RTD + LinCIR (CLIP G/14)

Reported on 6 benchmarks across 2 tasks · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Computer Vision6 results

Image RetrievalonFashion IQ
(Recall@10+Recall@50)/2· 2024-06-13
56.74
best: 71.77 (DQU-CIR)
An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval arXiv:2406.09188
Image RetrievalonCIRCO
mAP@10· 2024-06-13
22.29
best: 43.4 (MMRet-MLLM)
An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval arXiv:2406.09188
Image RetrievalonCIRR
R@5· 2024-06-13
67.47
best: 81.08 (CoVR-BLIP-2)
An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval arXiv:2406.09188
Composed Image Retrieval (CoIR)onFashion IQ
(Recall@10+Recall@50)/2· 2024-06-13
56.74
best: 60.57 (CoVR-BLIP-2)
An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval arXiv:2406.09188
Composed Image Retrieval (CoIR)onCIRCO
mAP@10· 2024-06-13
22.29
best: 43.4 (MMRet-MLLM)
An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval arXiv:2406.09188
Composed Image Retrieval (CoIR)onCIRR
R@5· 2024-06-13
67.47
best: 81.08 (CoVR-BLIP-2)
An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval arXiv:2406.09188