Diffusion-LM Improves Controllable Text Generation

TL;DR

This paper introduces Diffusion-LM, a non-autoregressive diffusion-based language model that enables fine-grained controllable text generation without re-training, outperforming previous methods on complex control tasks.

Contribution

The paper presents a novel diffusion-based language model that allows for complex, controllable text generation through a simple gradient-based method, addressing limitations of prior approaches.

Findings

Successfully controls six fine-grained text attributes

Outperforms prior methods significantly on control tasks

Demonstrates the effectiveness of diffusion models in NLP

Abstract

Controlling the behavior of language models (LMs) without re-training is a major open problem in natural language generation. While recent works have demonstrated successes on controlling simple sentence attributes (e.g., sentiment), there has been little progress on complex, fine-grained controls (e.g., syntactic structure). To address this challenge, we develop a new non-autoregressive language model based on continuous diffusions that we call Diffusion-LM. Building upon the recent successes of diffusion models in continuous domains, Diffusion-LM iteratively denoises a sequence of Gaussian vectors into word vectors, yielding a sequence of intermediate latent variables. The continuous, hierarchical nature of these intermediate variables enables a simple gradient-based algorithm to perform complex, controllable generation tasks. We demonstrate successful control of Diffusion-LM for six challenging fine-grained control tasks, significantly outperforming prior work.