To Sink or Not to Sink: Visual Information Pathways in Large Vision-Language Models

• URL: http://arxiv.org/abs/2510.08510v1
• Date: Thu, 09 Oct 2025 17:44:42 GMT
• Title: To Sink or Not to Sink: Visual Information Pathways in Large Vision-Language Models
• Authors: Jiayun Luo, Wan-Cyuan Fan, Lyuyang Wang, Xiangteng He, Tanzila Rahman, Purang Abolmaesumi, Leonid Sigal,
• Abstract summary: Vision Transformer (ViT) encodes visual content into a sequence of image tokens.<n>Large Language Model (LLM) interprets these tokens to perform high-level reasoning.<n>We identify a class of high-norm visual tokens from ViT, referred to as ViT attention sinks.
• Score: 34.902254997726835
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large Vision Language Models (LVLMs) have recently emerged as powerful architectures capable of understanding and reasoning over both visual and textual information. These models typically rely on two key components: a Vision Transformer (ViT) and a Large Language Model (LLM). ViT encodes visual content into a sequence of image tokens and serves as the perceptual front-end -- the eyes of the model. In contrast, the LLM interprets these tokens to perform high-level reasoning, generates responses, and functions as the cognitive core -- the brain of the model. However, it remains unclear which visual tokens contribute most significantly to understanding and reasoning, and how effectively these signals are propagated from ViT to the LLM. While most existing works have focused on identifying attention sinks, low-semantic tokens receiving disproportionately high attention, within the LLM, we shift the focus to the vision encoder by identifying a class of high-norm visual tokens from ViT, referred to as ViT attention sinks -- a problem that has been rarely studied but is indeed very important for LVLMs. Our findings show that these ViT sinks encapsulate high-level semantic concepts from images, allowing the LLM to perform more effective understanding and reasoning. Despite their importance, these sink tokens are often overlooked in existing LVLM architectures. To explore their contribution, we present both qualitative and quantitative analyses of the information embedded in these sink tokens. We also propose both training-free and training-based approaches to better leverage how this information is interpreted by the LLM, and to what extent. By explicitly utilizing these tokens, we demonstrate substantial improvements across a range of LVLMs and visual reasoning tasks, highlighting the untapped potential of ViT attention sinks in enhancing visual reasoning.

Related papers

• Seeing Clearly, Reasoning Confidently: Plug-and-Play Remedies for Vision Language Model Blindness [36.09698262750699]
  We introduce an efficient plug-and-play module that substantially improves vision language models' reasoning over rare objects.<n>We learn multi-modal class embeddings for rare objects by leveraging prior knowledge from vision foundation models and synonym-augmented text descriptions.<n> Experiments on two benchmarks show consistent and substantial gains for pretrained VLMs in rare object recognition and reasoning.
  arXiv  Detail & Related papers  (2026-02-23T09:02:40Z)
• Unleashing the Intrinsic Visual Representation Capability of Multimodal Large Language Models [58.91911788912665]
  We propose Latent Visual Reconstruction (LaVer), a novel training framework that facilitates MLLMs in learning more discrimi visual representations.<n>Our method offers direct visual activation to MLLMs, which exhibit increased visual attention allocation, indicating enhanced utilization of visual information.
  arXiv  Detail & Related papers  (2025-12-06T04:20:13Z)
• Attention Guided Alignment in Efficient Vision-Language Models [56.20286899428444]
  Large Vision-Language Models (VLMs) rely on effective multimodal alignment between pre-trained vision encoders and Large Language Models (LLMs)<n>This paper presents a comprehensive analysis of attention patterns in efficient VLMs.<n>We introduce Attention-Guided Efficient Vision-Language Models (AGE-VLM), a novel framework that enhances visual grounding through interleaved cross-attention layers.
  arXiv  Detail & Related papers  (2025-11-21T21:36:48Z)
• HERO: Rethinking Visual Token Early Dropping in High-Resolution Large Vision-Language Models [60.028070589466445]
  We propose HERO, a framework that integrates content-adaptive token budget allocation with function-aware token selection.<n>This study provides both empirical insights and practical solutions toward efficient inference in HR-LVLMs.
  arXiv  Detail & Related papers  (2025-09-16T13:22:08Z)
• Revisit What You See: Disclose Language Prior in Vision Tokens for LVLM Decoding [6.612630497074871]
  Large Vision-Language Models (LVLMs) achieve strong performance across multimodal tasks by integrating visual perception with language understanding.<n>We propose ReVisiT, a training-free decoding method that references vision tokens to guide text generation.
  arXiv  Detail & Related papers  (2025-06-11T08:46:55Z)
• PAINT: Paying Attention to INformed Tokens to Mitigate Hallucination in Large Vision-Language Model [0.0]
  hallucinations often arise from the progressive weakening of attention weight to visual tokens.<n>textbfPAINT (textbfPaying textbfAttention to textbfINformed textbfTokens) is a plug-and-play framework that intervenes in the self-attention mechanism of the Large Vision Language Models.
  arXiv  Detail & Related papers  (2025-01-21T15:22:31Z)
• [CLS] Token Tells Everything Needed for Training-free Efficient MLLMs [66.5266435598799]
  Multi-language Large Language Models (MLLMs) have recently demonstrated strong performance across a wide range of vision tasks.<n>However, their efficient deployment remains a substantial challenge due to high computational costs and memory requirements.<n>We introduce a simple yet effective method for train-free visual compression, called VTC- compression.
  arXiv  Detail & Related papers  (2024-12-08T05:29:39Z)
• Beyond Sight: Towards Cognitive Alignment in LVLM via Enriched Visual Knowledge [24.538839144639653]
  Large Vision-Language Models (LVLMs) integrate separately pre-trained vision and language components. These models frequently encounter a core issue of "cognitive misalignment" between the vision encoder (VE) and the large language model (LLM)
  arXiv  Detail & Related papers  (2024-11-25T18:33:14Z)
• FoPru: Focal Pruning for Efficient Large Vision-Language Models [11.36025001578531]
  We propose Focal Pruning (FoPru), a training-free method that prunes visual tokens based on the attention-based token significance derived from the vision encoder. Our method can prune a large number of redundant tokens while maintaining high accuracy, leading to significant improvements in inference efficiency.
  arXiv  Detail & Related papers  (2024-11-21T14:22:38Z)
• Break the Visual Perception: Adversarial Attacks Targeting Encoded Visual Tokens of Large Vision-Language Models [15.029014337718849]
  Large vision-language models (LVLMs) integrate visual information into large language models, showcasing remarkable multi-modal conversational capabilities. In general, LVLMs rely on vision encoders to transform images into visual tokens, which are crucial for the language models to perceive image contents effectively. We propose a non-targeted attack method referred to as VT-Attack, which constructs adversarial examples from multiple perspectives.
  arXiv  Detail & Related papers  (2024-10-09T09:06:56Z)
• Towards Semantic Equivalence of Tokenization in Multimodal LLM [149.11720372278273]
  Vision tokenization is essential for semantic alignment between vision and language.<n>This paper proposes a novel dynamic Semantic-Equivalent Vision Tokenizer (SeTok)<n>SeTok groups visual features into semantic units via a dynamic clustering algorithm.<n>The resulting vision tokens effectively preserve semantic integrity and capture both low-frequency and high-frequency visual features.
  arXiv  Detail & Related papers  (2024-06-07T17:55:43Z)
• Unified Language-Vision Pretraining in LLM with Dynamic Discrete Visual Tokenization [52.935150075484074]
  We introduce a well-designed visual tokenizer to translate the non-linguistic image into a sequence of discrete tokens like a foreign language. The resulting visual tokens encompass high-level semantics worthy of a word and also support dynamic sequence length varying from the image. This unification empowers LaVIT to serve as an impressive generalist interface to understand and generate multi-modal content simultaneously.
  arXiv  Detail & Related papers  (2023-09-09T03:01:38Z)

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.