Text Retrieval on Flickr30K (test)
100R@1GLEAM
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| GLEAMSurrogate=ALBEF, Target=ALBEF2026.06 | 100 | — | — | — | — | — | — | |
| DeBias-AttackSurrogate=ALBEF, Target=ALBEF2026.06 | 100 | — | — | — | — | — | — | |
| SGASurrogate=TCL, Target=TCL2026.06 | 100 | — | — | — | — | — | — | |
| SA-AETSurrogate=TCL, Target=TCL2026.06 | 100 | — | — | — | — | — | — | |
| GLEAMSurrogate=TCL, Target=TCL2026.06 | 100 | — | — | — | — | — | — | |
| DeBias-AttackSurrogate=TCL, Target=TCL2026.06 | 100 | — | — | — | — | — | — | |
| SGASurrogate=CLIPViT, Target=CLIPViT2026.06 | 100 | — | — | — | — | — | — | |
| DRASurrogate=CLIPViT, Target=CLIPViT2026.06 | 100 | — | — | — | — | — | — | |
| TMMSurrogate=CLIPViT, Target=CLIPViT2026.06 | 100 | — | — | — | — | — | — | |
| SA-AETSurrogate=CLIPViT, Target=CLIPViT2026.06 | 100 | — | — | — | — | — | — | |
| DeBias-AttackSurrogate=CLIPViT, Target=CLIPViT2026.06 | 100 | — | — | — | — | — | — | |
| SA-AETSurrogate=CLIPCNN, Target=CLIPCNN2026.06 | 100 | — | — | — | — | — | — | |
| GLEAMSurrogate=CLIPCNN, Target=CLIPCNN2026.06 | 100 | — | — | — | — | — | — | |
| DeBias-AttackSurrogate=CLIPCNN, Target=CLIPCNN2026.06 | 100 | — | — | — | — | — | — | |
| SA-AETSurrogate=ALBEF, Target=ALBEF2026.06 | 99.91 | — | — | — | — | — | — | |
| DRASurrogate=TCL, Target=TCL2026.06 | 99.91 | — | — | — | — | — | — | |
| GLEAMSurrogate=CLIPViT, Target=CLIPViT2026.06 | 99.88 | — | — | — | — | — | — | |
| SGASurrogate=ALBEF, Target=ALBEF2026.06 | 99.86 | — | — | — | — | — | — | |
| SGASurrogate=CLIPCNN, Target=CLIPCNN2026.06 | 99.86 | — | — | — | — | — | — | |
| DRASurrogate=CLIPCNN, Target=CLIPCNN2026.06 | 99.84 | — | — | — | — | — | — | |
| DRASurrogate=ALBEF, Target=ALBEF2026.06 | 99.83 | — | — | — | — | — | — | |
| SA-AETSurrogate=TCL, Target=ALBEF2026.06 | 98.73 | — | — | — | — | — | — | |
| GLEAMSurrogate=TCL, Target=ALBEF2026.06 | 98.17 | — | — | — | — | — | — | |
| TMMSurrogate=TCL, Target=TCL2026.06 | 97.87 | — | — | — | — | — | — | |
| TMMSurrogate=CLIPCNN, Target=CLIPCNN2026.06 | 97.74 | — | — | — | — | — | — | |
| DeBias-AttackSurrogate=ALBEF, Target=TCL2026.06 | 97.68 | — | — | — | — | — | — | |
| TMMSurrogate=ALBEF, Target=ALBEF2026.06 | 97.53 | — | — | — | — | — | — | |
| UniMoCoZero-shot=true, Base Model=Qwen2-VL-7B2025.05 | 96.8 | 99.7 | 99.9 | — | — | — | — | |
| Co-AttackSurrogate=ALBEF, Target=ALBEF2026.06 | 96.72 | — | — | — | — | — | — | |
| GLEAMSurrogate=ALBEF, Target=TCL2026.06 | 96.27 | — | — | — | — | — | — | |
| SA-AETSurrogate=ALBEF, Target=TCL2026.06 | 96.2 | — | — | — | — | — | — | |
| ALBEFPre-training Images=14M, Evaluation Protocol=Finetuning2021.11 | 95.9 | 99.8 | 100 | — | — | — | — | |
| ALBEF(14M)Pre-training Data Scale=> 10M images, Evaluation Protocol=fine-tuned2022.11 | 95.9 | 99.8 | 100 | — | — | — | — | |
| SCLPre-training Data Scale=< 10M images, Evaluation Protocol=fine-tuned2022.11 | 95.9 | 99.8 | 100 | — | — | — | — | |
| Co-AttackSurrogate=CLIPCNN, Target=CLIPCNN2026.06 | 95.66 | — | — | — | — | — | — | |
| DRASurrogate=TCL, Target=ALBEF2026.06 | 95.45 | — | — | — | — | — | — | |
| ALIGNPre-training Data Scale=> 10M images, Evaluation Protocol=fine-tuned2022.11 | 95.3 | 99.8 | 100 | — | — | — | — | |
| TCLPre-training Data Scale=< 10M images, Evaluation Protocol=fine-tuned2022.11 | 94.9 | 99.5 | 99.8 | — | — | — | — | |
| VLM2VECZero-shot=true2025.05 | 94.6 | 99.5 | 99.8 | — | — | — | — | |
| ALBEFPre-training Images=4M, Evaluation Protocol=Finetuning2021.11 | 94.3 | 99.4 | 99.8 | — | — | — | — | |
| METER-CLIP-ViT-BasePre-training Images=4M, Backbone=CLIP-ViT-Base, Evaluation Protocol=Finetuning2021.11 | 94.3 | 99.6 | 99.9 | — | — | — | — | |
| ALBEF(4M)Pre-training Data Scale=< 10M images, Evaluation Protocol=fine-tuned2022.11 | 94.3 | 99.4 | 99.8 | — | — | — | — | |
| METERPre-training Data Scale=< 10M images, Evaluation Protocol=fine-tuned2022.11 | 94.3 | 99.6 | 99.9 | — | — | — | — | |
| EVA-02-CLIP-E/14+zero-shot=true, backbone=ViT-E/14+2023.03 | 93.9 | 99.4 | 99.8 | — | — | — | — | |
| DeBias-AttackSurrogate=TCL, Target=ALBEF2026.06 | 93.43 | — | — | — | — | — | — | |
| ALBEF BaseParams=419, TFLOPS=7.412023.10 | 93.4 | 99.5 | 99.6 | — | — | — | — | |
| Open CLIP-G/14zero-shot=true, backbone=ViT-G/142023.03 | 92.9 | 99.3 | 99.8 | — | — | — | — | |
| SGASurrogate=TCL, Target=ALBEF2026.06 | 92.71 | — | — | — | — | — | — | |
| METER-Swin-BasePre-training Images=4M, Backbone=Swin-Base, Evaluation Protocol=Finetuning2021.11 | 92.4 | 99 | 99.5 | — | — | — | — | |
| EVA-02-CLIP-E/14zero-shot=true, backbone=ViT-E/142023.03 | 92.4 | 99.3 | 99.9 | — | — | — | — | |
| PGDSurrogate=ALBEF, Target=ALBEF2026.06 | 92.21 | — | — | — | — | — | — | |
| ALBEF LoRAParams=431, TFLOPS=7.492023.10 | 92.1 | 99.2 | 99 | — | — | — | — | |
| DRASurrogate=ALBEF, Target=TCL2026.06 | 91.88 | — | — | — | — | — | — | |
| EVA-01-CLIP-g/14+zero-shot=true, backbone=ViT-g/14+2023.03 | 91.6 | 99.3 | 99.8 | — | — | — | — | |
| PELAParams=173, TFLOPS=2.582023.10 | 91.6 | 99.3 | 99.6 | — | — | — | — | |
| Open CLIP-g/14zero-shot=true, backbone=ViT-g/142023.03 | 91.4 | 99.2 | 99.6 | — | — | — | — | |
| Co-AttackSurrogate=TCL, Target=TCL2026.06 | 91.29 | — | — | — | — | — | — | |
| Open CLIP-H/14zero-shot=true, backbone=ViT-H/142023.03 | 90.8 | 99.3 | 99.7 | — | — | — | — | |
| EVA-02-CLIP-L/14zero-shot=true, backbone=ViT-L/142023.03 | 89.7 | 98.6 | 99.2 | — | — | — | — | |
| PGDSurrogate=CLIPCNN, Target=CLIPCNN2026.06 | 89.54 | — | — | — | — | — | — | |
| UNIMO-LargePre-training Images=<10M, Evaluation Protocol=Finetuning2021.11 | 89.4 | 98.9 | 99.8 | — | — | — | — | |
| EVA-02-CLIP-L/14+zero-shot=true, backbone=ViT-L/14+2023.03 | 89.2 | 98.9 | 99.6 | — | — | — | — | |
| PGDSurrogate=TCL, Target=TCL2026.06 | 89.08 | — | — | — | — | — | — | |
| ERNIE-ViLEvaluation Protocol=Finetune2021.02 | 88.7 | 98 | 99.2 | — | — | — | — | |
| Open CLIP-L/14zero-shot=true, backbone=ViT-L/142023.03 | 88.7 | 98.4 | 99.2 | — | — | — | — | |
| EVA-01-CLIP-g/14zero-shot=true, backbone=ViT-g/142023.03 | 88.3 | 98.3 | 99.3 | — | — | — | — | |
| CLIPEvaluation Protocol=Zero-Shot2021.02 | 88 | 98.7 | 99.4 | — | — | — | — | |
| VILLAEvaluation Protocol=Finetune2021.02 | 87.9 | 97.5 | 98.8 | — | — | — | — | |
| VILLA-LargePre-training Images=<10M, Evaluation Protocol=Finetuning2021.11 | 87.9 | 97.5 | 98.8 | — | — | — | — | |
| VILLAParams=1102023.10 | 87.9 | 97.5 | 98.8 | — | — | — | — | |
| SGASurrogate=ALBEF, Target=TCL2026.06 | 87.42 | — | — | — | — | — | — | |
| OpenAI CLIP-L/14+zero-shot=true, backbone=ViT-L/14+2023.03 | 87.4 | 98.3 | 99.3 | — | — | — | — | |
| UniterEvaluation Protocol=Finetune2021.02 | 87.3 | 98 | 99.2 | — | — | — | — | |
| UNITER-LargePre-training Images=<10M, Evaluation Protocol=Finetuning2021.11 | 87.3 | 98 | 99.2 | — | — | — | — | |
| UNITERPre-training Data Scale=< 10M images, Evaluation Protocol=fine-tuned2022.11 | 87.3 | 98 | 99.2 | — | — | — | — | |
| UNITERParams=110, TFLOPS=0.372023.10 | 87.3 | 98 | 99.2 | — | — | — | — | |
| PixelBERTPre-training Images=<10M, Evaluation Protocol=Finetuning2021.11 | 87 | 98.9 | 99.5 | — | — | — | — | |
| Visual ParsingPre-training Images=<10M, Evaluation Protocol=Finetuning2021.11 | 87 | 98.4 | 99.5 | — | — | — | — | |
| PixelBERTPre-training Data Scale=< 10M images, Evaluation Protocol=fine-tuned2022.11 | 87 | 98.9 | 99.5 | — | — | — | — | |
| ERNIE-ViLParams=210M, Architecture=Two stream2021.06 | 86.7 | 97.8 | 99 | — | — | — | — | |
| SOHOPre-training Data Scale=< 10M images, Evaluation Protocol=fine-tuned2022.11 | 86.5 | 98.1 | 99.3 | — | — | — | — | |
| Open CLIP-B/16zero-shot=true, backbone=ViT-B/162023.03 | 86.3 | 97.9 | 99.4 | — | — | — | — | |
| E2E-VLPParams=94M, Architecture=End2End2021.06 | 86.24 | 97.5 | 98.92 | — | — | — | — | |
| Unicoder-VLEvaluation Protocol=Finetune2021.02 | 86.2 | 96.3 | 99 | — | — | — | — | |
| Unicoder-VLParams=110M, Architecture=Single stream2021.06 | 86.2 | 96.3 | 99 | — | — | — | — | |
| UNITERParams=110M, Architecture=Single stream2021.06 | 85.9 | 97.1 | 98.8 | — | — | — | — | |
| EVA-02-CLIP-B/16zero-shot=true, backbone=ViT-B/162023.03 | 85.7 | 96.7 | 98.9 | — | — | — | — | |
| FFF-ViT-B/32zero-shot=true, pre-train dataset=YFCC15M, backbone=ViT-B/322024.05 | 85.3 | 97.5 | 99.4 | — | — | — | — | |
| OpenAI CLIP-L/14zero-shot=true, backbone=ViT-L/142023.03 | 85.2 | 97.3 | 99 | — | — | — | — | |
| CLIP-L/14Zero-shot=true2025.05 | 85.2 | 97.3 | 99 | — | — | — | — | |
| UniterEvaluation Protocol=Zero-Shot2021.02 | 83.6 | 95.7 | 97.7 | — | — | — | — | |
| ViLTPre-training Images=<10M, Evaluation Protocol=Finetuning2021.11 | 83.5 | 96.7 | 98.6 | — | — | — | — | |
| CODERImage Encoder=Faster-RCNN, Text Encoder=BERT2022.08 | 83.2 | 96.5 | 98 | 520.9 | — | — | — | |
| DeBias-AttackSurrogate=CLIPViT, Target=CLIPCNN2026.06 | 83.01 | — | — | — | — | — | — | |
| OpenAI CLIP-B/16zero-shot=true, backbone=ViT-B/162023.03 | 81.9 | 96.2 | 98.8 | — | — | — | — | |
| GLEAMSurrogate=CLIPViT, Target=CLIPCNN2026.06 | 79.25 | — | — | — | — | — | — | |
| MMnasNet2020.04 | 78.3 | 94.6 | 97.4 | — | — | — | — | |
| Co-AttackSurrogate=CLIPViT, Target=CLIPViT2026.06 | 78.21 | — | — | — | — | — | — | |
| GPOImage Encoder=Faster-RCNN, Text Encoder=BERT2022.08 | 78.1 | 94.1 | 97.8 | 502.3 | — | — | — | |
| SGRAFImage Encoder=Faster-RCNN2022.08 | 77.8 | 94.1 | 97.4 | 499.6 | — | — | — |