mlsanityData to 7 Oct 2026

VLM in a flash: I/O-Efficient Sparsification of Vision-Language Model via Neuron Chunking

arXiv 2511.18692 ↗First posted 24 Nov 2025cs.LG cs.AI cs.CVcited by 0

On the map: Multimodal LLMs / vision-language models → Visual token pruning and efficiency

Abstract (arXiv)

Edge deployment of large Vision-Language Models (VLMs) increasingly relies on flash-based weight offloading, where activation sparsification is used to reduce I/O overhead. However, conventional sparsification remains model-centric, selecting neurons solely by activation magnitude and neglecting how access patterns influence flash performance. We present Neuron Chunking, an I/O-efficient sparsification strategy that operates on chunks (i.e., groups of contiguous neurons in memory) and couples neuron importance with storage access cost. The method models I/O latency through a lightweight abstraction of access contiguity and selects chunks with high utility, defined as neuron importance normalized by estimated latency. By aligning sparsification decisions with the underlying storage behavior, Neuron Chunking improves I/O efficiency by up to 4.65x and 5.76x on Jetson Orin Nano and Jetson AGX Orin, respectively.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Visual token pruning and efficiency, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 34 of its 56 references that are in the map. First posted on arXiv 24 Nov 2025.

Builds on

34 of 56 references.

References · 34

  1. 2024Qwen2.5 Technical Reportcited by 4,713
  2. 2022FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awarenesscited by 3,611
  3. 2016Gaussian Error Linear Units (GELUs)cited by 3,431
  4. 2024LLaVA-OneVision: Easy Visual Task Transfercited by 2,669
  5. 2015Learning both Weights and Connections for Efficient Neural Networkscited by 1,998
  6. 2020GLU Variants Improve Transformercited by 1,724
  7. 2022SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Modelscited by 1,338
  8. 2023AWQ: Activation-aware Weight Quantization for LLM Compression and Accelerationcited by 1,234
  9. 2021NExT-QA:Next Phase of Question-Answering to Explaining Temporal Actionscited by 735
  10. 2023VILA: On Pre-training for Visual Language Modelscited by 686
  11. 2023LLM-Pruner: On the Structural Pruning of Large Language Modelscited by 643
  12. 2023FlexGen: High-Throughput Generative Inference of Large Language Models with a Single GPUcited by 567
22 more
  1. 2019Patient Knowledge Distillation for BERT Model Compressioncited by 502
  2. 2024Long Context Transfer from Language to Visioncited by 502
  3. 2022DeepSpeed Inference: Enabling Efficient Inference of Transformer Models at Unprecedented Scalecited by 403
  4. 2024TempCompass: Do Video LLMs Really Understand Videos?cited by 352
  5. 2024VideoLLM-online: Online Video Large Language Model for Streaming Videocited by 216
  6. 2023PowerInfer: Fast Large Language Model Serving with a Consumer-grade GPUcited by 210
  7. 2024NVILA: Efficient Frontier Visual Language Modelscited by 206
  8. 2023LLM in a flash: Efficient Large Language Model Inference with Limited Memorycited by 170
  9. 2024LongVLM: Efficient Long Video Understanding via Large Language Modelscited by 146
  10. 2022The Lazy Neuron Phenomenon: On Emergence of Activation Sparsity in Transformerscited by 121
  11. 2022The Optimal BERT Surgeon: Scalable and Accurate Second-Order Pruning for Large Language Modelscited by 115
  12. 2024PowerInfer-2: Fast Large Language Model Inference on a Smartphonecited by 101
  13. 2023ReLU Strikes Back: Exploiting Activation Sparsity in Large Language Modelscited by 91
  14. 2020Contrastive Distillation on Intermediate Representations for Language Model Compressioncited by 57
  15. 2024Training-Free Activation Sparsity in Large Language Modelscited by 51
  16. 2024ProSparse: Introducing and Enhancing Intrinsic Activation Sparsity within Large Language Modelscited by 43
  17. 2024CATS: Contextually-Aware Thresholding for Sparsity in Large Language Modelscited by 41
  18. 2021Rethinking Network Pruning -- under the Pre-train and Fine-tune Paradigmcited by 40
  19. 2024Turbo Sparse: Achieving LLM SOTA Performance with Minimal Activated Parameterscited by 37
  20. 2020Meta-KD: A Meta Knowledge Distillation Framework for Language Model Compression across Domainscited by 33
  21. 2024Galaxy: A Resource-Efficient Collaborative Edge AI System for In-situ Transformer Inferencecited by 25
  22. 2022STI: Turbocharge NLP Inference at the Edge via Elastic Pipeliningcited by 14

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025CoreMatching: A Co-adaptive Sparse Inference Framework with Token and Neuron Pruning for Comprehensive Acceleration of Vision-Language Models
  2. 2024VL-Cache: Sparsity and Modality-Aware KV Cache Compression for Vision-Language Model Inference Acceleration
  3. 2025SLIM: A Heterogeneous Accelerator for Edge Inference of Sparse Large Language Model via Adaptive Thresholding
  4. 2026NeuronEye: Query-Guided Visual Concept Activation for Vision-Language Reasoning
  5. 2026Mind the Spike: Mechanisms and Brittleness of Visual Massive Activations in Large Vision-Language Models
  6. 2025Flash-VL 2B: Optimizing Vision-Language Model Performance for Ultra-Low Latency and High Throughput
  7. 2025BlindSight: Harnessing Sparsity for Efficient Vision-Language Models
  8. 2025Sparse Neuron Ablation Triggers Catastrophic Collapse of the Language Core in Large Vision-Language Models