Dynamic Mixture-of-Experts for Visual Autoregressive Model

Jort Vincenti, Metod Jazbec, Guoxuan Xia

Keywords: Visual Autoregressive Models, Mixture-of-Experts, Dynamic Computation, Efficient Image Generation, Scale-Aware Inference

Abstract: Visual Autoregressive Models (VAR) offer efficient and high-quality image generation but suffer from computational redundancy due to repeated Transformer calls at increasing resolutions. We introduce a dynamic Mixture-of-Experts router integrated into VAR. The new architecture allows to trade compute for quality through scale-aware thresholding. This thresholding strategy balances expert selection based on token complexity and resolution, without requiring additional training. As a result, we achieve 20% fewer FLOPs, 11% faster inference and match the image quality achieved by the dense baseline.

Submission Number: 4