inference/research
verifiedhigh · 0.92updated 2026-10-03

Diffusion-LM Improves Controllable Text Generation

paper

Fields

id
sources/arch-diffusion-lm
URL
https://arxiv.org/abs/2205.14217
Publisher
Li et al. (arXiv)
Kind
paper
Published
2022-05-27
Accessed
2026-10-03
notes
The original continuous-diffusion text LM: iteratively denoises a sequence of Gaussian vectors into word vectors. Establishes the non-autoregressive diffusion-LM inference shape that LLaDA and Dream discretise.

Notes

The original continuous-diffusion text LM: iteratively denoises a sequence of Gaussian vectors into word vectors. Establishes the non-autoregressive diffusion-LM inference shape that LLaDA and Dream discretise.

Cross-references

This record references no other records.