Model Editing on CounterFact 3,000 samples (test)
9,980ReliabilityCrispEdit
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CrispEditbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing2026.02 | 9,980 | 73 | 55.2 | |
| LocBF-FTbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing2026.02 | 9,970 | 72.7 | 44.6 | |
| FTbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing2026.02 | 9,970 | 82 | 47.9 | |
| Adam-NSCLbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing2026.02 | 9,950 | 81.5 | 47.7 | |
| CrispEdit-Seqbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing, editing_mode=sequential2026.02 | 9,950 | 62.9 | 52.3 | |
| AlphaEditbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing2026.02 | 9,430 | 72 | 69.1 | |
| LoRAbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing2026.02 | 9,380 | 17.8 | 42.9 | |
| FT Sequentialbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing, editing_mode=sequential2026.02 | 7,860 | 59.8 | 51.6 | |
| UltraEditbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing2026.02 | 2,800 | 18.7 | 51.4 | |
| LoRA Sequentialbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing, editing_mode=sequential2026.02 | 2,090 | 10.4 | 57 | |
| Llama 3 8B Instructbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing2026.02 | 90 | 1.2 | 89.4 | |
| MEMITbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing2026.02 | 0 | 0 | 49.4 | |
| MENDbackbone=LLaMA-3-8B-Instruct, pipeline=teacher-forcing2026.02 | 0 | 0 | 0 |