Image Captioning on MSCOCO 1k
0.94ΔCIDErMIE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MIESource Model=InternVL3.5-4B, Target Model=InternVL3.5-4B, ε_img=8/2552025.12 | 0.94 | 2,305 | |
| XTASource Model=LLaVA-1.5-7B, Target Model=LLaVA-1.5-7B, ε_img=8/2552025.12 | 0.91 | 113 | |
| MIESource Model=Qwen2.5-VL-7B-Instruct, Target Model=Qwen2.5-VL-7B-Instruct, ε_img=8/2552025.12 | 0.89 | 1,395 | |
| XTASource Model=InternVL3.5-4B, Target Model=InternVL3.5-4B, ε_img=8/2552025.12 | 0.89 | 38 | |
| MIESource Model=LLaVA-1.5-7B, Target Model=LLaVA-1.5-7B, ε_img=8/2552025.12 | 0.89 | 2,129 | |
| EGASource Model=Qwen2.5-VL-7B-Instruct, Target Model=Qwen2.5-VL-7B-Instruct, ε_img=8/2552025.12 | 0.88 | 4,253 | |
| EGASource Model=InternVL3.5-4B, Target Model=InternVL3.5-4B, ε_img=8/2552025.12 | 0.86 | 3,729 | |
| XTASource Model=Qwen2.5-VL-7B-Instruct, Target Model=Qwen2.5-VL-7B-Instruct, ε_img=8/2552025.12 | 0.85 | 86 | |
| EGASource Model=LLaVA-1.5-7B, Target Model=LLaVA-1.5-7B, ε_img=8/2552025.12 | 0.84 | 4,705 | |
| XTASource Model=Qwen2.5-VL-7B-Instruct, Target Model=LLaVA-1.5-7B, ε_img=8/2552025.12 | 0.81 | 2 | |
| XTASource Model=Qwen2.5-VL-7B-Instruct, Target Model=InternVL3.5-4B, ε_img=8/2552025.12 | 0.77 | 44 | |
| XTASource Model=InternVL3.5-4B, Target Model=Qwen2.5-VL-7B-Instruct, ε_img=8/2552025.12 | 0.74 | 18 | |
| XTASource Model=LLaVA-1.5-7B, Target Model=Qwen2.5-VL-7B-Instruct, ε_img=8/2552025.12 | 0.74 | 9 | |
| XTASource Model=InternVL3.5-4B, Target Model=LLaVA-1.5-7B, ε_img=8/2552025.12 | 0.73 | 6 | |
| XTASource Model=LLaVA-1.5-7B, Target Model=InternVL3.5-4B, ε_img=8/2552025.12 | 0.73 | 27 | |
| EGASource Model=Qwen2.5-VL-7B-Instruct, Target Model=InternVL3.5-4B, ε_img=8/2552025.12 | 0.42 | 1,927 | |
| EGASource Model=InternVL3.5-4B, Target Model=Qwen2.5-VL-7B-Instruct, ε_img=8/2552025.12 | 0.39 | 2,162 | |
| EGASource Model=LLaVA-1.5-7B, Target Model=Qwen2.5-VL-7B-Instruct, ε_img=8/2552025.12 | 0.39 | 2,432 | |
| EGASource Model=InternVL3.5-4B, Target Model=LLaVA-1.5-7B, ε_img=8/2552025.12 | 0.37 | 2,363 | |
| EGASource Model=LLaVA-1.5-7B, Target Model=InternVL3.5-4B, ε_img=8/2552025.12 | 0.36 | 2,620 | |
| EGASource Model=Qwen2.5-VL-7B-Instruct, Target Model=LLaVA-1.5-7B, ε_img=8/2552025.12 | 0.33 | 1,737 | |
| MIESource Model=Qwen2.5-VL-7B-Instruct, Target Model=LLaVA-1.5-7B, ε_img=8/2552025.12 | 0.32 | 993 | |
| MIESource Model=InternVL3.5-4B, Target Model=LLaVA-1.5-7B, ε_img=8/2552025.12 | 0.31 | 1,236 | |
| MIESource Model=InternVL3.5-4B, Target Model=Qwen2.5-VL-7B-Instruct, ε_img=8/2552025.12 | 0.3 | 1,163 | |
| MIESource Model=LLaVA-1.5-7B, Target Model=Qwen2.5-VL-7B-Instruct, ε_img=8/2552025.12 | 0.29 | 1,256 | |
| MIESource Model=LLaVA-1.5-7B, Target Model=InternVL3.5-4B, ε_img=8/2552025.12 | 0.29 | 1,345 | |
| MIESource Model=Qwen2.5-VL-7B-Instruct, Target Model=InternVL3.5-4B, ε_img=8/2552025.12 | 0.23 | 1,065 |