Discriminative Flow Matching Via Local Generative Predictors

ArXi:2603.13928v1 Announce Type: cross Traditional discriminative computer vision relies predominantly on static projections, mapping input features to outputs in a single computational step. Although efficient, this paradigm lacks the iterative refinement and robustness inherent in biological vision and modern generative modelling. In this paper, we propose Discriminative Flow Matching, a framework that reformulates classification and object detection as a conditional transport process.