DLM-One: Diffusion Language Models for One-Step Sequence Generation

TL;DR

DLM-One introduces a score-distillation framework for one-step sequence generation with continuous diffusion language models, significantly improving inference speed while maintaining performance.

Contribution

It presents a novel one-step diffusion approach that eliminates iterative refinement, enabling faster language generation with continuous diffusion models.

Findings

Achieves up to ~500x inference speedup

Maintains competitive performance on benchmark tasks

Demonstrates generality across multiple datasets

Abstract

This paper introduces DLM-One, a score-distillation-based framework for one-step sequence generation with continuous diffusion language models (DLMs). DLM-One eliminates the need for iterative refinement by aligning the scores of a student model's outputs in the continuous token embedding space with the score function of a pretrained teacher DLM. We investigate whether DLM-One can achieve substantial gains in sampling efficiency for language modeling. Through comprehensive experiments on DiffuSeq -- a representative continuous DLM -- we show that DLM-One achieves up to ~500x speedup in inference time while maintaining competitive performance on benchmark text generation tasks used to evaluate the teacher models. We further analyze the method's empirical behavior across multiple datasets, providing initial insights into its generality and practical applicability. Our findings position one-step diffusion as a promising direction for efficient, high-quality language generation and broader adoption of continuous diffusion models operating in embedding space for natural language processing.