Reproduce “Coupled Guidance for Flow Matching” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Mateo Clémente, Léo Maxime Brunswic, Rui Heng Yang, Amir Rasouli
Coupled guidance steers flow-matching generation by correcting both predicted data and latent endpoints at inference time, rather than changing only the data endpoint. It retains task-directed guidance while adjusting the latent correction to better align with the learned coupling. In ManiSkill3’s Pick Cube collision-avoidance test, success was 81% versus 36% across 256 seeds; this result concerns the experiment, not real-world deployment.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.