Simplax：通过单纯形松弛增强离散扩散模型

AI 摘要

Simplax是一种新的离散扩散模型增强方法，通过Dirichlet-categorical增强来丰富均匀离散扩散，在不改变底层分类损坏过程的情况下改进反向采样和生成质量。在OpenWebText生成任务上，Simplax改善了生成质量与熵的权衡；在Sudoku任务上，仅用30线索谜题训练的模型在所有线索密度下均达到最高准确率，包括17线索的最小唯一可解情况。

正文节选

Simplex Relaxation for Discrete Diffusion Abstract Simplax enriches uniform discrete diffusion via Dirichlet-categorical augmentation to improve reverse sampling and generative quality on text and Sudoku tasks. Discrete diffusion models for categorical generation are defined by a corruption kernel, which determines the intermediate state space and the associated reverse prediction problem. We study uniform discrete diffusion and ask whether its training objective and reverse transitions can be e

发布时间：—

抓取时间：2026-08-13 15:16

来源机构：Hugging Face