Abstract
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.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Bajpai, D. J., Verma, A., & Hanawal, M. K. (2026). Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation. https://omanscience.com/en/articles/contextual-flow-matching-adaptive-step-selection-in-flow-models-for-efficient-visual-generation
MLA 9
Bajpai, Divya Jyoti, et al. "Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation." https://omanscience.com/en/articles/contextual-flow-matching-adaptive-step-selection-in-flow-models-for-efficient-visual-generation.
Chicago (author–date)
Bajpai, Divya Jyoti, Arun Verma, and Manjesh Kumar Hanawal. 2026. "Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation." https://omanscience.com/en/articles/contextual-flow-matching-adaptive-step-selection-in-flow-models-for-efficient-visual-generation.
Harvard
Bajpai, D. J., Verma, A. and Hanawal, M. K. (2026) 'Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation', Available at: https://omanscience.com/en/articles/contextual-flow-matching-adaptive-step-selection-in-flow-models-for-efficient-visual-generation.
Vancouver
Bajpai DJ, Verma A, Hanawal MK. Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation. https://omanscience.com/en/articles/contextual-flow-matching-adaptive-step-selection-in-flow-models-for-efficient-visual-generation
IEEE
D. J. Bajpai, A. Verma, and M. K. Hanawal, "Contextual Flow Matching: Adaptive Step Selection in Flow Models for Efficient Visual Generation," https://omanscience.com/en/articles/contextual-flow-matching-adaptive-step-selection-in-flow-models-for-efficient-visual-generation.