Flow-Guided Target-Space Alignment via Path Consistency
arXiv cs.AI · 7 Nov 2026, 04:44 AEST
FlowAlign is a new method for aligning source modalities with a fixed target space, focusing on preserving both global structure and instance-level correspondence. The approach utilizes conditional flow matching to enhance training-time alignment signals.
Key facts
- 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.
- 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.
- 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.
- 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.
Read aloud
FlowAlign is a new method for aligning source modalities with a fixed target space, focusing on preserving both global structure and instance-level correspondence. The approach utilizes conditional flow matching to enhance training-time alignment signals. 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. 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. 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. 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.
Sources in this cluster
- Flow-Guided Target-Space Alignment via Path Consistency
arXiv cs.AI · 7 Nov 2026, 04:44 AEST