[Resource] Awesome-Visual-Autoregressive-Modeling — A curated paper list built on VAR

[XIANGLONGYAN]

📚 Awesome Visual Autoregressive Modeling

Hi VAR team 👋

I've been following the VAR paper since its NeurIPS 2024 release and have been maintaining a curated reading list of all research inspired by or built upon the next-scale prediction paradigm.

Repository: https://github.com/XIANGLONGYAN/Awesome-Visual-Autoregressive-Modeling

---

What's inside

The list currently tracks 60+ papers organized in two ways:

By venue/year — Conference papers (NeurIPS, CVPR, ICLR, ICML, ICCV, AAAI) and arXiv preprints, from the original VAR (NeurIPS 2024) through ICLR 2026.

By research topic:

Topic	Example works
Image Generation	Infinity, HART, FlowAR, FlexVAR, MVAR
Efficiency & Acceleration	CoDe, FastVAR, SkipVAR, ScaleKV, FreqExit
Quantization & Compression	PTQ4ARVG, LiteVAR, Shift-and-Sum
Image Restoration & SR	VARFormer, VARSR, RestoreVAR, VARestorer
Controllable & Guided Gen	ControlVAR, CAR, VAREdit, AREdit
Video, 3D & Multi-modal	InfinityStar, SAR3D, VARGPT
Dense Prediction	Scalable Depth, Seg-VAR, DepthART
Safety & Watermarking	Safe-VAR, Closing the Safety Gap
Theory & Analysis	Fundamental Limits, VAR Beats Diffusion

---

Why I'm posting here

VAR is the foundational work that started this whole ecosystem. Researchers and practitioners who discover this repo often look for a broader map of the field. A reference in your README's Related Resources or Follow-up Works section would greatly help newcomers navigate the growing body of work.

If you're open to it, something like:

Community resources: Awesome Visual Autoregressive Modeling — a curated list of 60+ papers
built on the VAR paradigm.

I'm also happy to accept PRs / issues in my repo if there are papers or corrections you'd like to see added.

Thanks for the incredible work on VAR — it has sparked a genuinely exciting research direction!