Knowledge Editing on ZsRE (test)
0.65Normalized Editing TimeGRACE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GRACETarget Model=Llama2-7B2025.06 | 0.65 | — | |
| FT-LTarget Model=Llama2-7B2025.06 | 0.69 | — | |
| FT-LTarget Model=Mistral-7B-v0.12025.06 | 0.71 | — | |
| GRACETarget Model=Mistral-7B-v0.12025.06 | 0.72 | — | |
| FT-LTarget Model=GPT-J-6B2025.06 | 0.73 | — | |
| GRACETarget Model=GPT-J-6B2025.06 | 0.84 | — | |
| ONCEEDITTarget Model=Llama2-7B2025.06 | 1 | — | |
| ONCEEDITTarget Model=Mistral-7B-v0.12025.06 | 1 | — | |
| ONCEEDITTarget Model=GPT-J-6B2025.06 | 1 | — | |
| WISETarget Model=GPT-J-6B2025.06 | 1.26 | — | |
| WISETarget Model=Mistral-7B-v0.12025.06 | 1.33 | — | |
| WISETarget Model=Llama2-7B2025.06 | 1.35 | — | |
| DEFERTarget Model=GPT-J-6B2025.06 | 1.4 | — | |
| DEFERTarget Model=Mistral-7B-v0.12025.06 | 1.47 | — | |
| DEFERTarget Model=Llama2-7B2025.06 | 1.49 | — | |
| ROMETarget Model=Llama2-7B2025.06 | 2.29 | — | |
| ROMETarget Model=GPT-J-6B2025.06 | 2.5 | — | |
| ROMETarget Model=Mistral-7B-v0.12025.06 | 3.39 | — | |
| DEFERTarget Model=Total2025.06 | — | 4.36 | |
| FT-LTarget Model=Total2025.06 | — | 2.13 | |
| GRACETarget Model=Total2025.06 | — | 2.21 | |
| ONCEEDITTarget Model=Total2025.06 | — | 1 | |
| ROMETarget Model=Total2025.06 | — | 8.18 | |
| WISETarget Model=Total2025.06 | — | 3.94 |