A Reparameterized Discrete Diffusion Model for Text Generation

Lin Zheng, Jianbo Yuan, Lei Yu, Lingpeng Kong

Research Area: Compute efficient LMs, Learning algorithms for LMs, Inference algorithms for LMs

Keywords: discrete diffusion, text generation, non-autoregressive generation

TL;DR: This work develops a reparameterized discrete diffusion model that implies effective training and sampling algorithms and delivers significant performance improvements for text generation..

Abstract: This work studies discrete diffusion probabilistic models with applications to natural language generation. We derive an alternative yet equivalent formulation of the sampling from discrete diffusion processes and leverage this insight to develop a family of reparameterized discrete diffusion models. The derived generic framework is highly flexible, offers a fresh perspective of the generation process in discrete diffusion models, and features more effective training and decoding techniques. We conduct extensive experiments to evaluate the text generation capability of our model, demonstrating significant improvements over existing diffusion models.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the COLM Code of Ethics on https://colmweb.org/CoE.html

Author Guide: I certify that this submission complies with the submission instructions as described on https://colmweb.org/AuthorGuide.html

Submission Number: 554