Image Captioning on MSCOCO (test)
149.6CIDErOFA
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| OFA2022.05 | 149.6 | 43.5 | 31.9 | 26.1 | — | — | — | — | — | |
| LEMONoptimization=SCST2022.05 | 145.5 | 42.6 | 31.4 | 25.5 | — | — | — | — | — | |
| OFA*Condition=Original score (no attack)2023.10 | 145.43 | 42.81 | 31.3 | 25.37 | — | — | — | — | — | |
| CoCa2022.05 | 143.6 | 40.9 | 33.9 | 24.7 | — | — | — | — | — | |
| SimVLM2022.05 | 143.3 | 40.6 | 33.7 | 25.4 | — | — | — | — | — | |
| VinVL2022.05 | 140.9 | 41 | 31.1 | 25.4 | — | — | — | — | — | |
| LEMON2022.05 | 139.1 | 41.5 | 30.8 | 24.1 | — | — | — | — | — | |
| BLIP2022.05 | 136.7 | 40.4 | — | — | — | — | — | — | — | |
| CLIP-ViL2022.05 | 134.2 | 40.2 | 29.7 | 23.8 | — | — | — | — | — | |
| BaselineSource Model=CLIP ViT-B/16, Target Model=BLIP2026.01 | 131.38 | 39.31 | 30.69 | 23.52 | 59.62 | — | — | — | — | |
| BaselineSource Model=ALBEF, Target Model=BLIP2026.01 | 131.38 | 39.31 | 30.69 | 23.52 | 59.62 | — | — | — | — | |
| AdvCLIPSource Model=CLIP ViT-B/16, Target Model=BLIP2026.01 | 127.14 | 38.2 | 30.03 | 22.88 | 58.7 | — | — | — | — | |
| AdvCLIPSource Model=ALBEF, Target Model=BLIP2026.01 | 121.79 | 36.59 | 29.15 | 21.95 | 57.35 | — | — | — | — | |
| SGASource Model=CLIP ViT-B/16, Target Model=BLIP2026.01 | 120.47 | 36.1 | 28.89 | 21.85 | 57.25 | — | — | — | — | |
| FD-UAPSource Model=CLIP ViT-B/16, Target Model=BLIP2026.01 | 119.63 | 35.91 | 28.78 | 21.65 | 57.15 | — | — | — | — | |
| C-PGCSource Model=CLIP ViT-B/16, Target Model=BLIP2026.01 | 117.33 | 35.63 | 28.44 | 21.38 | 56.75 | — | — | — | — | |
| ETUSource Model=CLIP ViT-B/16, Target Model=BLIP2026.01 | 114.81 | 34.83 | 28.07 | 21.02 | 56.29 | — | — | — | — | |
| FD-UAPSource Model=ALBEF, Target Model=BLIP2026.01 | 113.26 | 34.37 | 28.09 | 20.86 | 56.02 | — | — | — | — | |
| C-PGCSource Model=ALBEF, Target Model=BLIP2026.01 | 112.69 | 33.97 | 27.73 | 20.5 | 55.5 | — | — | — | — | |
| SGASource Model=ALBEF, Target Model=BLIP2026.01 | 112.28 | 33.92 | 27.67 | 20.62 | 55.41 | — | — | — | — | |
| ETUSource Model=ALBEF, Target Model=BLIP2026.01 | 109.34 | 33.42 | 27.28 | 20.11 | 54.96 | — | — | — | — | |
| HRASource Model=ALBEF, Target Model=BLIP2026.01 | 108.19 | 32.67 | 27.21 | 20.15 | 54.62 | — | — | — | — | |
| HRASource Model=CLIP ViT-B/16, Target Model=BLIP2026.01 | 106.6 | 32.41 | 26.8 | 19.61 | 54.31 | — | — | — | — | |
| DRTarget Model=OFA2023.10 | 95.52 | 30.26 | 24.47 | 17.89 | — | — | — | — | — | |
| FDATarget Model=OFA2023.10 | 55.75 | 17.77 | 17.92 | 11.36 | — | — | — | — | — | |
| SSPTarget Model=OFA2023.10 | 23.54 | 10.99 | 12.52 | 5.67 | — | — | — | — | — | |
| InfoNCE+LM+CSTraining Strategy=Blip2 stage one, Alignment Component=CS-Aligner2025.02 | 16.7 | — | — | — | — | — | — | — | 41.3 | |
| InfoNCE+LMTraining Strategy=Blip2 stage one2025.02 | 14.3 | — | — | — | — | — | — | — | 40.4 | |
| BSATarget Model=OFA2023.10 | 2.16 | 3.04 | 8.08 | 1.5 | — | — | — | — | — | |
| COATarget Model=Qwen2.5-VL-7B-Instruct, epsilon_img=8/2552025.12 | — | — | — | — | — | 93.59 | 0.882 | 1.47 | — | |
| COATarget Model=InternVL3.5-4B, epsilon_img=8/2552025.12 | — | — | — | — | — | 95.38 | 0.926 | 1.53 | — | |
| COATarget Model=LLaVA-1.5-7B, epsilon_img=8/2552025.12 | — | — | — | — | — | 94.74 | 0.917 | 0 | — | |
| EGATarget Model=Qwen2.5-VL-7B-Instruct, epsilon_img=8/2552025.12 | — | — | — | — | — | 94.81 | 0.883 | 42.53 | — | |
| EGATarget Model=InternVL3.5-4B, epsilon_img=8/2552025.12 | — | — | — | — | — | 93.75 | 0.862 | 37.29 | — | |
| EGATarget Model=LLaVA-1.5-7B, epsilon_img=8/2552025.12 | — | — | — | — | — | 93.12 | 0.846 | 47.05 | — | |
| MIETarget Model=Qwen2.5-VL-7B-Instruct, epsilon_img=8/2552025.12 | — | — | — | — | — | 94.18 | 0.892 | 13.95 | — | |
| MIETarget Model=InternVL3.5-4B, epsilon_img=8/2552025.12 | — | — | — | — | — | 94.83 | 0.905 | 23.05 | — | |
| MIETarget Model=LLaVA-1.5-7B, epsilon_img=8/2552025.12 | — | — | — | — | — | 93.59 | 0.893 | 21.29 | — | |
| PGDTarget Model=Qwen2.5-VL-7B-Instruct, epsilon_img=8/2552025.12 | — | — | — | — | — | 91.16 | 0.842 | 1.28 | — | |
| PGDTarget Model=InternVL3.5-4B, epsilon_img=8/2552025.12 | — | — | — | — | — | 88.98 | 0.793 | 1.41 | — | |
| PGDTarget Model=LLaVA-1.5-7B, epsilon_img=8/2552025.12 | — | — | — | — | — | 90.75 | 0.834 | 1.52 | — | |
| VLATarget Model=Qwen2.5-VL-7B-Instruct, epsilon_img=8/2552025.12 | — | — | — | — | — | 89.22 | 0.801 | 1.49 | — | |
| VLATarget Model=InternVL3.5-4B, epsilon_img=8/2552025.12 | — | — | — | — | — | 89.49 | 0.804 | 1.4 | — | |
| VLATarget Model=LLaVA-1.5-7B, epsilon_img=8/2552025.12 | — | — | — | — | — | 87.75 | 0.778 | 0 | — |