LGAIMay 17

Drift Flow Matching

arXiv:2605.1724489.4
Predicted impact top 5% in LG · last 90 daysOriginality Highly original
AI Analysis

For generative modeling practitioners, DFM provides a flexible framework that adapts sampling computation to different quality-efficiency requirements, bridging a gap between one-step and iterative methods.

Drift Flow Matching (DFM) connects one-step drift models with multi-step flow matching, enabling adjustable inference computation for quality-efficiency trade-offs. Experiments show effectiveness across tasks and datasets.

Iterative generative models such as Flow Matching and Diffusion models have demonstrated 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. DFM preserves the efficiency of direct transport maps while enabling generation to be refined through multiple inference steps when desired. This bridges the gap between one-step Drift Models and multi-step Flow Matching methods, and provides a novel generative paradigm that can adapt sampling computation to different quality--efficiency requirements. Extensive experiments across different tasks and datasets demonstrate the effectiveness and generality of the proposed framework.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes