Image Retrieval on Flickr30K
2,620R@1m-CNNENS
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| m-CNNENSBackbone=VGG2015.04 | 2,620 | 5,630 | 6,960 | 4 | |
| RTP (weighted distance)2015.04 | 2,600 | 5,600 | 6,930 | — | |
| RTP (SAE)2015.04 | 2,540 | 5,520 | 6,860 | — | |
| FV (GMM+HGLMM)2015.04 | 2,500 | 5,270 | 6,600 | 5 | |
| FV (HGLMM)2015.04 | 2,440 | 5,210 | 6,560 | 5 | |
| FV (GMM)2015.04 | 2,390 | 5,160 | 6,490 | 5 | |
| FV (LMM)2015.04 | 2,360 | 5,120 | 6,440 | 5 | |
| m-RNN-vggBackbone=VGG2015.04 | 2,280 | 5,070 | 6,310 | 5 | |
| RTP (joint training)2015.04 | 2,200 | 5,070 | 6,200 | — | |
| FV (Mean Vec)2015.04 | 2,050 | 4,630 | 5,930 | 6.8 | |
| m-CNNphsBackbone=VGG2015.04 | 1,970 | 4,820 | 6,220 | 6 | |
| m-CNNstBackbone=VGG2015.04 | 1,970 | 4,840 | 6,230 | 6 | |
| m-CNNphlBackbone=VGG2015.04 | 1,940 | 4,930 | 6,240 | 6 | |
| m-CNNwdBackbone=VGG2015.04 | 1,820 | 4,720 | 6,090 | 6 | |
| LRCN2015.04 | 1,750 | 4,030 | 5,080 | 9 | |
| NIC2015.04 | 1,700 | — | 5,700 | 7 | |
| MNLM-vggBackbone=VGG2015.04 | 1,680 | 4,200 | 5,650 | 8 | |
| m-CNNENSBackbone=OverFeat2015.04 | 1,590 | 4,030 | 5,190 | 9.5 | |
| DVSA (BRNN)2015.04 | 1,520 | 3,770 | 5,050 | 9.2 | |
| DVSA (DepTree)2015.04 | 1,500 | 3,650 | 4,820 | 10.4 | |
| RVP (T)2015.04 | 1,280 | 3,290 | 4,450 | 13 | |
| RVP (T+I)2015.04 | 1,270 | 3,310 | 4,490 | 12.5 | |
| m-RNN2015.04 | 1,260 | 3,120 | 4,150 | 16 | |
| DCCA2015.04 | 1,260 | 3,100 | 4,300 | 15 | |
| m-CNNstBackbone=OverFeat2015.04 | 1,250 | 3,280 | 4,420 | 14 | |
| m-CNNphsBackbone=OverFeat2015.04 | 1,240 | 3,330 | 4,470 | 14 | |
| MNLM2015.04 | 1,180 | 3,400 | 4,630 | 13 | |
| m-CNNwdBackbone=OverFeat2015.04 | 1,160 | 3,210 | 4,420 | 14 | |
| m-CNNphlBackbone=OverFeat2015.04 | 1,160 | 3,270 | 4,410 | 14 | |
| Deep Fragment2015.04 | 1,020 | 3,080 | 4,420 | 14 | |
| SDT RNN2015.04 | 890 | 2,980 | 4,110 | 16 | |
| DeViSE2015.04 | 670 | 2,190 | 3,270 | 25 | |
| HQA-VLAttackSurrogate Model=CLIPViT, Victim Model=CLIPViT, Attack Type=White-box2026.04 | 100 | — | — | — | |
| HQA-VLAttackSurrogate Model=CLIPCNN, Victim Model=CLIPCNN, Attack Type=White-box2026.04 | 100 | — | — | — | |
| HQA-VLAttackSurrogate Model=ALBEF, Victim Model=ALBEF, Attack Type=White-box2026.04 | 99.98 | — | — | — | |
| HQA-VLAttackSurrogate Model=TCL, Victim Model=TCL, Attack Type=White-box2026.04 | 99.93 | — | — | — | |
| SGASurrogate Model=ALBEF, Victim Model=ALBEF, Attack Type=White-box2026.04 | 97.28 | — | — | — | |
| DRASurrogate Model=ALBEF, Victim Model=ALBEF, Attack Type=White-box2026.04 | 96.63 | — | — | — | |
| BEIT-3protocol=finetuned, averaging=average of top-1 image-to-text and text-to-image2022.08 | 94.2 | — | — | — | |
| Florenceprotocol=finetuned, averaging=average of top-1 image-to-text and text-to-image2022.08 | 92.6 | — | — | — | |
| BEIT-3protocol=zero-shot, averaging=average of top-1 image-to-text and text-to-image2022.08 | 88.2 | — | — | — | |
| BLIP-BASE# Pre-train Images=129M, Model scale=Base2023.05 | 87.3 | — | — | — | |
| ManagerTower-BASE# Pre-train Images=4M, Model scale=Base2023.05 | 86.56 | — | — | — | |
| CoCaprotocol=zero-shot, averaging=average of top-1 image-to-text and text-to-image2022.08 | 86.5 | — | — | — | |
| BridgeTower-BASE# Pre-train Images=4M, Model scale=Base2023.05 | 85.83 | — | — | — | |
| ALBEF-BASE (14M)# Pre-train Images=14M, Model scale=Base2023.05 | 85.6 | — | — | — | |
| VladVABackbone=LLAVA-1.5-7B, Zero-shot=true2024.12 | 85 | — | 98.5 | — | |
| Co-AttackSurrogate Model=ALBEF, Victim Model=ALBEF, Attack Type=White-box2026.04 | 83.86 | — | — | — | |
| LARE (SigLIP 2)ViT=So/16, Zero-shot=true2026.06 | 83.76 | 96.38 | 98.34 | — | |
| SigLIP 2ViT=So/16, Zero-shot=true2026.06 | 83.72 | 96.34 | 98.32 | — | |
| EVA-CLIP-18BZero-shot=true2024.02 | 83.3 | 96.3 | 98.3 | 87.8 | |
| EVA-CLIPParameters=18B, Zero-shot=true2024.12 | 83.3 | — | 97.9 | — | |
| SigLIPViT=So/14, Zero-shot=true2026.06 | 82.94 | 96.08 | 98 | — | |
| LARE (SigLIP)ViT=So/14, Zero-shot=true2026.06 | 82.94 | 96.12 | 98 | — | |
| ALBEF-BASE# Pre-train Images=4M, Model scale=Base2023.05 | 82.8 | — | — | — | |
| METER-CLIP-BASE# Pre-train Images=4M, Model scale=Base2023.05 | 82.22 | — | — | — | |
| InternVL-CZero-shot=true2024.02 | 82.1 | 96 | 98.1 | 86.6 | |
| DFN5B-CLIP-H/14+Zero-shot=true2024.02 | 82.1 | 96 | 97.9 | 87.2 | |
| OpenCLIP-G/14+Zero-shot=true2024.02 | 80.9 | 95.1 | 97.2 | 85.7 | |
| EVA-CLIP-8BZero-shot=true2024.02 | 80.8 | 95.5 | 97.6 | 86.2 | |
| EVA-CLIPParameters=8B, Zero-shot=true2024.12 | 80.3 | — | 97.2 | — | |
| DFN5B-CLIP-H/14Zero-shot=true2024.02 | 80.1 | 95.2 | 97.3 | 86.6 | |
| VLM2VecBackbone=Mistral-7B, Zero-shot=true2024.12 | 80.1 | — | 97.3 | — | |
| EVA-02-CLIP-E/14+Zero-shot=true2024.02 | 79.7 | 94.9 | 97.3 | 85.7 | |
| Uni-Perceiver-v2 LARGEparameters=446M, task-specific fine-tuning=false2022.11 | 79.6 | — | — | — | |
| Flamingo-3Bparameters=3.2B, task-specific fine-tuning=false2022.11 | 79.5 | — | — | — | |
| OpenCLIPBackbone=ViT-BigG/14, Zero-shot=true2024.12 | 79.5 | — | 97.1 | — | |
| E5-VBackbone=LLAVA-Next-8B, Zero-shot=true2024.12 | 79.5 | — | 97.6 | — | |
| VLMo-BASE# Pre-train Images=4M, Model scale=Base2023.05 | 79.3 | — | — | — | |
| EVA-01-CLIP-g/14+Zero-shot=true2024.02 | 79.2 | 95.2 | 97.3 | 85.3 | |
| FreeRet_{r20}Train Data=-, MLLM=Qwen2.5-VL 7B2025.09 | 79.1 | 93.2 | 94.9 | — | |
| METER-Swin-BASE# Pre-train Images=4M, Model scale=Base2023.05 | 79.02 | — | — | — | |
| EVA-02-CLIPBackbone=ViT-E/14+, Zero-shot=true2024.12 | 78.8 | — | 96.8 | — | |
| UNIMO-LARGE# Pre-train Images=4M, Model scale=Large2023.05 | 78.04 | — | — | — | |
| OpenCLIPBackbone=ViT-G/14, Zero-shot=true2024.12 | 77.8 | — | 96.9 | — | |
| HQA-VLAttackSurrogate Model=ALBEF, Victim Model=TCL, Attack Type=Black-box2026.04 | 77.6 | — | — | — | |
| FreeRet_{r20}Train Data=-, MLLM=Qwen2.5-VL 3B2025.09 | 77 | 89.6 | 90.8 | — | |
| E5-VBackbone=LLAVA-1.5-7B, Zero-shot=true2024.12 | 76.7 | — | 96.9 | — | |
| Uni-Perceiver-MOE LARGEparameters=505M, task-specific fine-tuning=false2022.11 | 75.9 | — | — | — | |
| UNITER-LARGE# Pre-train Images=4M, Model scale=Large2023.05 | 75.56 | — | — | — | |
| Uni-Perceivertuning_setting=fine-tuning, data_percentage=100%2021.12 | 74.9 | 93.5 | 96 | — | |
| UNIMO-BASE# Pre-train Images=4M, Model scale=Base2023.05 | 74.66 | — | — | — | |
| BLIP2Backbone=ViT-L, Zero-shot=true2024.12 | 74.5 | — | 97.2 | — | |
| Uni-Perceiver LARGEparameters=354M, task-specific fine-tuning=false2022.11 | 74.2 | — | — | — | |
| SuperCLIPPretrain=L-12.8B, Zero-shot=true2025.12 | 74.2 | — | — | — | |
| Sep-AttackSurrogate Model=ALBEF, Victim Model=ALBEF, Attack Type=White-box2026.04 | 73.95 | — | — | — | |
| Uni-Perceiver-v2 BASEparameters=308M, task-specific fine-tuning=false2022.11 | 73.8 | — | — | — | |
| UNITER LARGEPretraining Data=CC+SUB+COCO+VG2019.12 | 73.66 | 93.06 | 95.98 | — | |
| CLIPPretrain=L-12.8B, Zero-shot=true2025.12 | 72.7 | — | — | — | |
| UNITER-BASE# Pre-train Images=4M, Model scale=Base2023.05 | 72.52 | — | — | — | |
| UNITER-Btuning_setting=fine-tuning2021.12 | 72.5 | 92.4 | 96.1 | — | |
| Uni-Perceivertuning_setting=prompt-tuning, data_percentage=10%2021.12 | 72.5 | 92.3 | 95.7 | — | |
| EVA-01-CLIP-g/14Zero-shot=true2024.02 | 72.5 | 91.5 | 95.4 | 80.9 | |
| Uni-Perceiver-MOE BASEparameters=167M, task-specific fine-tuning=false2022.11 | 72.4 | — | — | — | |
| HQA-VLAttackSurrogate Model=TCL, Victim Model=ALBEF, Attack Type=Black-box2026.04 | 71.7 | — | — | — | |
| UNITER BASEPretraining Data=CC+SUB+COCO+VG2019.12 | 71.5 | 91.16 | 95.2 | — | |
| Unicoder-VLtuning_setting=fine-tuning2021.12 | 71.5 | 91.2 | 95.2 | — | |
| Uni-Perceivertuning_setting=prompt-tuning, data_percentage=1%2021.12 | 71.1 | 91.6 | 95.1 | — | |
| Uni-Perceiver BASEparameters=124M, task-specific fine-tuning=false2022.11 | 71.1 | — | — | — | |
| BLIPBackbone=ViT-L, Zero-shot=true2024.12 | 70 | — | 95.2 | — |