# M-VAR: Decoupled Scale-wise Autoregressive Modeling for High-Quality Image Generation

arXiv: https://arxiv.org/abs/2411.10433 · HTML: https://mlsanity.com/paper/2411.10433/

## Key facts

- Cited by 28 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 15 Nov 2024.

## Built on by (most cited first, 28 of 28)

- [Unified Multimodal Understanding and Generation Models: Advances, Challenges, and Opportunities](https://mlsanity.com/paper/2505.02567/) (2025)
- [FlowAR: Scale-wise Autoregressive Image Generation Meets Flow Matching](https://mlsanity.com/paper/2412.15205/) (2024)
- [Beyond Next-Token: Next-X Prediction for Autoregressive Visual Generation](https://mlsanity.com/paper/2502.20388/) (2025)
- [Scaling Laws in Patchification: An Image Is Worth 50,176 Tokens And More](https://mlsanity.com/paper/2502.03738/) (2025)
- [FlexVAR: Flexible Visual Autoregressive Modeling without Residual Prediction](https://mlsanity.com/paper/2502.20313/) (2025)
- [Lumos-1: On Autoregressive Video Generation with Discrete Diffusion from a Unified Model Perspective](https://mlsanity.com/paper/2507.08801/) (2025)
- [Head-Aware KV Cache Compression for Efficient Visual Autoregressive Modeling](https://mlsanity.com/paper/2504.09261/) (2025)
- [RestoreVAR: Visual Autoregressive Generation for All-in-One Image Restoration](https://mlsanity.com/paper/2505.18047/) (2025)
- [HIERAMP: Coarse-to-Fine Autoregressive Amplification for Generative Dataset Distillation](https://mlsanity.com/paper/2603.06932/) (2026)
- [Frequency-Aware Flow Matching for High-Quality Image Generation](https://mlsanity.com/paper/2604.15521/) (2026)
- [ViT-Linearizer: Distilling Quadratic Knowledge into Linear-Time Vision Models](https://mlsanity.com/paper/2504.00037/) (2025)
- [High-Fidelity Functional Ultrasound Reconstruction via A Visual Auto-Regressive Framework](https://mlsanity.com/paper/2505.21530/) (2025)
- [Multi-Scale Local Speculative Decoding for Image Generation](https://mlsanity.com/paper/2601.05149/) (2026)
- [Markovian Scale Prediction: A New Era of Visual Autoregressive Generation](https://mlsanity.com/paper/2511.23334/) (2025)
- [Autoregressive Image Generation Needs Only a Few Lines of Cached Tokens](https://mlsanity.com/paper/2512.04857/) (2025)
- [Knowledge Distillation for Visual Autoregressive Models](https://mlsanity.com/paper/2606.06078/) (2026)
- [Next-Scale Autoregressive Models are Zero-Shot Single-Image Object View Synthesizers](https://mlsanity.com/paper/2503.13588/) (2025)
- [RNN as Linear Transformer: A Closer Investigation into Representational Potentials of Visual Mamba Models](https://mlsanity.com/paper/2511.18380/) (2025)
- [LSRS: Latent Scale Rejection Sampling for Visual Autoregressive Modeling](https://mlsanity.com/paper/2512.03796/) (2025)
- [Adaptive Visual Autoregressive Acceleration via Dual-Linkage Entropy Analysis](https://mlsanity.com/paper/2602.01345/) (2026)
- [HACK++: Towards More Effective Head-Aware Key-Value Compression for Efficient Visual Autoregressive Modeling](https://mlsanity.com/paper/2606.08302/) (2026)
- [DiverseVAR: Balancing Diversity and Quality of Next-Scale Visual Autoregressive Models](https://mlsanity.com/paper/2511.21415/) (2025)
- [Rethinking Training Dynamics in Scale-wise Autoregressive Generation](https://mlsanity.com/paper/2512.06421/) (2025)
- [Depth Adaptive Efficient Visual Autoregressive Modeling](https://mlsanity.com/paper/2604.17286/) (2026)
- [Edit the Bits, Diff the Codes: Bitwise Residual Editing for Visual Autoregressive Models](https://mlsanity.com/paper/2606.13558/) (2026)
- [MotionMAR: Multi-scale Auto-Regressive Human Motion Reconstruction from Sparse Observations](https://mlsanity.com/paper/2606.23000/) (2026)
- [Photorealistic Novel View Synthesis of Human Faces using Next-Scale Transformers](https://mlsanity.com/paper/2608.23410/) (2026)
- [Logit Refiner: Improving Visual Autoregressive Models via Intra-Scale Dependency Modeling](https://mlsanity.com/paper/2609.11804/) (2026)

## Abstract (arXiv)

There exists recent work in computer vision, named VAR, that proposes a new autoregressive paradigm for image generation. Diverging from the vanilla next-token prediction, VAR structurally reformulates the image generation into a coarse to fine next-scale prediction. In this paper, we show that this scale-wise autoregressive framework can be effectively decoupled into textitintra-scale modeling, which captures local spatial dependencies within each scale, and textitinter-scale modeling, which models cross-scale relationships progressively from coarse-to-fine scales. This decoupling structure allows to rebuild VAR in a more computationally efficient manner. Specifically, for intra-scale modeling -- crucial for generating high-fidelity images -- we retain the original bidirectional self-attention design to ensure comprehensive modeling; for inter-scale modeling, which semantically connects different scales but is computationally intensive, we apply linear-complexity mechanisms like Mamba to substantially reduce computational overhead. We term this new framework M-VAR. Extensive experiments demonstrate that our method outperforms existing models in both image quality and generation speed. For example, our 1.5B model, with fewer parameters and faster inference speed, outperforms the largest VAR-d30-2B. Moreover, our largest model M-VAR-d32 impressively registers 1.78 FID on ImageNet 256×256 and outperforms the prior-art autoregressive models LlamaGen/VAR by 0.4/0.19 and popular diffusion models LDM/DiT by 1.82/0.49, respectively. Code is avaiable at urlhttps://github.com/OliverRensu/MVAR.