🌟 FasterVAR

Plug-and-Play Acceleration for Visual Autoregressive Models

Senmao Li^1,3, Kai Wang², Salman Khan³, Fahad Shahbaz Khan^3,4, Jian Yang¹, Yaxing Wang¹,

¹ Nankai University, ² City University of Hong Kong (Dongguan), China, ³ MBZUAI, ⁴ Linkoping University

Figure 1. Overview of the proposed FasterVAR framework. We retain the original VAR inference process for the semantic and structure establishment stages, while exploiting semantic irrelevance and low-rank properties in the fidelity refinement stage to accelerate inference.

🖼️ Qualitative Results

Figure 2. Qualitative comparison with the vanilla Infinity-2B, Infinity-8B, and STAR models (1st, 3rd, and 5th rows). Our StageVAR (2nd, 4th, and 6th rows) achieves a 3.4x, 2.7x, and 1.74x speedup while maintaining performance.

📊 Quantitative Results

Quantitative Results on the GenEval and DPG benchmarks

📄 Citation

Please cite our paper if you find this work useful for your research:

@inproceedings{li2026icml,
  title     = {FasterVAR: Plug-and-Play Acceleration for Visual Autoregressive Models},
  author    = {Li, Senmao and Wang, Kai and Khan, Salman and Khan, Fahad Shahbaz and Yang, Jian and Wang, Yaxing},
  booktitle = {ICML},
  year      = {2026},
}

⭐ If FasterVAR is helpful to your projects, please help star this repo. Thanks! 🤗