VLMs Need Words: Vision Language Models Ignore Visual Detail In Favor of Semantic Anchors

• URL: http://arxiv.org/abs/2604.02486v1
• Date: Thu, 02 Apr 2026 19:40:56 GMT
• Title: VLMs Need Words: Vision Language Models Ignore Visual Detail In Favor of Semantic Anchors
• Abstract summary: Vision Language Models (VLMs) achieve impressive performance across a wide range of multimodal tasks.<n>However, they often fail even when the required information is present in their internal representations.<n>We show that this gap arises from their narrow training pipeline which focuses on moving visual information to the textual space.
• Score: 14.288057170664983
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Vision Language Models (VLMs) achieve impressive performance across a wide range of multimodal tasks. However, on some tasks that demand fine-grained visual perception, they often fail even when the required information is present in their internal representations. In this work, we demonstrate that this gap arises from their narrow training pipeline which focuses on moving visual information to the textual space. Consequently, VLMs can only reason about visual entities that can be mapped to known concepts in the language space, leaving vision-focused tasks such as visual correspondence and reasoning about novel visual entities poorly supported. As a result, VLMs are severely limited in several important multimodal capabilities because they rely on brittle, hallucinated textual descriptions of visual entities that they cannot map to textual representations. We verify this behavior through visual correspondence tasks, in which VLMs must detect matching entities between two images. Testing across semantic, shape, and face correspondence tasks, we find that VLMs perform much better when the relevant entities are nameable in language than when they are unnameable. Mechanistically, our Logit Lens analyses confirm that VLMs explicitly assign semantic labels to nameable entities and surface more unique corresponding tokens compared to unnameable entities. Furthermore, we show that teaching completely arbitrary names for unknown entities improves performance, yet task-specific finetuning yields even stronger generalization without relying on language priors. Our findings suggest that current VLM failures on visual tasks reflect learned shortcuts from their training, rather than a fundamental limitation of multimodal architectures.

Related papers

• Seeing Beyond Redundancy: Task Complexity's Role in Vision Token Specialization in VLLMs [4.469560591280309]
  Vision capabilities in vision large language models (VLLMs) have consistently lagged behind their linguistic capabilities.<n>We investigate how various types of visual information are processed by the model and what types of visual information are discarded.
  arXiv  Detail & Related papers  (2026-02-06T18:13:01Z)
• 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)
• Hidden in plain sight: VLMs overlook their visual representations [48.83628674170634]
  We compare vision language models (VLMs) to their visual encoders to understand their ability to integrate across these modalities.<n>We find that VLMs perform substantially worse than their visual encoders, dropping to near-chance performance.
  arXiv  Detail & Related papers  (2025-06-09T17:59:54Z)
• Vision language models are unreliable at trivial spatial cognition [0.2902243522110345]
  Vision language models (VLMs) are designed to extract relevant visuospatial information from images.<n>We develop a benchmark dataset -- TableTest -- whose images depict 3D scenes of objects arranged on a table, and used it to evaluate state-of-the-art VLMs.<n>Results show that performance could be degraded by minor variations of prompts that use equivalent descriptions.
  arXiv  Detail & Related papers  (2025-04-22T17:38:01Z)
• Teaching VLMs to Localize Specific Objects from In-context Examples [56.797110842152]
  We find that present-day Vision-Language Models (VLMs) lack a fundamental cognitive ability: learning to localize specific objects in a scene by taking into account the context.<n>This work is the first to explore and benchmark personalized few-shot localization for VLMs.
  arXiv  Detail & Related papers  (2024-11-20T13:34:22Z)
• Losing Visual Needles in Image Haystacks: Vision Language Models are Easily Distracted in Short and Long Contexts [65.04791072532106]
  We present LoCoVQA, a benchmark generator for evaluating long-context extractive reasoning in vision language models (VLMs) LoCoVQA augments test examples for mathematical reasoning, VQA, and character recognition tasks with increasingly long visual contexts. This test assesses how well VLMs can ignore irrelevant information when answering queries.
  arXiv  Detail & Related papers  (2024-06-24T17:58:03Z)
• ClawMachine: Learning to Fetch Visual Tokens for Referential Comprehension [71.03445074045092]
  We propose ClawMachine, offering a new methodology that explicitly notates each entity using token collectives groups of visual tokens.<n>Our method unifies the prompt and answer of visual referential tasks without using additional syntax.<n>ClawMachine achieves superior performance on scene-level and referential understanding tasks with higher efficiency.
  arXiv  Detail & Related papers  (2024-06-17T08:39:16Z)
• 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.