Knowledge Editing on Temporal τ=75 GPT-J-6B (OOD)
98ReliabilityWISE
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| WISEBase Model=GPT-J-6B2026.06 | 98 | 38 | 99 | 78 | |
| LOKIBase Model=GPT-J-6B2026.06 | 98 | 36 | 81 | 72 | |
| FTBase Model=GPT-J-6B2026.06 | 93 | 32 | 62 | 62 | |
| AlphaEditBase Model=GPT-J-6B2026.06 | 87 | 28 | 97 | 71 | |
| w/o EditingBase Model=GPT-J-6B2026.06 | 56 | 21 | — | — | |
| GRACEBase Model=GPT-J-6B2026.06 | 56 | 22 | 100 | 59 | |
| DEFERBase Model=GPT-J-6B2026.06 | 55 | 21 | 95 | 57 | |
| MEMITBase Model=GPT-J-6B2026.06 | 43 | 14 | 49 | 35 | |
| ROMEBase Model=GPT-J-6B2026.06 | 2 | 0 | 2 | 1 |