | 85.23 | 62.6 | 99.79 | 99.31 | 87.17 | 69.69 | 78.87 | |
| FT-MBackbone=Llama2-chat-7B2025.02 | 51.86 | 51.24 | 98.18 | 55.3 | 100 | 99.71 | 79.28 | 61.98 | 74.69 | |
| MEMITBackbone=Llama2-chat-7B2025.02 | 45.37 | 44.95 | 99.07 | 54.65 | 75.19 | 73.45 | 96.02 | 51.44 | 67.52 | |
| ROMEBackbone=Llama2-chat-7B2025.02 | 44.14 | 43.8 | 97.92 | 52.66 | 72.24 | 70.12 | 96.48 | 48.15 | 65.69 | |
| FT-LBackbone=Llama2-chat-7B2025.02 | 40.8 | 40.66 | 94.8 | 55.24 | 54.72 | 53.68 | 66.51 | 48.75 | 56.89 |