Knowledge Editing on UnKEBench Paraphrased questions (Para.)
96.6Bert ScoreAnyEdit
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AnyEditLLM=Llama3-8B-It2025.02 | 96.6 | 95.6 | |
| AnyEdit*LLM=Qwen2.5-7B-It2025.02 | 94.81 | 82.6 | |
| AnyEdit*LLM=Llama3-8B-It2025.02 | 94.7 | 85.75 | |
| AnyEditLLM=Qwen2.5-7B-It2025.02 | 93.56 | 79.98 | |
| UnKELLM=Llama3-8B-It2025.02 | 93.38 | 78.42 | |
| UnKELLM=Qwen2.5-7B-It2025.02 | 89.17 | 67 | |
| AlphaEditLLM=Qwen2.5-7B-It2025.02 | 78.38 | 38.26 | |
| MEMITLLM=Qwen2.5-7B-It2025.02 | 76.5 | 28.65 | |
| MEMITLLM=Llama3-8B-It2025.02 | 74.25 | 28.65 | |
| AlphaEditLLM=Llama3-8B-It2025.02 | 72.96 | 25.92 | |
| ROMELLM=Qwen2.5-7B-It2025.02 | 71.9 | 28.12 | |
| ROMELLM=Llama3-8B-It2025.02 | 70.54 | 22.39 | |
| Pre-editedLLM=Qwen2.5-7B-It2025.02 | 64.39 | 24.02 | |
| MENDLLM=Llama3-8B-It2025.02 | 64.11 | 28.05 | |
| Pre-editedLLM=Llama3-8B-It2025.02 | 62.73 | 23.52 | |
| MENDLLM=Qwen2.5-7B-It2025.02 | 62.01 | 27.92 | |
| FT-LLLM=Qwen2.5-7B-It2025.02 | 40.33 | 12.93 | |
| FT-LLLM=Llama3-8B-It2025.02 | 37.29 | 8.51 |