# MMInference: Accelerating Pre-filling for Long-Context VLMs via Modality-Aware Permutation Sparse Attention

arXiv: https://arxiv.org/abs/2504.16083 · HTML: https://mlsanity.com/paper/2504.16083/

## Key facts

- Cited by 26 papers in the map (data through 14 Oct 2026).
- Placed in Visual token pruning and efficiency, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 52 of its 58 references that are in the map.
- First posted on arXiv 22 Apr 2025.

## Placed in

- Multimodal LLMs / vision-language models / [Visual token pruning and efficiency](https://mlsanity.com/mllm/visual-token-pruning-and-efficiency/)

## Builds on (most cited first, 30 of 52)

- [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://mlsanity.com/paper/2103.14030/) (2021)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context](https://mlsanity.com/paper/2403.05530/) (2024)
- [Mistral 7B](https://mlsanity.com/paper/2310.06825/) (2023)
- [FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning](https://mlsanity.com/paper/2307.08691/) (2023)
- [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)
- [Video-MME: The First-Ever Comprehensive Evaluation Benchmark of Multi-modal LLMs in Video Analysis](https://mlsanity.com/paper/2405.21075/) (2024)
- [Video-ChatGPT: Towards Detailed Video Understanding via Large Vision and Language Models](https://mlsanity.com/paper/2306.05424/) (2023)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [SnapKV: LLM Knows What You are Looking for Before Generation](https://mlsanity.com/paper/2404.14469/) (2024)
- [NExT-QA:Next Phase of Question-Answering to Explaining Temporal Actions](https://mlsanity.com/paper/2105.08276/) (2021)
- [Qwen2.5-Omni Technical Report](https://mlsanity.com/paper/2503.20215/) (2025)
- [EgoSchema: A Diagnostic Benchmark for Very Long-form Video Language Understanding](https://mlsanity.com/paper/2308.09126/) (2023)
- [An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Models](https://mlsanity.com/paper/2403.06764/) (2024)
- [GAIA-1: A Generative World Model for Autonomous Driving](https://mlsanity.com/paper/2309.17080/) (2023)
- [LLaVA-Video: Video Instruction Tuning With Synthetic Data](https://mlsanity.com/paper/2410.02713/) (2024)
- [ActivityNet-QA: A Dataset for Understanding Complex Web Videos via Question Answering](https://mlsanity.com/paper/1906.02467/) (2019)
- [Long Context Transfer from Language to Vision](https://mlsanity.com/paper/2406.16852/) (2024)
- [Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention](https://mlsanity.com/paper/2502.11089/) (2025)
- [MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attention](https://mlsanity.com/paper/2407.02490/) (2024)
- [LongVU: Spatiotemporal Adaptive Compression for Long Video-Language Understanding](https://mlsanity.com/paper/2410.17434/) (2024)
- [Vista: A Generalizable Driving World Model with High Fidelity and Versatile Controllability](https://mlsanity.com/paper/2405.17398/) (2024)
- [Perception Test: A Diagnostic Benchmark for Multimodal Video Models](https://mlsanity.com/paper/2305.13786/) (2023)
- [VisionZip: Longer is Better but Not Necessary in Vision Language Models](https://mlsanity.com/paper/2412.04467/) (2024)
- [LongVILA: Scaling Long-Context Visual Language Models for Long Videos](https://mlsanity.com/paper/2408.10188/) (2024)
- [GR-2: A Generative Video-Language-Action Model with Web-Scale Knowledge for Robot Manipulation](https://mlsanity.com/paper/2410.06158/) (2024)
- [Zero-Shot Robotic Manipulation with Pretrained Image-Editing Diffusion Models](https://mlsanity.com/paper/2310.10639/) (2023)
- [Neighborhood Attention Transformer](https://mlsanity.com/paper/2204.07143/) (2022)

## Built on by (most cited first, 26 of 26)

- [Sparse VideoGen2: Accelerate Video Generation with Sparse Attention via Semantic-Aware Permutation](https://mlsanity.com/paper/2505.18875/) (2025)
- [Reinforcement Learning Optimization for Large-Scale Learning: An Efficient and User-Friendly Scaling Library](https://mlsanity.com/paper/2506.06122/) (2025)
- [LiveVLM: Efficient Online Video Understanding via Streaming-Oriented KV Cache and Retrieval](https://mlsanity.com/paper/2505.15269/) (2025)
- [Video-XL-2: Towards Very Long-Video Understanding Through Task-Aware KV Sparsification](https://mlsanity.com/paper/2506.19225/) (2025)
- [Bidirectional Sparse Attention for Faster Video Diffusion Training](https://mlsanity.com/paper/2509.01085/) (2025)
- [Mixing Importance with Diversity: Joint Optimization for KV Cache Compression in Large Vision-Language Models](https://mlsanity.com/paper/2510.20707/) (2025)
- [VideoNSA: Native Sparse Attention Scales Video Understanding](https://mlsanity.com/paper/2510.02295/) (2025)
- [Sparse Forcing: Native Trainable Sparse Attention for Real-time Autoregressive Diffusion Video Generation](https://mlsanity.com/paper/2604.21221/) (2026)
- [Compact Attention: Exploiting Structured Spatio-Temporal Sparsity for Fast Video Generation](https://mlsanity.com/paper/2508.12969/) (2025)
- [TimeScope: Towards Task-Oriented Temporal Grounding In Long Videos](https://mlsanity.com/paper/2509.26360/) (2025)
- [Sparser Block-Sparse Attention via Token Permutation](https://mlsanity.com/paper/2510.21270/) (2025)
- [One Token per Highly Selective Frame: Towards Extreme Compression for Long Video Understanding](https://mlsanity.com/paper/2604.14149/) (2026)
- [SpecVLM: Fast Speculative Decoding in Vision-Language Models](https://mlsanity.com/paper/2509.11815/) (2025)
- [Sparsity Forcing: Reinforcing Token Sparsity of MLLMs](https://mlsanity.com/paper/2504.18579/) (2025)
- [OmniSparse: Training-Aware Fine-Grained Sparse Attention for Long-Video MLLMs](https://mlsanity.com/paper/2511.12201/) (2025)
- [POP: Prefill-Only Pruning for Efficient Large Model Inference](https://mlsanity.com/paper/2602.03295/) (2026)
- [VecAttention: Vector-wise Sparse Attention for Accelerating Long Context Inference](https://mlsanity.com/paper/2603.29494/) (2026)
- [AdaptInfer: Adaptive Token Pruning for Vision-Language Model Inference with Dynamical Text Guidance](https://mlsanity.com/paper/2508.06084/) (2025)
- [MTraining: Distributed Dynamic Sparse Attention for Efficient Ultra-Long Context Training](https://mlsanity.com/paper/2510.18830/) (2025)
- [S2O: Early Stopping for Sparse Attention via Online Permutation](https://mlsanity.com/paper/2602.22575/) (2026)
- [BlindSight: Harnessing Sparsity for Efficient Vision-Language Models](https://mlsanity.com/paper/2507.09071/) (2025)
- [The Hybrid Multimodal Graph Index (HMGI): A Comprehensive Framework for Integrated Relational and Vector Search](https://mlsanity.com/paper/2510.10123/) (2025)
- [Rethinking Inference Placement for Deep Learning across Edge and Cloud Platforms: A Multi-Objective Optimization Perspective and Future Directions](https://mlsanity.com/paper/2510.22909/) (2025)
- [APB-V: Accelerating Long-Video Understanding via Sequence-Parallelism-aware Approximate Attention](https://mlsanity.com/paper/2601.21444/) (2026)
- [AdaSpark: Adaptive Sparsity for Efficient Long-Video Understanding](https://mlsanity.com/paper/2604.08077/) (2026)
- [HHR: Hierarchical Hash Retrieval for Efficient LLM Generation](https://mlsanity.com/paper/2610.01230/) (2026)

## Abstract (arXiv)

The integration of long-context capabilities with visual understanding unlocks unprecedented potential for Vision Language Models (VLMs). However, the quadratic attention complexity during the pre-filling phase remains a significant obstacle to real-world deployment. To overcome this limitation, we introduce MMInference (Multimodality Million tokens Inference), a dynamic sparse attention method that accelerates the prefilling stage for long-context multi-modal inputs. First, our analysis reveals that the temporal and spatial locality of video input leads to a unique sparse pattern, the Grid pattern. Simultaneously, VLMs exhibit markedly different sparse distributions across different modalities. We introduce a permutation-based method to leverage the unique Grid pattern and handle modality boundary issues. By offline search the optimal sparse patterns for each head, MMInference constructs the sparse distribution dynamically based on the input. We also provide optimized GPU kernels for efficient sparse computations. Notably, MMInference integrates seamlessly into existing VLM pipelines without any model modifications or fine-tuning. Experiments on multi-modal benchmarks-including Video QA, Captioning, VisionNIAH, and Mixed-Modality NIAH-with state-of-the-art long-context VLMs (LongVila, LlavaVideo, VideoChat-Flash, Qwen2.5-VL) show that MMInference accelerates the pre-filling stage by up to 8.3x at 1M tokens while maintaining accuracy. Our code is available at https://aka.ms/MMInference.