Categorical Flow Maps

TL;DR

Categorical Flow Maps introduce a flow-matching approach for accelerated categorical data generation, leveraging continuous trajectories and self-distillation to improve efficiency and performance across various data types.

Contribution

The paper presents a novel continuous flow-matching method for categorical data that enables fast, accurate generation with existing distillation and guidance techniques.

Findings

State-of-the-art few-step results on images, molecular graphs, and text.

Strong performance in single-step generation.

Effective use of continuous trajectories for categorical data.

Abstract

We introduce Categorical Flow Maps, a flow-matching method for accelerated few-step generation of categorical data via self-distillation. Building on recent variational formulations of flow matching and the broader trend towards accelerated inference in diffusion and flow-based models, we define a flow map towards the simplex that transports probability mass toward a predicted endpoint, yielding a parametrisation that naturally constrains model predictions. Since our trajectories are continuous rather than discrete, Categorical Flow Maps can be trained with existing distillation techniques, as well as a new objective based on endpoint consistency. This continuous formulation also automatically unlocks test-time inference: we can directly reuse existing guidance and reweighting techniques in the categorical setting to steer sampling toward downstream objectives. Empirically, we achieve state-of-the-art few-step results on images, molecular graphs, and text, with strong performance even in single-step generation.