Flowalign improves mapping between different types of data spaces
Flow-Guided Target-Space Alignment via Path Consistency
Artificial IntelligenceComputer Vision and Pattern RecognitionMachine Learning
Summary
Sometimes computers need to match information from one type, like pictures, to another, like words, in a special space where we can compare them directly. The authors show that existing methods either focus on matching pairs or global shapes but miss some details about how things connect in that space. They introduce FlowAlign, a new method that uses a learned flow to better match examples while keeping the overall space organized. Tests show that this approach improves how well the matches respect both individual pairs and the overall structure.
What this means in practice
- •For data scientists: Improve integration of paired datasets like images and text by better aligning them in a shared fixed space, enhancing downstream analysis.
- •For biomedical engineers: Boost alignment quality in single-cell multiome data integration, supporting more accurate biological interpretation.
Authors
Ruizhi Yuan, Zeqiu Yu, Wei Gao, Wei Chen, Xiwei Tang, Lu Tang
Abstract
Many multimodal alignment problems require mapping a source modality into a fixed space defined by a target modality, where the target space is not merely an intermediate embedding but is used directly for downstream analysis. In this setting, successful alignment should preserve both the global structure of the target distribution and the instance-level correspondence observed in paired source--target training data. Existing objectives typically address only part of this goal: contrastive losses encourage relative similarity, distribution-matching losses improve marginal agreement, and pointwise regression fits each source sample to its paired target without explicitly modeling target-space path geometry. We propose FlowAlign, a flow-guided target-space alignment method that repurposes conditional flow matching as a training-time alignment signal rather than an inference-time transport model. FlowAlign uses the learned flow field to guide the alignment of source-side representations with the fixed target space, encouraging them to preserve paired correspondence while remaining compatible with the global target-space structure. We further provide theoretical analysis showing how FlowAlign helps improve paired correspondence by controlling endpoint accuracy and capturing local target-space geometry. Experiments on synthetic data, paired single-cell multiome benchmarks, and image--text alignment demonstrate that FlowAlign improves instance level correspondence while preserving target-space structure across different modality settings.