AI RESEARCH
Drift Flow Matching
arXiv CS.LG
•
ArXi:2605.17244v1 Announce Type: new Iterative generative models such as Flow Matching and Diffusion models have nstrated strong test-time scaling behavior, where additional inference computation can improve generation quality. In contrast, Drift Models offer efficient one-step generation, but their direct generation paradigm limits such flexibility. In this work, we propose Drift Flow Matching (DFM), a framework that connects drifting generative modeling with flow-based iterative generation.