Multimodal Knowledge Editing on MMQAKE Rephrased Image
51.27M-AccHybrid-DMKG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Hybrid-DMKGBackbone=LLaVA (7B)2025.11 | 51.27 | 26.16 | |
| Hybrid-DMKGBackbone=BLIP-2 (3.8B)2025.11 | 45.27 | 26.08 | |
| IKEBackbone=LLaVA (7B)2025.11 | 37.61 | 16.91 | |
| Hybrid-DMKGBackbone=MiniGPT-4 (7.8B)2025.11 | 33.41 | 22.23 | |
| IKEBackbone=BLIP-2 (3.8B)2025.11 | 14.44 | 6.06 | |
| IKEBackbone=MiniGPT-4 (7.8B)2025.11 | 9.86 | 5.76 | |
| FT(QFor)Backbone=LLaVA (7B)2025.11 | 5.71 | 0.53 | |
| FT(QFor)Backbone=MiniGPT-4 (7.8B)2025.11 | 4.17 | 0.77 | |
| FT(All)Backbone=LLaVA (7B)2025.11 | 1.61 | 0 | |
| SERACBackbone=BLIP-2 (3.8B)2025.11 | 1.04 | 0 | |
| SERACBackbone=LLaVA (7B)2025.11 | 0.97 | 0.02 | |
| FT(QFor)Backbone=BLIP-2 (3.8B)2025.11 | 0.84 | 0.11 | |
| SERACBackbone=MiniGPT-4 (7.8B)2025.11 | 0.13 | 0 | |
| FT(All)Backbone=MiniGPT-4 (7.8B)2025.11 | 0.11 | 0 | |
| MENDBackbone=LLaVA (7B)2025.11 | 0.06 | 0 | |
| FT(All)Backbone=BLIP-2 (3.8B)2025.11 | 0.04 | 0 | |
| MENDBackbone=MiniGPT-4 (7.8B)2025.11 | 0.04 | 0 | |
| MENDBackbone=BLIP-2 (3.8B)2025.11 | 0.02 | 0 |