# Make It Efficient: Dynamic Sparse Attention for Autoregressive Image Generation

arXiv: https://arxiv.org/abs/2506.18226 · HTML: https://mlsanity.com/paper/2506.18226/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Placed in Fast autoregressive visual generation, one of 12 directions in Efficient LLM inference and serving.
- Builds on 35 of its 37 references that are in the map.
- First posted on arXiv 23 Jun 2025.

## Placed in

- Efficient LLM inference and serving / [Fast autoregressive visual generation](https://mlsanity.com/inference/fast-autoregressive-visual-generation/)

## Builds on (most cited first, 30 of 35)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Neural Discrete Representation Learning](https://mlsanity.com/paper/1711.00937/) (2017)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [Zero-Shot Text-to-Image Generation](https://mlsanity.com/paper/2102.12092/) (2021)
- [Qwen Technical Report](https://mlsanity.com/paper/2309.16609/) (2023)
- [Improved Techniques for Training GANs](https://mlsanity.com/paper/1606.03498/) (2016)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [Generating Long Sequences with Sparse Transformers](https://mlsanity.com/paper/1904.10509/) (2019)
- [MaskGIT: Masked Generative Image Transformer](https://mlsanity.com/paper/2202.04200/) (2022)
- [Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction](https://mlsanity.com/paper/2404.02905/) (2024)
- [Autoregressive Model Beats Diffusion: Llama for Scalable Image Generation](https://mlsanity.com/paper/2406.06525/) (2024)
- [Vector-quantized Image Modeling with Improved VQGAN](https://mlsanity.com/paper/2110.04627/) (2021)
- [Autoregressive Image Generation without Vector Quantization](https://mlsanity.com/paper/2406.11838/) (2024)
- [Scaling up GANs for Text-to-Image Synthesis](https://mlsanity.com/paper/2303.05511/) (2023)
- [MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attention](https://mlsanity.com/paper/2407.02490/) (2024)
- [When Attention Sink Emerges in Language Models: An Empirical View](https://mlsanity.com/paper/2410.10781/) (2024)
- [InfLLM: Training-Free Long-Context Extrapolation for LLMs with an Efficient Context Memory](https://mlsanity.com/paper/2402.04617/) (2024)
- [UniTok: A Unified Tokenizer for Visual Generation and Understanding](https://mlsanity.com/paper/2502.20321/) (2025)
- [Fluid: Scaling Autoregressive Text-to-image Generative Models with Continuous Tokens](https://mlsanity.com/paper/2410.13863/) (2024)
- [RetrievalAttention: Accelerating Long-Context LLM Inference via Vector Retrieval](https://mlsanity.com/paper/2409.10516/) (2024)
- [ImageFolder: Autoregressive Image Generation with Folded Tokens](https://mlsanity.com/paper/2410.01756/) (2024)
- [RandAR: Decoder-only Autoregressive Visual Generation in Random Orders](https://mlsanity.com/paper/2412.01827/) (2024)
- [Neighboring Autoregressive Modeling for Efficient Visual Generation](https://mlsanity.com/paper/2503.10696/) (2025)
- [Frequency Autoregressive Image Generation with Continuous Tokens](https://mlsanity.com/paper/2503.05305/) (2025)
- [Subobject-level Image Tokenization](https://mlsanity.com/paper/2402.14327/) (2024)
- [SemHiTok: A Unified Image Tokenizer via Semantic-Guided Hierarchical Codebook for Multimodal Understanding and Generation](https://mlsanity.com/paper/2503.06764/) (2025)

## Built on by (most cited first, 7 of 7)

- [Group Critical-token Policy Optimization for Autoregressive Image Generation](https://mlsanity.com/paper/2509.22485/) (2025)
- [Geometry-Aware Implicit Memory for Video World Models](https://mlsanity.com/paper/2606.02436/) (2026)
- [Autoregressive Image Generation Needs Only a Few Lines of Cached Tokens](https://mlsanity.com/paper/2512.04857/) (2025)
- [TourPhysics: Bringing Physics to World Models for Exploration and Manipulation from a Single Image](https://mlsanity.com/paper/2609.04911/) (2026)
- [Search-to-World: Evaluation of 3D World Delivery from User Request through Web Search](https://mlsanity.com/paper/2609.07605/) (2026)
- [Uncertainty DMD: Restoring Diversity in Few-Step Autoregressive Video Distillation](https://mlsanity.com/paper/2609.11265/) (2026)
- [Understanding and Exploiting Diagonal Attention Sparsity in Autoregressive Image Generation](https://mlsanity.com/paper/2609.19702/) (2026)

## Abstract (arXiv)

Autoregressive conditional image generation models have emerged as a dominant paradigm in text-to-image synthesis. These methods typically convert images into one-dimensional token sequences and leverage the self-attention mechanism, which has achieved remarkable success in natural language processing, to capture long-range dependencies, model global context, and ensure semantic coherence. However, excessively long contexts during inference lead to significant memory overhead caused by KV-cache and computational delays. To alleviate these challenges, we systematically analyze how global semantics, spatial layouts, and fine-grained textures are formed during inference, and propose a novel training-free context optimization method called Adaptive Dynamic Sparse Attention (ADSA). Conceptually, ADSA dynamically identifies historical tokens crucial for maintaining local texture consistency and those essential for ensuring global semantic coherence, thereby efficiently streamlining attention computation. Additionally, we introduce a dynamic KV-cache update mechanism tailored for ADSA, reducing GPU memory consumption during inference by approximately 50%. Extensive qualitative and quantitative experiments demonstrate the effectiveness and superiority of our approach in terms of both generation quality and resource efficiency.