AI Paper Atlas

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

Mask to Align, Weight to Disambiguate: Reliable Unsupervised Cross-Modal Hashing with Masked-Weight Contrast

Fan Yang, Yuanzhi Zhao, Haimei Zhao, Yudong Zhao, Haikun Xu

CVPR 2026 · Vision-language and multimodal › Cross-modal learning · approach: Cross-modal hashing

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
add L_CCA and L_SSR to L_WMCL objective (ablation)UWMCH · MIRFLICKR-25K I→T 16-bitmAP (%): 87.43 -> 88.64Improved
add L_CCA and L_SSR to L_WMCL objective (ablation)UWMCH · MIRFLICKR-25K T→I 16-bitmAP (%): 86.28 -> 88.29Improved
add L_CCA and L_SSR to L_WMCL objective (ablation)UWMCH · MIRFLICKR-25K I→T 128-bitmAP (%): 91.07 -> 91.60Improved
add L_CCA (cluster-centroid agreement) to L_WMCL (ablation)UWMCH · MIRFLICKR-25K I→T 16-bitmAP (%): 87.43 -> 88.40Improved
add L_SSR (semantic structure regularization) to L_WMCL (ablation)UWMCH · MIRFLICKR-25K I→T 16-bitmAP (%): 87.43 -> 88.16Improved
add L_CCA to L_WMCL (ablation)UWMCH · MIRFLICKR-25K T→I 128-bitmAP (%): 89.88 -> 90.40Improved
UWMCH vs strongest baselinesUWMCH · MS COCO I→T 128-bitmAP (%): +1.19 mAPImproved
UWMCH vs strongest baselinesUWMCH · NUS-WIDE I→T 16-bitmAP (%): +1.28 mAPImproved
UWMCH vs RSHNL on T→IUWMCH · MIRFLICKR-25K T→I 128-bitmAP (%): 91.21 -> 90.59Hurt
UWMCH across all settings vs baselinesUWMCH · MIRFLICKR-25K, NUS-WIDE, MS COCOmAP (%): best in 21 of 24 settingsMixed

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