A Field Guide to the Diffusion Literature Jungle

What it does This repository is a manually curated bibliography of diffusion-model research papers, sorted into roughly two dozen subareas such as visual illusion, color control, object counting, poster generation, acceleration, inpainting, and continual learning. Each entry links to the paper, project page, and code when available. It is essentially a reading list maintained as a Markdown file.

The interesting bit The categorization is task-centric rather than chronological or architectural, which makes it useful when you know you want, say, “few-shot personalization” or “drag-based editing” but do not know the paper titles. The granularity is unusual: there are dedicated sections for niche problems like accurate object counting and virtual try-on, alongside broader headings such as image restoration and segmentation.

Key highlights

• Covers highly specific subareas (e.g., Count Guidance, Drag Edit, Remove Concept, Virtual Try On) alongside standard topics like super-resolution and layout generation.
• Entries include direct links to arXiv, project pages, and GitHub code repositories where they exist.
• Reflects recent conference and workshop publications (CVPR, NeurIPS, ECCV, ICLR, WACV, ACM MM).
• Maintained as a plain, browsable README with no dependencies or build steps.

Caveats

• The list is a compilation of external work; the repository contains no original code, models, or datasets.
• Some sections are sparsely populated or contain only paper titles without code links.
• A few commented-out sections (e.g., Text Generation, Video Generation, Video Editing) suggest the taxonomy is still expanding.

Verdict Worth bookmarking if you are a researcher or practitioner trying to navigate the flood of diffusion literature by task rather than by citation graph. Skip it if you are looking for a reusable framework or training scripts.