AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

IsoCLIP: Decomposing CLIP Projectors for Efficient Intra-modal Alignment

Simone Magistri, Dipam Goswami, Marco Mistretta, Bartłomiej Twardowski, Joost van de Weijer, Andrew D. Bagdanov

CVPR 2026 · Vision-language and multimodal › Cross-modal learning · approach: Contrastive vision-language alignment · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace standard projector with IsoCLIP isotropic-subspace projectorCLIP ViT-B/16 · Image-to-image retrievalmAP avg: 46.3 -> 52.8Improved
replace standard projector with IsoCLIPCLIP ViT-L/14 · Image-to-image retrievalmAP avg: 54.3 -> 60.7Improved
replace standard projector with IsoCLIPPE-Core-B-16 · Image-to-image retrievalmAP avg: 58.6 -> 62.8Improved
replace OTI inversion with IsoCLIPCLIP ViT-B/16 · Image-to-image retrievalmAP avg: 49.4 -> 52.8Improved
replace OTI with IsoCLIPCLIP ViT-B/16-open · Image-to-image retrievalmAP avg: 58.7 -> 58.6No clear effect
replace OTI with IsoCLIPCLIP ViT-B/32 · Image-to-image retrievalquery latency (ms): 1879 -> 7Improved
replace standard text projector with IsoCLIPCLIP ViT-L/14 · Text-to-text retrievalmAP avg: 38.2 -> 43.0Improved
replace OVI with IsoCLIPCLIP ViT-B/16-open · Text-to-text retrievalmAP avg: 45.3 -> 45.3No clear effect
replace OVI with IsoCLIPCLIP ViT-B/32 · Text-to-text retrievalquery latency (ms): 11549 -> 6Improved
replace image-image NCM with IsoCLIP NCMCLIP ViT-B/16 · Image classificationaccuracy avg: 77.4 -> 80.1Improved
replace pre-projection features with IsoCLIP projector (ablation) (ablation)CLIP ViT-B/16 · Image-to-image retrievalmAP avg: 48.4 -> 52.8Improved
replace whitened projector with IsoCLIP projector (ablation) (ablation)CLIP ViT-B/16 · Image-to-image retrievalmAP avg: 49.0 -> 52.8Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy