RePercENT: Scaling Disentangled Representation Learning Beyond Two Modalities

2026-06-03Machine Learning

Machine Learning
AI summary

The authors developed RePercENT, a new method to analyze data coming from many different sources (like text, images, and sounds) at the same time. Unlike earlier methods that only worked well with two types of data, their approach can handle multiple types efficiently. It breaks down the information into shared parts and unique parts for each type without needing lots of extra training. Their technique works with existing data representations and offers mathematical guarantees for its effectiveness while using less computing power.

multimodal datadisentangled representationsself-supervised learningpairwise disentanglementmodality-specific informationjoint optimizationpre-extracted embeddingsfoundation modelscomputational complexitytheoretical guarantees
Authors
Vasiliki Rizou, Pascal Frossard, Dorina Thanou
Abstract
To leverage the full potential of multimodal data, we need representations that go beyond the state-of-the-art alignment and fusion approaches and exploit all cross-modal interactions without sacrificing modality-specific information. Learning disentangled representations is a principled way to identify these underlying shared and unique factors that are hidden in observational data. However, while multimodal disentanglement is a compelling paradigm, existing methods are largely confined to the two-modality regime due to its inherent scalability bottleneck. To address this, we propose RePercENT, a self-supervised framework designed to surpass these limitations and unlocks scalable pairwise disentanglement beyond two modalities. Through a multimodal `plug-and-play' architecture, our approach operates directly on pre-extracted embeddings, eliminating the need for extensive joint pre-training while making no assumptions regarding the underlying modalities or foundation model backbones. Moreover, we introduce a joint optimization objective for simultaneously deriving the shared and unique components, and provide formal theoretical guarantees that characterize the optimality of our solution. Across diverse modalities and tasks, RePercENT successfully recovers disentangled components while maintaining competitive performance and significantly reducing computational complexity.