Sequential Knowledge Editing on ZsRE (10,000 Samples)
97.8Efficacy Success (Eff)REVIVE
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| REVIVEBase Model=GPT-J, Baseline=NSE2026.01 | 97.8 | 91.75 | 26.84 | |
| REVIVEBase Model=GPT-J, Baseline=AlphaEdit2026.01 | 97.53 | 91.33 | 23.4 | |
| REVIVEBase Model=LLaMA3, Baseline=NSE2026.01 | 94.37 | 90.57 | 32.17 | |
| REVIVEBase Model=LLaMA3, Baseline=AlphaEdit2026.01 | 93.4 | 89.31 | 31.72 | |
| AlphaEditBase Model=LLaMA32026.01 | 90.57 | 85.66 | 30.5 | |
| REVIVEBase Model=GPT-J, Baseline=MEMIT2026.01 | 88.88 | 83.22 | 27.87 | |
| AlphaEditBase Model=GPT-J2026.01 | 87.84 | 78.65 | 22.31 | |
| REVIVEBase Model=LLaMA3, Baseline=RECT2026.01 | 84.2 | 80.27 | 31.92 | |
| REVIVEBase Model=LLaMA3, Baseline=MEMIT2026.01 | 83.45 | 79.9 | 32.01 | |
| REVIVEBase Model=GPT-J, Baseline=RECT2026.01 | 81.28 | 74.78 | 28.2 | |
| NSE+REVIVEModel=GPT2-XL, Enhanced with=REVIVE2026.01 | 77.83 | 70.55 | 24.84 | |
| MEMIT+REVIVEModel=GPT2-XL, Enhanced with=REVIVE2026.01 | 68.2 | 60.8 | 27.09 | |
| AlphaEdit+REVIVEModel=GPT2-XL, Enhanced with=REVIVE2026.01 | 68.1 | 57.17 | 20.35 | |
| REVIVEBase Model=GPT-J, Baseline=PRUNE2026.01 | 63.08 | 58.9 | 26.03 | |
| RECT+REVIVEModel=GPT2-XL, Enhanced with=REVIVE2026.01 | 62.45 | 55.17 | 26.2 | |
| REVIVEBase Model=LLaMA3, Baseline=PRUNE2026.01 | 56.61 | 53.3 | 27.74 | |
| MEMITModel=GPT2-XL2026.01 | 53 | 46.27 | 12.76 | |
| AlphaEditModel=GPT2-XL2026.01 | 53 | 46.27 | 12.76 | |
| NSEBase Model=LLaMA32026.01 | 45.61 | 45.04 | 31.27 | |
| NSEBase Model=GPT-J2026.01 | 44.03 | 42.39 | 24.86 | |
| PRUNE+REVIVEModel=GPT2-XL, Enhanced with=REVIVE2026.01 | 40.92 | 37.61 | 25.29 | |
| LLaMA3Model=LLaMA3, Status=Base Model2026.01 | 35.67 | 34.81 | 31.83 | |
| NSEModel=GPT2-XL2026.01 | 33.71 | 32.31 | 22.7 | |
| RECTModel=GPT2-XL2026.01 | 29.8 | 27.17 | 6.94 | |
| GPT-JModel=GPT-J, Status=Base Model2026.01 | 26.45 | 25.74 | 27.04 | |
| RECTBase Model=GPT-J2026.01 | 23.6 | 22.02 | 12.44 | |
| GPT2-XLModel=GPT2-XL2026.01 | 22.17 | 21.28 | 24.2 | |
| PRUNEModel=GPT2-XL2026.01 | 0.21 | 0.19 | 2.06 | |
| MEMITBase Model=GPT-J2026.01 | 0.1 | 0.1 | 0.17 | |
| MEMITBase Model=LLaMA32026.01 | 0.08 | 0.08 | 1.36 | |
| PRUNEBase Model=GPT-J2026.01 | 0.03 | 0.02 | 0.05 | |
| PRUNEBase Model=LLaMA32026.01 | 0 | 0 | 0.08 | |
| RECTBase Model=LLaMA32026.01 | 0 | 0 | 0 |