Divergence-Suppressing Couplings for Rectified Flow
The paper discusses divergence-suppressing couplings for Rectified Flow, aimed at improving trajectory generation. It identifies issues with trajectory entanglement caused by nonzero divergence in the learned velocity field. The proposed method offers a correction that enhances performance without increasing computational costs during deployment.
- ▪The authors are Yimeng Min and Carla P. Gomes.
- ▪The proposed couplings aim to reduce distortion in generated trajectories.
- ▪Improvements were observed in both 2D synthetic benchmarks and image generation.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.17733 (cs) [Submitted on 18 May 2026] Title:Divergence-Suppressing Couplings for Rectified Flow Authors:Yimeng Min, Carla P. Gomes View a PDF of the paper titled Divergence-Suppressing Couplings for Rectified Flow, by Yimeng Min and 1 other authors View PDF HTML (experimental) Abstract:The promise of Rectified Flow rests on producing self-generated couplings whose trajectories are straight, or nearly so. In practice, trajectories generated by the base flow model can bend and intertwine, and the resulting coupling inherits this distortion.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
