# Diffusion LM - [Diffusion-LM Improves Controllable Text Generation](https://arxiv.org/abs/2205.14217) - Controlling the behavior of language models (LMs) without re-training is a major open problem in natural language generation - non-[Autoregressive](Autoregressive.md) language model based on continuous diffusions - substantial departure from the current paradigm of discrete [Autoregressive](Autoregressive.md) generation - iteratively denoises a sequence of Gaussian vectors into word vectors, yielding a sequence of intermediate latent variables - continuous, hierarchical nature of these intermediate variables enables a simple gradient-based algorithm to perform complex, controllable generation tasks - successful control of Diffusion-LM for six challenging fine-grained control tasks