COFLOW：根据提示自适应选择步数，视觉生成提速超2.5倍
Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation
一篇论文提出 COFLOW，按 prompt 动态决定扩散步数，图像视频都能提速 2.5 倍还不用重训模型，做生成的可以看看。
arXiv 论文提出 COFLOW，一种推理时方法，根据 prompt 特征为每次生成自适应选择步数。该方法以无监督奖励在线训练，平衡推理效率与生成保真度，无需对底层生成模型重训练，即插即用。实验覆盖图像和视频生成，在保持感知与语义质量的同时实现超过 2.5 倍加速。论文还给出标准正则条件下的 O(1/K) 前向欧拉离散化误差界。
Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation
Flow Matching enables high-quality visual generation via continuous-time dynamics, but inference remains costly due to multiple sequential function evaluations. Existing acceleration methods reduce the number of function evaluations but often introduce additional training overhead, degrade quality, or fail to account for input-dependent variability. We propose COFLOW, an inference-time method that adaptively selects the step counts each generation based on the prompt features. Our context-aware COFLOW is trained online with an unsupervised reward that balances inference efficiency and generation fidelity. Our method is plug-and-play, requiring no retraining of the underlying generative model. It generalizes to image and video generation, achieving over 2.5x speedup while preserving perceptual and semantic quality. We further provide a theoretical analysis establishing an O(1/K) forward-Euler discretization error bound under standard regularity conditions.