Knowledge Editing Reliability
99.72Target-based EMFT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| FTModel=LLaVA2026.04 | 99.72 | 99.75 | 99.98 | 100 | |
| FTModel=InstructBLIP2026.04 | 98.24 | 97.78 | 99.39 | 99.88 | |
| FTModel=Qwen-VL2026.04 | 98.21 | 98.95 | 98.83 | 99.63 | |
| DECODEModel=InstructBLIP2026.04 | 96.27 | 96.79 | 91.69 | 92.16 | |
| DECODEModel=LLaVA2026.04 | 94.49 | 93.27 | 91.05 | 91.36 | |
| FiNEModel=LLaVA2026.04 | 92.74 | 91.86 | 97.01 | 94.88 | |
| DECODEModel=Qwen-VL2026.04 | 92.69 | 91.98 | 97.59 | 97.35 | |
| IKEModel=Qwen-VL2026.04 | 84.68 | 93.21 | 25.16 | 23.44 | |
| FiNEModel=Qwen-VL2026.04 | 79.84 | 76.19 | 92.28 | 88.83 | |
| FiNEModel=InstructBLIP2026.04 | 68.33 | 72.4 | 86.46 | 90.53 | |
| IKEModel=LLaVA2026.04 | 36.47 | 80.57 | 13.11 | 73.78 | |
| SERACModel=InstructBLIP2026.04 | 28.1 | 27.39 | 99.63 | 99.14 | |
| SERACModel=Qwen-VL2026.04 | 27.75 | 27.39 | 99.76 | 99.07 | |
| SERACModel=LLaVA2026.04 | 26.55 | 28.62 | 99.58 | 98.89 | |
| IKEModel=InstructBLIP2026.04 | 10.83 | 66.07 | 3.83 | 63.23 |