Purrception: Variational Flow Matching for Vector-Quantized Image Generation

ICLR 2026 Conference Submission12228 Authors

Keywords: generative models, flow matching, vector quantized, image generation, computer vision, variational flow matching

TL;DR: We apply variational flow matching to VQ latent image generation through a hybrid discrete-continuous approach for improved image generation.

Abstract: We introduce Purrception, a variational flow matching approach for vector-quantized image generation that provides explicit categorical supervision while maintaining continuous transport dynamics. Our method adapts Variational Flow Matching to vector-quantized latents by learning categorical posteriors over codebook indices while computing velocity fields in the continuous embedding space. This combines the geometric awareness of continuous methods with the discrete supervision of categorical approaches, enabling uncertainty quantification over plausible codes and temperature-controlled generation. We evaluate Purrception on ImageNet-1k $256 \times 256$ generation. Training converges faster than both continuous flow matching and discrete flow matching baselines while achieving competitive FID scores with state-of-the-art models. This demonstrates that Variational Flow Matching can effectively bridge continuous transport and discrete supervision for improved training efficiency in image generation.

Primary Area: generative models

Submission Number: 12228