Look Less, Reason More: Block-wise Attention Skipping for Efficient Multimodal LLMs

• URL: http://arxiv.org/abs/2606.08511v1
• Date: Sun, 07 Jun 2026 08:32:13 GMT
• Title: Look Less, Reason More: Block-wise Attention Skipping for Efficient Multimodal LLMs
• Abstract summary: We present Visual-Skip, a training-free inference paradigm that decouples spatial interaction from semantic evolution.<n>V-Skip effectively bypasses redundant vision attention to achieve block-wise sparsity, maintaining a 94.16% to 100.31% performance retention across diverse MLLMs.
• Score: 89.7106332677868
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multimodal Large Language Models (MLLMs) face a significant inference bottleneck due to the quadratic computational cost of self-attention over long visual token sequences. However, we identify a critical inefficiency in current architectures: Visual Attention Saturation. Our analysis reveals that visual tokens rapidly establish their spatial structure and intra-modal relationships in early layers, rendering visual-to-visual self-attention in deeper layers computationally redundant. Conversely, Feed-Forward Networks (FFNs) in these layers remain essential for projecting visual features into the evolving textual semantic space. Leveraging this insight, we present Visual-Skip (V-Skip), a training-free inference paradigm that decouples spatial interaction from semantic evolution. Rather than discarding tokens, V-Skip imposes block-wise structured sparsity by selectively bypassing saturated visual self-attention modules. Furthermore, recognizing that varying downstream tasks demand distinct reasoning depths, V-Skip employs a lightweight, few-shot calibration to dynamically route the task-optimal sparsity path. Extensive experiments demonstrate that V-Skip effectively bypasses redundant vision attention to achieve block-wise sparsity, maintaining a 94.16% to 100.31% performance retention across diverse MLLMs. Ultimately, we prove that to reason more effectively, models do not need to discard what they see -- they simply need to "look less" at the right depth.

Related papers

• Capturing Token Tendencies for Training-Free Token Pruning in Multimodal Large Language Models [86.02826269126308]
  We propose Trend-aware Pruning, a novel framework that elevates pruning from a local snapshot decision to a temporal trajectory modeling problem.<n>This enables a dynamic rectification mechanism that selectively reactivates "late-blooming" tokens, those initially undervalued but exhibiting rising semantic importance, thereby preventing the loss of critical visual cues.
  arXiv  Detail & Related papers  (2026-07-30T15:07:00Z)
• On the Limits of Token Reduction for Efficient Unified Vision Language Training [15.69600993792123]
  Unified vision-language models (VLMs) integrate visual understanding and visual generation within a single autoregressive backbone.<n>We study the feasibility and limits of token-reduction-based acceleration for unified VLM training.
  arXiv  Detail & Related papers  (2026-05-31T23:59:12Z)
• Do Vision Language Models Need to Process Image Tokens? [37.581930147059445]
  Vision Language Models (VLMs) have achieved remarkable success by integrating visual encoders with large language models (LLMs)<n>It remains fundamentally unclear whether sustained image-token processing is necessary for their performance or visual representations meaningfully evolve from early to later layers.
  arXiv  Detail & Related papers  (2026-04-10T15:38:00Z)
• VISion On Request: Enhanced VLLM efficiency with sparse, dynamically selected, vision-language interactions [51.41587958253802]
  We introduce VISion On Request (VISOR), a method that reduces inference cost without discarding visual information.<n>VISOR improves efficiency by sparsifying the interaction between image and text tokens.<n>Experiments show that VISOR drastically reduces computational cost while matching or exceeding state-of-the-art results.
  arXiv  Detail & Related papers  (2026-03-24T17:58:17Z)
• Look Carefully: Adaptive Visual Reinforcements in Multimodal Large Language Models for Hallucination Mitigation [51.743225614196774]
  Multimodal large language models (MLLMs) have achieved remarkable progress in vision-language reasoning.<n>They remain vulnerable to hallucination, where generated content deviates from visual evidence.<n>Recent vision enhancement methods attempt to address this issue by reinforcing visual tokens during decoding.<n>We propose Adaptive Visual Reinforcement (AIR), a training-free framework for MLLMs.
  arXiv  Detail & Related papers  (2026-02-27T14:18:51Z)
• Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokens [54.18057944158818]
  Chain-of-Visual-Thought (COVT) is a framework that enables Vision-Language Models (VLMs) to reason through continuous visual tokens.<n>Within a small budget of roughly 20 tokens, COVT distills knowledge from lightweight vision experts.<n>During training, the VLM with COVT autoregressively predicts visual tokens to reconstruct dense supervision signals.
  arXiv  Detail & Related papers  (2025-11-24T18:55:19Z)
• CoViPAL: Layer-wise Contextualized Visual Token Pruning for Large Vision-Language Models [75.88232735646018]
  Large Vision-Language Models (LVLMs) process multimodal inputs consisting of text tokens and vision tokens extracted from images or videos.<n>Existing methods attempt to prune redundant vision tokens, revealing substantial redundancy in visual representations.<n>We propose CoViPAL, a layer-wise contextualized visual token pruning method that employs a Plug-and-Play Pruning Module (PPM) to predict and remove redundant vision tokens before they are processed by the LVLM.
  arXiv  Detail & Related papers  (2025-08-24T07:47:00Z)
• Rethinking Visual Token Reduction in LVLMs under Cross-modal Misalignment [38.04426918886084]
  Vision-Language Models (LVLMs) encode visual inputs as dense sequences of patch-level tokens to capture fine-grained semantics.<n>Previous efforts have explored visual token reduction either prior to or within the large language models (LLMs)<n>We introduce VisionDrop, a training-free, visual-only pruning framework that selects informative visual tokens based on intra-modal (visual-to-visual) attention.
  arXiv  Detail & Related papers  (2025-06-27T14:55:40Z)
• AdaFV: Rethinking of Visual-Language alignment for VLM acceleration [7.9213473377478865]
  Some approaches to reduce the visual tokens according to the self-attention of VLMs, which are biased, result in inaccurate responses.<n>We propose a self-adaptive cross-modality attention mixture mechanism that dynamically leverages the effectiveness of visual saliency and text-to-image similarity.<n>The proposed approach achieves state-of-the-art training-free VLM acceleration performance, especially when the reduction rate is sufficiently large.
  arXiv  Detail & Related papers  (2025-01-16T13:34:33Z)

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.