Scaling Categorical Flow Maps

• URL: http://arxiv.org/abs/2605.07820v2
• Date: Mon, 11 May 2026 13:50:18 GMT
• Title: Scaling Categorical Flow Maps
• Abstract summary: Continuous diffusion and flow matching models could represent a powerful alternative to autoregressive approaches for language modelling.<n>We show how to generate diverse, high-quality text in as few as $4$ inference steps while maintaining near-data-level token entropy.<n>We uncover some of the challenges that arise from training these models at scale, and we provide prescriptive insights on loss weighting and time scheduling.
• Score: 29.25832277355597
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Continuous diffusion and flow matching models could represent a powerful alternative to autoregressive approaches for language modelling (LM), as they unlock a host of advantages currently reserved for continuous modalities, including accelerated sampling and tilting. Recently, several works have demonstrated the possibility of generating discrete data continuously by a simple flow matching process between a Gaussian and the one-hot encoded data distribution. They have further shown the feasibility of accelerated sampling via Categorical Flow Maps (CFMs), resulting in competitive sample quality in the few-step regime. However, this method had only been evaluated at relatively modest scales ($<1$B), leaving the question of its scalability completely open. In this article, we train a $1.7$B-parameter base flow model on $2.1$T tokens and self-distill it into a CFM that generates diverse, high-quality text in as few as $4$ inference steps while maintaining near-data-level token entropy. Furthermore, we introduce a likelihood bound for CFMs in the semi-discrete setting, and show that they can be used to score the model on standard LM benchmarks, achieving results in the same range as discrete diffusion methods. Finally, we uncover some of the challenges that arise from training these models at scale, and we provide prescriptive insights on loss weighting and time scheduling.

Related papers

• Synergizing Transport-Based Generative Models and Latent Geometry for Stochastic Closure Modeling [1.665466637453776]
  We show that flow matching in a lower-dimensional latent space is suited for fast sampling of closure models.<n>We control the latent space distortion and thus ensure the physical fidelity of the sampled closure term.
  arXiv  Detail & Related papers  (2026-02-19T05:24:00Z)
• Joint Distillation for Fast Likelihood Evaluation and Sampling in Flow-based Models [100.28111930893188]
  Some of today's best generative models still require hundreds to thousands of neural function evaluations to compute a single likelihood.<n>We present fast flow joint distillation (F2D2), a framework that simultaneously reduces the number of NFEs required for both sampling and likelihood evaluation by two orders of magnitude.<n>F2D2 is modular, compatible with existing flow-based few-step sampling models, and requires only an additional divergence prediction head.
  arXiv  Detail & Related papers  (2025-12-02T10:48:20Z)
• Inference-Time Scaling of Diffusion Language Models with Particle Gibbs Sampling [70.8832906871441]
  We study how to steer generation toward desired rewards without retraining the models.<n>Prior methods typically resample or filter within a single denoising trajectory, optimizing rewards step-by-step without trajectory-level refinement.<n>We introduce particle Gibbs sampling for diffusion language models (PG-DLM), a novel inference-time algorithm enabling trajectory-level refinement while preserving generation perplexity.
  arXiv  Detail & Related papers  (2025-07-11T08:00:47Z)
• Almost Linear Convergence under Minimal Score Assumptions: Quantized Transition Diffusion [25.542593757387095]
  We propose Quantized Transition Diffusion (QTD), a novel approach that integrates data quantization with discrete diffusion dynamics.<n>Our method first transforms the continuous data distribution $p_*$ into a discrete one $q_*$ via histogram approximation and binary encoding.<n>For reverse-time sampling, we introduce a textittruncated uniformization technique to simulate the reverse CTMC.
  arXiv  Detail & Related papers  (2025-05-28T02:10:11Z)
• Feynman-Kac Correctors in Diffusion: Annealing, Guidance, and Product of Experts [64.34482582690927]
  We provide an efficient and principled method for sampling from a sequence of annealed, geometric-averaged, or product distributions derived from pretrained score-based models.<n>We propose Sequential Monte Carlo (SMC) resampling algorithms that leverage inference-time scaling to improve sampling quality.
  arXiv  Detail & Related papers  (2025-03-04T17:46:51Z)
• Local Flow Matching Generative Models [19.859984725284896]
  Flow Matching (FM) is a simulation-free method for learning a continuous and invertible flow to interpolate between two distributions.<n>We introduce a stepwise FM model called Local Flow Matching (LFM), which consecutively learns a sequence of FM sub-models.<n>We empirically demonstrate improved training efficiency and competitive generative performance of LFM compared to FM.
  arXiv  Detail & Related papers  (2024-10-03T14:53:10Z)
• Discrete Flow Matching [74.04153927689313]
  We present a novel discrete flow paradigm designed specifically for generating discrete data. Our approach is capable of generating high-quality discrete data in a non-autoregressive fashion.
  arXiv  Detail & Related papers  (2024-07-22T12:33:27Z)
• Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion [56.38386580040991]
  Consistency Trajectory Model (CTM) is a generalization of Consistency Models (CM) CTM enables the efficient combination of adversarial training and denoising score matching loss to enhance performance. Unlike CM, CTM's access to the score function can streamline the adoption of established controllable/conditional generation methods.
  arXiv  Detail & Related papers  (2023-10-01T05:07:17Z)
• Bayesian Flow Networks [4.197165999892042]
  This paper introduces Bayesian Flow Networks (BFNs), a new class of generative model in which the parameters of a set of independent distributions are modified with Bayesian inference.<n>Starting from a simple prior and iteratively updating the two distributions yields a generative procedure similar to the reverse process of diffusion models.<n>BFNs achieve competitive log-likelihoods for image modelling on dynamically binarized MNIST and CIFAR-10, and outperform all known discrete diffusion models on the text8 character-level language modelling task.
  arXiv  Detail & Related papers  (2023-08-14T09:56:35Z)
• Score-based Continuous-time Discrete Diffusion Models [102.65769839899315]
  We extend diffusion models to discrete variables by introducing a Markov jump process where the reverse process denoises via a continuous-time Markov chain. We show that an unbiased estimator can be obtained via simple matching the conditional marginal distributions. We demonstrate the effectiveness of the proposed method on a set of synthetic and real-world music and image benchmarks.
  arXiv  Detail & Related papers  (2022-11-30T05:33:29Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.