AI Paper Atlas

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

Stable Spectral Copula Alignment for Robust Multimodal Learning

Hongkang Zhang, Shao-Lun Huang, Yanlong Wang, Ercan Engin KURUOGLU

ICML 2026 · Vision-language and multimodal › Cross-modal learning · approach: Unimodal-to-multimodal representation transfer

Read the paper

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 copula stabilization (ablation)SSCA · CMU-MOSEIdegradation under 10x scaling: 4.8 -> 1.8Improved
add copula stabilization (ablation)SSCA · MELDdegradation under 10x scaling: 4.5 -> 1.5Improved
add multiway hub coupling instead of pairwise (ablation)SSCA · CMU-MOSEIdegradation under 10x scaling: 3.8 -> 1.8Improved
add multiway hub coupling instead of pairwise (ablation)SSCA · MELDdegradation under 10x scaling: 3.5 -> 1.5Improved
add spectral alignment instead of contrastive loss (ablation)SSCA · CMU-MOSEIdegradation under 10x scaling: 3.5 -> 1.8Improved
add spectral alignment instead of contrastive loss (ablation)SSCA · MELDdegradation under 10x scaling: 3.2 -> 1.5Improved
add hub coupling and spectral stabilization to copula-only SigLIPCopula-only + SigLIP · CMU-MOSEIdegradation under 10x scaling: 2.3 -> 1.8Improved
add hub coupling and spectral stabilization to copula-only SigLIPCopula-only + SigLIP · MELDdegradation under 10x scaling: 2.2 -> 1.5Improved
add copula stabilization to SigLIPSigLIP · CMU-MOSEIdegradation under 10x scaling: 6.3 -> 2.3Improved
use SSCA instead of MMMLMMML · CMU-MOSEIdegradation under 10x scaling: 3.2 -> 1.8Improved
use SSCA instead of CLIP baseline under JPEG Q=50CLIP ViT-B/16 · CC3M-500K Recall@1degradation under 10x scaling: 5.1 -> 2.5Improved
use SSCA instead of MMML on average real-world drift degradationMMML · CMU-MOSEI and MELD driftsdegradation under 10x scaling: 2.2 -> 1.6Improved

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