Q Cache: Visual Attention is Valuable in Less than Half of Decode Layers for Multimodal Large Language Model

• URL: http://arxiv.org/abs/2602.01901v1
• Date: Mon, 02 Feb 2026 10:08:00 GMT
• Title: Q Cache: Visual Attention is Valuable in Less than Half of Decode Layers for Multimodal Large Language Model
• Authors: Jiedong Zhuang, Lu Lu, Ming Dai, Rui Hu, Jian Chen, Qiang Liu, Haoji Hu,
• Abstract summary: Multimodal large language models (MLLMs) are plagued by exorbitant inference costs attributable to the profusion of visual tokens.<n>Existing approaches focus on token-wise optimization, leveraging diverse token pruning techniques to eliminate non-crucial visual tokens.<n>We propose Lazy Attention, an efficient attention mechanism that enables cross-layer sharing of similar attention patterns.
• Score: 21.206033754351786
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multimodal large language models (MLLMs) are plagued by exorbitant inference costs attributable to the profusion of visual tokens within the vision encoder. The redundant visual tokens engenders a substantial computational load and key-value (KV) cache footprint bottleneck. Existing approaches focus on token-wise optimization, leveraging diverse intricate token pruning techniques to eliminate non-crucial visual tokens. Nevertheless, these methods often unavoidably undermine the integrity of the KV cache, resulting in failures in long-text generation tasks. To this end, we conduct an in-depth investigation towards the attention mechanism of the model from a new perspective, and discern that attention within more than half of all decode layers are semantic similar. Upon this finding, we contend that the attention in certain layers can be streamlined by inheriting the attention from their preceding layers. Consequently, we propose Lazy Attention, an efficient attention mechanism that enables cross-layer sharing of similar attention patterns. It ingeniously reduces layer-wise redundant computation in attention. In Lazy Attention, we develop a novel layer-shared cache, Q Cache, tailored for MLLMs, which facilitates the reuse of queries across adjacent layers. In particular, Q Cache is lightweight and fully compatible with existing inference frameworks, including Flash Attention and KV cache. Additionally, our method is highly flexible as it is orthogonal to existing token-wise techniques and can be deployed independently or combined with token pruning approaches. Empirical evaluations on multiple benchmarks demonstrate that our method can reduce KV cache usage by over 35% and achieve 1.5x throughput improvement, while sacrificing only approximately 1% of performance on various MLLMs. Compared with SOTA token-wise methods, our technique achieves superior accuracy preservation.

Related papers

• Vision Token Reduction via Attention-Driven Self-Compression for Efficient Multimodal Large Language Models [34.12135666939555]
  Multimodal Large Language Models (MLLMs) incur significant computational cost from processing numerous vision tokens through all layers.<n>We introduce Attention-Driven Self-Compression (ADSC), a simple, broadly applicable method that progressively reduces vision tokens using only the LLM's attention mechanism.<n>ADSC reduces FLOPs by 53.7% and peak KV-cache memory by 56.7%, while preserving 98.2% of the original model performance.
  arXiv  Detail & Related papers  (2026-02-13T04:49:27Z)
• Don't Just Chase "Highlighted Tokens" in MLLMs: Revisiting Visual Holistic Context Retention [50.97683288777336]
  Multimodal Large Language Models (MLLMs) suffer from considerable computational overhead due to their reliance on massive visual tokens.<n>Recent studies have explored token pruning to alleviate this problem, which typically uses text-vision cross-attention.<n>We propose HoloV, a plug-and-play visual token pruning framework for efficient inference.
  arXiv  Detail & Related papers  (2025-10-03T11:33:40Z)
• LaCo: Efficient Layer-wise Compression of Visual Tokens for Multimodal Large Language Models [62.240460476785934]
  We propose LaCo (Layer-wise Visual Token Compression), a novel framework that enables effective token compression within the intermediate layers of the vision encoder.<n>LaCo introduces two core components: 1) a layer-wise pixel-shuffle mechanism that systematically merges adjacent tokens through space-to-channel transformations, and 2) a residual learning architecture with non-parametric shortcuts.
  arXiv  Detail & Related papers  (2025-07-03T03:42:54Z)
• Efficient LLaMA-3.2-Vision by Trimming Cross-attended Visual Features [24.33252753245426]
  We exploit the sparse nature in cross-attention maps to selectively prune redundant visual features.<n>Our model can reduce inference latency and memory usage while achieving benchmark parity.
  arXiv  Detail & Related papers  (2025-04-01T09:10:32Z)
• AirCache: Activating Inter-modal Relevancy KV Cache Compression for Efficient Large Vision-Language Model Inference [11.73134417321505]
  We propose AirCache, a novel KV cache compression method aimed at accelerating LVLMs inference.<n>We show that our method achieves comparable performance to the full cache while retaining only 10% of visual KV cache.
  arXiv  Detail & Related papers  (2025-03-31T11:13:18Z)
• TopV: Compatible Token Pruning with Inference Time Optimization for Fast and Low-Memory Multimodal Vision Language Model [56.43860351559185]
  We introduce textbfTopV, a compatible textbfTOken textbfPruning with inference Time Optimization for fast and low-memory textbfVLM.<n>Our framework incorporates a visual-aware cost function to measure the importance of each source visual token, enabling effective pruning of low-importance tokens.
  arXiv  Detail & Related papers  (2025-03-24T01:47:26Z)
• RedundancyLens: Revealing and Exploiting Visual Token Processing Redundancy for Efficient Decoder-Only MLLMs [38.34856927170692]
  We propose a training-free framework for analyzing trained Multimodal Large Language Model (MLLM)<n>It consists of Probe-Activated Dynamic FFN and Hollow Attention, which enable adjustable reductions in computations for visual tokens.<n>Experiments demonstrate substantial, structured, and clustered redundancy unique to decoder-only MLLMs.
  arXiv  Detail & Related papers  (2025-01-31T11:09:16Z)
• Cross-Self KV Cache Pruning for Efficient Vision-Language Inference [19.062950348441426]
  KV cache pruning has emerged as a promising technique for reducing memory and computation costs in long-context auto-regressive generation.<n>We propose decomposing attention scores into intra-modality attention (within the same modality) and inter-modality attention (across modalities)<n>Our final training-free method, textbfCross-textbfSelf textbfPruning (CSP), achieves competitive performance compared to models with full KV caches.
  arXiv  Detail & Related papers  (2024-12-05T22:47:17Z)
• PrefixKV: Adaptive Prefix KV Cache is What Vision Instruction-Following Models Need for Efficient Generation [97.41972925670508]
  Large vision-language models (LVLMs) incur significant computational and memory overhead during inference.<n>We present PrefixKV, where "Prefix" means the top-ranked KV based on importance rather than position in the original sequence.<n>Our method achieves the state-of-the-art performance compared with others.
  arXiv  Detail & Related papers  (2024-12-04T15:48:59Z)
• A Stitch in Time Saves Nine: Small VLM is a Precise Guidance for Accelerating Large VLMs [65.00970402080351]
  A promising approach to accelerating large vision-language models (VLMs) is using partial information, such as attention maps from specific layers, to assess token importance and prune less essential tokens.<n>Our study reveals three key insights: (i) Partial attention information is insufficient for accurately identifying critical visual tokens, resulting in suboptimal performance, especially at low token retention ratios; (ii) Global attention information, such as the attention map aggregated across all layers, more effectively preserves essential tokens and maintains comparable performance under aggressive pruning; and (iii) The global attention map aggregated from a small VLM closely resembles that of a large VLM,
  arXiv  Detail & Related papers  (2024-12-04T13:56:44Z)
• Anchor Attention, Small Cache: Code Generation with Large Language Models [15.94784908771546]
  Current practices in NLP often use sparse attention which may, unfortunately, lead to substantial inaccuracies, or hallucinations, in code generation tasks. We propose a novel approach, AnchorCoder, which features token-wise anchor attention designed to extract and compress contextual information. It can consistently achieve a significant (at least 70%) reduction in KV cache requirements, while preserving the majority of model's performance.
  arXiv  Detail & Related papers  (2024-11-11T02:47:05Z)
• ThinK: Thinner Key Cache by Query-Driven Pruning [63.13363917871414]
  Large Language Models (LLMs) have revolutionized the field of natural language processing, achieving unprecedented performance across a variety of applications.<n>This paper focuses on the long-context scenario, addressing the inefficiencies in KV cache memory consumption during inference.<n>We propose ThinK, a novel query-dependent KV cache pruning method designed to minimize attention weight loss while selectively pruning the least significant channels.
  arXiv  Detail & Related papers  (2024-07-30T17:59:08Z)
• Efficient Inference of Vision Instruction-Following Models with Elastic Cache [76.44955111634545]
  We introduce Elastic Cache, a novel strategy for efficient deployment of instruction-following large vision-language models. We propose an importance-driven cache merging strategy to prune redundancy caches. For instruction encoding, we utilize the frequency to evaluate the importance of caches. Results on a range of LVLMs demonstrate that Elastic Cache not only boosts efficiency but also notably outperforms existing pruning methods in language generation.
  arXiv  Detail & Related papers  (2024-07-25T15:29:05Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.