# SparseMM: Head Sparsity Emerges from Visual Concept Responses in MLLMs

arXiv: https://arxiv.org/abs/2506.05344 · HTML: https://mlsanity.com/paper/2506.05344/

## Key facts

- Cited by 20 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 43 of its 54 references that are in the map.
- First posted on arXiv 5 Jun 2025.

## Placed in

- Multimodal LLMs / vision-language models / [Visual token pruning and efficiency](https://mlsanity.com/mllm/visual-token-pruning-and-efficiency/)
- Efficient LLM inference and serving / [Visual token pruning in multimodal models](https://mlsanity.com/inference/visual-token-pruning-in-multimodal-models/)

## Builds on (most cited first, 30 of 43)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [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)
- [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://mlsanity.com/paper/2301.12597/) (2023)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [Flamingo: a Visual Language Model for Few-Shot Learning](https://mlsanity.com/paper/2204.14198/) (2022)
- [Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context](https://mlsanity.com/paper/2403.05530/) (2024)
- [Sigmoid Loss for Language Image Pre-Training](https://mlsanity.com/paper/2303.15343/) (2023)
- [LLaVA-OneVision: Easy Visual Task Transfer](https://mlsanity.com/paper/2408.03326/) (2024)
- [InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks](https://mlsanity.com/paper/2312.14238/) (2023)
- [Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering](https://mlsanity.com/paper/1612.00837/) (2016)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [MMBench: Is Your Multi-modal Model an All-around Player?](https://mlsanity.com/paper/2307.06281/) (2023)
- [Qwen2 Technical Report](https://mlsanity.com/paper/2407.10671/) (2024)
- [Mixtral of Experts](https://mlsanity.com/paper/2401.04088/) (2024)
- [Towards VQA Models That Can Read](https://mlsanity.com/paper/1904.08920/) (2019)
- [Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling](https://mlsanity.com/paper/2412.05271/) (2024)
- [ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning](https://mlsanity.com/paper/2203.10244/) (2022)
- [Video-LLaVA: Learning United Visual Representation by Alignment Before Projection](https://mlsanity.com/paper/2311.10122/) (2023)
- [GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints](https://mlsanity.com/paper/2305.13245/) (2023)
- [DocVQA: A Dataset for VQA on Document Images](https://mlsanity.com/paper/2007.00398/) (2020)
- [MiniCPM-V: A GPT-4V Level MLLM on Your Phone](https://mlsanity.com/paper/2408.01800/) (2024)
- [SnapKV: LLM Knows What You are Looking for Before Generation](https://mlsanity.com/paper/2404.14469/) (2024)
- [H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models](https://mlsanity.com/paper/2306.14048/) (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)
- [LLaVA-NeXT-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal Models](https://mlsanity.com/paper/2407.07895/) (2024)
- [mPLUG-Owl2: Revolutionizing Multi-modal Large Language Model with Modality Collaboration](https://mlsanity.com/paper/2311.04257/) (2023)

## Built on by (most cited first, 20 of 20)

- [StreamKV: Streaming Video Question-Answering with Segment-based KV Cache Retrieval and Compression](https://mlsanity.com/paper/2511.07278/) (2025)
- [Mixing Importance with Diversity: Joint Optimization for KV Cache Compression in Large Vision-Language Models](https://mlsanity.com/paper/2510.20707/) (2025)
- [Efficient Inference for Large Vision-Language Models: Bottlenecks, Techniques, and Prospects](https://mlsanity.com/paper/2604.05546/) (2026)
- [Do MLLMs Really See It: Reinforcing Visual Attention in Multimodal LLMs](https://mlsanity.com/paper/2602.08241/) (2026)
- [Mechanistic Insights into Functional Sparsity in Multimodal LLMs via CoRe Heads](https://mlsanity.com/paper/2606.05843/) (2026)
- [HybridKV: Hybrid KV Cache Compression for Efficient Multimodal Large Language Model Inference](https://mlsanity.com/paper/2604.05887/) (2026)
- [Head-Aware Visual Cropping: Enhancing Fine-Grained VQA with Attention-Guided Subimage](https://mlsanity.com/paper/2601.22483/) (2026)
- [One Token, Two Fates: A Unified Framework via Vision Token Manipulation Against MLLMs Hallucination](https://mlsanity.com/paper/2603.10360/) (2026)
- [Bridging Visual Representation and Reinforcement Learning from Verifiable Rewards in Large Vision-Language Models](https://mlsanity.com/paper/2603.27375/) (2026)
- [AIA: Rethinking Architecture Decoupling Strategy In Unified Multimodal Model](https://mlsanity.com/paper/2511.22663/) (2025)
- [Last But Not Least: Boundary Attention CalibratiON for Multimodal KV Cache Compression](https://mlsanity.com/paper/2606.14782/) (2026)
- [VideoMM: Adaptive Macro-Micro Inference for Efficient Video MLLMs](https://mlsanity.com/paper/2609.16722/) (2026)
- [Salca: A Sparsity-Aware Hardware Accelerator for Efficient Long-Context Attention Decoding](https://mlsanity.com/paper/2604.24820/) (2026)
- [Vision-Core Guided Contrastive Learning for Balanced Multi-modal Prognosis Prediction of Stroke](https://mlsanity.com/paper/2605.14710/) (2026)
- [TGV-KV: Text-Grounded KV Eviction for Vision-Language Models](https://mlsanity.com/paper/2606.03075/) (2026)
- [MM-ShiftKV: Decode-Aware Prefill-Stage KV Selection for Multimodal Large Language Models](https://mlsanity.com/paper/2607.22586/) (2026)
- [WnW: Waxing-and-Waning KV Cache for Long-Form Speech LLMs](https://mlsanity.com/paper/2608.22704/) (2026)
- [ECOKV: Geometry-Aware KV Cache Eviction via Complementary Diversity Metrics](https://mlsanity.com/paper/2609.06663/) (2026)
- [MLLMs Hallucinate when Information Distribution Drifts in Synergy Heads](https://mlsanity.com/paper/2609.09206/) (2026)
- [HeadGuard: Selective Head Protection for Low-Bit VLM KV-Cache Quantization](https://mlsanity.com/paper/2609.35800/) (2026)

## Abstract (arXiv)

Multimodal Large Language Models (MLLMs) are commonly derived by extending pre-trained Large Language Models (LLMs) with visual capabilities. In this work, we investigate how MLLMs process visual inputs by analyzing their attention mechanisms. We reveal a surprising sparsity phenomenon: only a small subset (approximately less than 5%) of attention heads in LLMs actively contribute to visual understanding, termed visual heads. To identify these heads efficiently, we design a training-free framework that quantifies head-level visual relevance through targeted response analysis. Building on this discovery, we introduce SparseMM, a KV-Cache optimization strategy that allocates asymmetric computation budgets to heads in LLMs based on their visual scores, leveraging the sparity of visual heads for accelerating the inference of MLLMs. Compared with prior KV-Cache acceleration methods that ignore the particularity of visual, SparseMM prioritizes stress and retaining visual semantics during decoding. Extensive evaluations across mainstream multimodal benchmarks demonstrate that SparseMM achieves superior accuracy-efficiency trade-offs. Notably, SparseMM delivers 1.38x real-time acceleration and 52% memory reduction during generation while maintaining performance parity on efficiency test. Our project is open sourced at https://github.com/CR400AF-A/SparseMM.