VILA-Lab/Awesome-DLMs

The official GitHub repo for the survey paper "A Survey on Diffusion Language Models".

What it solves

This project addresses the need for a centralized, up-to-date repository of research and resources regarding Diffusion Language Models (DLMs). It helps researchers and developers navigate the rapidly evolving landscape of text generation models that use diffusion techniques, which differ from traditional autoregressive models.

How it works

It acts as a curated knowledge hub. It organizes academic papers, code implementations, and project demos into logical categories such as "Discrete DLMs," "Continuous DLMs," "Inference Optimization," and "Training Strategies." The repository tracks the evolution of the field through a timeline of milestones and provides direct links to arXiv papers, GitHub repositories, and interactive web demos.

Who it’s for

This collection is designed for AI researchers, machine learning engineers, and students who want to stay current with the latest advancements in diffusion-based text generation, ranging from foundational theory to state-of-the-art inference acceleration techniques.

Highlights

  • A comprehensive, frequently updated list of research papers and codebases.
  • Categorized sections covering foundations, training, inference, and multimodal applications.
  • A dedicated "Playground" section featuring live demos of various diffusion language models.
  • A visual timeline illustrating the shift from continuous to discrete diffusion models.
  • Inclusion of cutting-edge research like DARTree, which improves decoding speed and efficiency.

Related

  • Project
    HazyResearch/ncm

    Next‑Concept Modeling (NCM) is a two‑stage, parallel text‑generation framework that uses a multiscale VQ tokenizer and a transformer to generate sequences coarse‑to‑fine, achieving 9‑31× speed‑ups while matching diffusion baselines on summarisation, translation, language modelling and Sudoku. The repository includes installation instructions, evaluation scripts, and a collection of pretrained tokenizers and generators for several datasets.

  • Project
    anthropics/anthropic-sdk-go

    A Go client library for calling Anthropic’s Claude LLM API, offering typed helpers for building prompts and handling responses.

  • Project
    OpenTSLM/OpenTSLM

    OpenTSLM is an open‑source library that adds native time‑series handling to Llama 3.2 and Gemma LLMs, enabling natural‑language prompting and reasoning over medical signals (ECG, EEG, accelerometer, etc.). It provides pre‑trained checkpoints, a simple `OpenTSLM` Python API, demo scripts for inference on several benchmark tasks, and a curriculum‑learning training pipeline covering MCQ, captioning, and chain‑of‑thought reasoning stages. Install via `pip install opentslm`, load models from Hugging‑Face, and fine‑tune or evaluate using the supplied scripts. Licensed under MIT.

  • Project
    qwdwqfwq/topconf-paper-figure-gallery

    A curated, searchable gallery of Figure 1/teaser images from six top AI/ML conferences (2023–2026), paired with FigureForge, a browser tool that retrieves reference figures via CLIP+BM25 and uses LLMs to draft a publication-style Figure 1 for your own paper.