Disentangled Representation Learning via Flow Matching

2026-02-05

AI Summary

Existing disentangled representation learning methods often suffer from insufficient semantic alignment and information leakage among latent factors. This work proposes a novel flow-matching–based paradigm for disentangled learning, which models factor-conditional flows within a compact latent space and introduces a non-overlapping (orthogonality) regularization term to explicitly enhance semantic alignment of individual factors while suppressing cross-factor interference. Evaluated across multiple benchmark datasets, the proposed method significantly outperforms current baselines, achieving higher scores on established disentanglement metrics and demonstrating superior controllability in generation as well as improved sample fidelity.

Technology Category

• Machine Learning: Representation Learning
• Computer Vision: Representation Learning for Vision
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Disentangled representation learning aims to capture the underlying explanatory factors of observed data, enabling a principled understanding of the data-generating process. Recent advances in generative modeling have introduced new paradigms for learning such representations. However, existing diffusion-based methods encourage factor independence via inductive biases, yet frequently lack strong semantic alignment. In this work, we propose a flow matching-based framework for disentangled representation learning, which casts disentanglement as learning factor-conditioned flows in a compact latent space. To enforce explicit semantic alignment, we introduce a non-overlap (orthogonality) regularizer that suppresses cross-factor interference and reduces information leakage between factors. Extensive experiments across multiple datasets demonstrate consistent improvements over representative baselines, yielding higher disentanglement scores as well as improved controllability and sample fidelity.

Problem

• Disentangled Representation Learning
• Semantic Alignment
• Factor Independence
• Information Leakage
• Latent Factors

Innovation

• flow matching
• disentangled representation learning
• orthogonality regularizer
• factor-conditioned flows
• semantic alignment