Knowledge Editing on Counterfact
9,387EfficacyEAMET
Evaluation results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| EAMETModel=LLaMA3-8B2025.05 | 9,387 | 6,374 | 79.07 | 533.3 | — | — | — | — | — | — | — | — | — | |
| MEMITModel=LLaMA3-8B2025.05 | 9,376 | 6,198 | 77.69 | 526.47 | — | — | — | — | — | — | — | — | — | |
| EAMETModel=LLaMA2-13B2025.05 | 9,285 | 6,008 | 77.51 | 530.78 | — | — | — | — | — | — | — | — | — | |
| EAMETModel=Falcon-7B2025.05 | 9,237 | 6,391 | 78.94 | 528.98 | — | — | — | — | — | — | — | — | — | |
| EAMETModel=Qwen2.5-7B2025.05 | 9,049 | 6,437 | 72.18 | 536.67 | — | — | — | — | — | — | — | — | — | |
| MEMITModel=Qwen2.5-7B2025.05 | 9,006 | 6,386 | 70.53 | 529.27 | — | — | — | — | — | — | — | — | — | |
| EAMETModel=Deepseek-7B2025.05 | 8,974 | 5,998 | 77.73 | 513.93 | — | — | — | — | — | — | — | — | — | |
| MEMITModel=Falcon-7B2025.05 | 8,921 | 6,085 | 77.56 | 519.92 | — | — | — | — | — | — | — | — | — | |
| EAMETModel=LLaMA2-7B2025.05 | 8,909 | 6,121 | 72.19 | 519.89 | — | — | — | — | — | — | — | — | — | |
| ALPHAEDITModel=Falcon-7B2025.05 | 8,762 | 5,832 | 72.43 | 500.35 | — | — | — | — | — | — | — | — | — | |
| ALPHAEDITModel=Qwen2.5-7B2025.05 | 8,315 | 5,570 | 67.16 | 514.07 | — | — | — | — | — | — | — | — | — | |
| PMETModel=LLaMA2-13B2025.05 | 7,860 | 3,876 | 81.15 | 526.82 | — | — | — | — | — | — | — | — | — | |
| PMETModel=LLaMA3-8B2025.05 | 7,771 | 4,941 | 71.43 | 510.82 | — | — | — | — | — | — | — | — | — | |
| PMETModel=Falcon-7B2025.05 | 7,761 | 5,703 | 70.48 | 517.09 | — | — | — | — | — | — | — | — | — | |
| PMETModel=Deepseek-7B2025.05 | 7,452 | 4,349 | 79.01 | 514.58 | — | — | — | — | — | — | — | — | — | |
| PMETModel=LLaMA2-7B2025.05 | 7,422 | 4,645 | 72.47 | 507.1 | — | — | — | — | — | — | — | — | — | |
| PMETModel=Qwen2.5-7B2025.05 | 6,571 | 5,284 | 63.14 | 518.92 | — | — | — | — | — | — | — | — | — | |
| MEMITModel=Deepseek-7B2025.05 | 6,211 | 4,201 | 78.04 | 512.16 | — | — | — | — | — | — | — | — | — | |
| ALPHAEDITModel=LLaMA3-8B2025.05 | 5,897 | 3,302 | 85.16 | 537.91 | — | — | — | — | — | — | — | — | — | |
| ROMEModel=LLaMA3-8B2025.05 | 5,101 | 3,332 | 64.37 | 491.98 | — | — | — | — | — | — | — | — | — | |
| MEMITModel=LLaMA2-13B2025.05 | 4,798 | 3,475 | 71.61 | 517.63 | — | — | — | — | — | — | — | — | — | |
| MEMITModel=LLaMA2-7B2025.05 | 2,495 | 2,268 | 63.84 | 506.69 | — | — | — | — | — | — | — | — | — | |
| ALPHAEDITModel=Deepseek-7B2025.05 | 2,251 | 1,400 | 59.92 | 479.52 | — | — | — | — | — | — | — | — | — | |
| FTModel=Qwen2.5-7B2025.05 | 1,618 | 1,415 | 56.07 | 527.56 | — | — | — | — | — | — | — | — | — | |
| FTModel=Falcon-7B2025.05 | 1,470 | 1,354 | 56.34 | 167.18 | — | — | — | — | — | — | — | — | — | |
| ROMEModel=Falcon-7B2025.05 | 1,285 | 1,256 | 51.48 | 353.38 | — | — | — | — | — | — | — | — | — | |
| ROMEModel=LLaMA2-13B2025.05 | 405 | 152 | 50.44 | 525.12 | — | — | — | — | — | — | — | — | — | |
| ALPHAEDITModel=LLaMA2-13B2025.05 | 303 | 190 | 54.97 | 421.97 | — | — | — | — | — | — | — | — | — | |
| FTModel=LLaMA3-8B2025.05 | 268 | 130 | 58.16 | 434.67 | — | — | — | — | — | — | — | — | — | |
| FTModel=Deepseek-7B2025.05 | 261 | 249 | 81.43 | 519.35 | — | — | — | — | — | — | — | — | — | |
| FTModel=LLaMA2-13B2025.05 | 123 | 7 | 68.57 | 484.56 | — | — | — | — | — | — | — | — | — | |
| IKEModel=GPT-J2026.01 | 100 | — | 89.96 | — | 76.51 | 97.3 | — | — | — | — | — | — | — | |
| IKEModel=Qwen-7B2026.01 | 100 | — | 89.08 | — | 73.79 | 98.75 | — | — | — | — | — | — | — | |
| P-TokensModel=Qwen-7B2026.01 | 100 | — | 93.88 | — | 83.96 | 99.55 | — | — | — | — | — | — | — | |
| baselineModel=Llama32026.01 | 100 | — | 72.42 | — | 49.43 | 89.35 | — | — | — | — | — | — | — | |
| IKEModel=Llama32026.01 | 100 | — | 93.97 | — | 84.4 | 99.25 | — | — | — | — | — | — | — | |
| P-TokensModel=Llama32026.01 | 100 | — | 95.82 | — | 88.62 | 99.75 | — | — | — | — | — | — | — | |
| IKEModel=Qwen-14B2026.01 | 100 | — | 92.84 | — | 81.68 | 99.3 | — | — | — | — | — | — | — | |
| P-TokensModel=Qwen-14B2026.01 | 100 | — | 94.72 | — | 85.79 | 99.85 | — | — | — | — | — | — | — | |
| ROMELayer(s)=52026.01 | 100 | 99.5 | 76.3 | 620.9 | 78.8 | — | 90.9 | — | — | — | — | — | — | |
| ROMELayer(s)=102026.01 | 100 | 98.5 | 69.3 | 617.9 | 79.2 | — | 89.3 | — | — | — | — | — | — | |
| Redundant Editing (ROME)Layer(s)=5,152026.01 | 100 | 100 | 58.8 | 603.9 | 71 | — | 85.4 | — | — | — | — | — | — | |
| Redundant Editing (ROME)Layer(s)=5,202026.01 | 100 | 99 | 60.8 | 515.4 | 49.4 | — | 78.4 | — | — | — | — | — | — | |
| Redundant Editing (ROME)Layer(s)=5,10,202026.01 | 100 | 100 | 50.8 | 461.8 | 44.1 | — | 73.8 | — | — | — | — | — | — | |
| Redundant Editing (ROME)Layer(s)=5,10,15,202026.01 | 100 | 100 | 38.8 | 387.7 | 34.2 | — | 67.1 | — | — | — | — | — | — | |
| Redundant Editing (ROME)Layer(s)=5,9,13,17,202026.01 | 100 | 99.5 | 16.8 | 255.2 | 17.1 | — | 54.9 | — | — | — | — | — | — | |
| Redundant Editing (ROME)Layer(s)=5,8,11,15,17,202026.01 | 100 | 99.5 | 17 | 289.8 | 19.7 | — | 56.6 | — | — | — | — | — | — | |
| FTModel=GPT2-XL (1.5B)2025.05 | 100 | — | — | — | — | — | — | — | — | 89.1 | 64.4 | 39.5 | — | |
| FTModel=GPT-J (6B)2025.05 | 100 | — | — | — | — | — | — | — | — | 95.8 | 24.4 | 9.8 | — | |
| FTModel=LLaMA3 (8B)2025.05 | 100 | — | — | — | — | — | — | — | — | 97.7 | 18.4 | 7 | — | |
| MENDModel=LLaMA3 (8B)2025.05 | 100 | — | — | — | — | — | — | — | — | 73.6 | 58.4 | 36 | — | |
| ROMEModel=LLaMA3 (8B)2025.05 | 100 | — | — | — | — | — | — | — | — | 98.9 | 90.8 | 77.3 | — | |
| AlphaEditModel=LLaMA3 (8B)2025.05 | 100 | — | — | — | — | — | — | — | — | 93.5 | 89.7 | 78.5 | — | |
| FTModel=Qwen2.5-7B, Batch Edits=50002026.06 | 99.98 | 75.6 | 38.83 | — | — | — | — | — | — | — | 71.47 | — | — | |
| EVK-EditModel=GPT-J2026.02 | 99.9 | 94.35 | 80.52 | 620.54 | 40.73 | — | — | — | — | — | — | — | — | |
| P-TokensModel=GPT-J2026.01 | 99.9 | — | 92.39 | — | 81.3 | 98.4 | — | — | — | — | — | — | — | |
| baselineModel=Qwen-14B2026.01 | 99.9 | — | 63.86 | — | 38.56 | 90.65 | — | — | — | — | — | — | — | |
| SUITBackbone=LLaMA3, Batch size=102025.09 | 99.9 | 89.8 | 71.9 | — | — | — | — | 85.6 | — | — | — | — | — | |
| ROMEModel=GPT-J (6B)2025.05 | 99.9 | — | — | — | — | — | — | — | — | 99.7 | 90.3 | 75.9 | — | |
| DiKEModel=LLaMA3 (8B)2025.05 | 99.9 | — | — | — | — | — | — | — | — | 96.6 | 92.4 | 82.8 | — | |
| EVK-EditModel=GPT2-XL2026.02 | 99.8 | 94.05 | 72.31 | 611.33 | 40.15 | — | — | — | — | — | — | — | — | |
| baselineModel=Qwen-7B2026.01 | 99.8 | — | 62.07 | — | 36.7 | 90.4 | — | — | — | — | — | — | — | |
| DiKEModel=GPT-J (6B)2025.05 | 99.8 | — | — | — | — | — | — | — | — | 96.1 | 90.8 | 79.3 | — | |
| MEMITModel=LLaMA3 (8B)2025.05 | 99.8 | — | — | — | — | — | — | — | — | 94.6 | 91.1 | 80.9 | — | |
| AlphaEditModel=Qwen2.5-7B2026.06 | 99.8 | 95.8 | 82.88 | — | — | — | — | — | — | — | 92.83 | — | — | |
| AlphaEditModel=Qwen2.5-7B, Batch Edits=50002026.06 | 99.76 | 94.73 | 79.16 | — | — | — | — | — | — | — | 91.22 | — | — | |
| FTModel=LLaMA-3-8B2026.06 | 99.75 | 88.65 | 39.62 | — | — | — | — | — | — | — | 76.01 | — | — | |
| FTModel=Qwen2.5-7B2026.06 | 99.75 | 72.32 | 40.22 | — | — | — | — | — | — | — | 70.76 | — | — | |
| SAKEModel=GPT-J (6B)2026.06 | 99.71 | 96.47 | 76.29 | 620.25 | 42.12 | — | — | — | — | — | — | — | — | |
| SUITBackbone=LLaMA3, Batch size=1002025.09 | 99.7 | 90.3 | 74.2 | 631.2 | 38.2 | — | — | 86.8 | 63 | — | — | — | — | |
| SUITModel=LLaMA3, Batch size=1002025.09 | 99.7 | 90.3 | 74.2 | 631.2 | 38.2 | — | — | 86.8 | 61.8 | — | — | — | — | |
| ROMEModel=GPT2-XL (1.5B)2025.05 | 99.7 | — | — | — | — | — | — | — | — | 97.3 | 87.9 | 72.4 | — | |
| AlphaEditBLUEModel=GPT-J (6B)2026.06 | 99.62 | 96.42 | 75.58 | 618.29 | 42.21 | — | — | — | — | — | — | — | — | |
| AlphaEditModel=GPT2-XL2026.02 | 99.6 | 95.1 | 70.08 | 607.1 | 40.28 | — | — | — | — | — | — | — | — | |
| SUITBackbone=LLaMA3, Batch size=12025.09 | 99.6 | 89.5 | 71.9 | — | — | — | — | 85.4 | — | — | — | — | — | |
| MEMITModel=GPT-J (6B)2025.05 | 99.6 | — | — | — | — | — | — | — | — | 95.2 | 90.4 | 79.2 | — | |
| AlphaEditModel=GPT-J (6B)2025.05 | 99.6 | — | — | — | — | — | — | — | — | 96.9 | 91 | 79.3 | — | |
| PRUNEModel=Qwen2.5-7B2026.06 | 99.55 | 98.15 | 72.67 | — | — | — | — | — | — | — | 90.12 | — | — | |
| AlphaEditChainModel=GPT-J (6B)2026.06 | 99.53 | 96.34 | 74.45 | 616.72 | 42.09 | — | — | — | — | — | — | — | — | |
| RECTModel=GPT-J2026.02 | 99.5 | 95.25 | 72.8 | 611.89 | 41.47 | — | — | — | — | — | — | — | — | |
| AlphaEditModel=GPT-J2026.02 | 99.5 | 95.5 | 78.64 | 619.8 | 42.07 | — | — | — | — | — | — | — | — | |
| SUITModel=Qwen2.5, Batch size=1002025.09 | 99.5 | 86.8 | 74.4 | 626.2 | 37.4 | — | — | 85.7 | 23.7 | — | — | — | — | |
| DiKEModel=GPT2-XL (1.5B)2025.05 | 99.5 | — | — | — | — | — | — | — | — | 95 | 87.7 | 73.3 | — | |
| MEMITModel=Qwen2.5-7B2026.06 | 99.5 | 92.45 | 83.61 | — | — | — | — | — | — | — | 91.85 | — | — | |
| MEMITModel=Qwen2.5-7B, Batch Edits=50002026.06 | 99.42 | 92.39 | 81.2 | — | — | — | — | — | — | — | 91 | — | — | |
| AlphaEditModel=GPT2-XL (1.5B)2025.05 | 99.4 | — | — | — | — | — | — | — | — | 96.2 | 88.3 | 74.1 | — | |
| SUITBackbone=Qwen2.5, Batch size=12025.09 | 99.3 | 90.3 | 72 | — | — | — | — | 85.6 | — | — | — | — | — | |
| SUITBackbone=Qwen2.5, Batch size=102025.09 | 99.3 | 90.8 | 71.9 | — | — | — | — | 85.7 | — | — | — | — | — | |
| MEMITModel=GPT2-XL (1.5B)2025.05 | 99.3 | — | — | — | — | — | — | — | — | 93.2 | 87.4 | 73.9 | — | |
| PRISM-∆Model=QWEN3-8B2026.03 | 99.24 | — | — | — | — | — | — | — | — | 99.1 | — | — | — | |
| AlphaEditModel=GPT-J (6B)2026.06 | 99.23 | 95.94 | 73.63 | 611.85 | 40.63 | — | — | — | — | — | — | — | — | |
| SUITModel=GPT-J (6B)2026.06 | 99.21 | 95.25 | 75.98 | 617.32 | 42.34 | — | — | — | — | — | — | — | — | |
| MEMITModel=GPT-J2026.02 | 99.2 | 95.3 | 70.82 | 610.86 | 41.73 | — | — | — | — | — | — | — | — | |
| AlphaEditBackbone=GPT-J, Batch size=102025.09 | 99.2 | 96 | 53.1 | — | — | — | — | 76.3 | — | — | — | — | — | |
| SUITBackbone=GPT-J, Batch size=1002025.09 | 99.2 | 94.6 | 67.7 | 619.4 | 49.4 | — | — | 84.7 | 20.4 | — | — | — | — | |
| SUITBackbone=Qwen2.5, Batch size=1002025.09 | 99.2 | 91.6 | 72.3 | 626.2 | 37.4 | — | — | 86.1 | 30.8 | — | — | — | — | |
| OREModel=Qwen2.5-7B2026.06 | 99.18 | 95.5 | 84.73 | — | — | — | — | — | — | — | 93.14 | — | — | |
| GLAMEModel=GPT-J (6B)2026.06 | 99.15 | 95.78 | 74.92 | 614.1 | 41.45 | — | — | — | — | — | — | — | — | |
| PRISM-∆Model=QWEN3-4B2026.03 | 99.14 | — | — | — | — | — | — | — | — | 98.52 | — | — | — | |
| CoT2EditModel=Falcon3-10B2026.04 | 99.13 | — | — | — | — | — | — | — | — | 89.54 | — | 94.55 | — |