Visual Representations inside the Language Model

• URL: http://arxiv.org/abs/2510.04819v1
• Date: Mon, 06 Oct 2025 14:01:39 GMT
• Title: Visual Representations inside the Language Model
• Authors: Benlin Liu, Amita Kamath, Madeleine Grunde-McLaughlin, Winson Han, Ranjay Krishna,
• Abstract summary: We study flow of visual information through the language model, finding that image value tokens encode sufficient information to perform several perception-heavy tasks.<n>We find that while the language model does augment the visual information received from the projection of input visual encodings, it contains less visual information on several tasks than the equivalent visual encoder (SigLIP)<n>Next, we discuss controlling visual information in the language model, showing that adding a text prefix to the image input improves perception capabilities of visual representations.
• Score: 36.35124375782294
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Despite interpretability work analyzing VIT encoders and transformer activations, we don't yet understand why Multimodal Language Models (MLMs) struggle on perception-heavy tasks. We offer an under-studied perspective by examining how popular MLMs (LLaVA-OneVision, Qwen2.5-VL, and Llama-3-LLaVA-NeXT) process their visual key-value tokens. We first study the flow of visual information through the language model, finding that image value tokens encode sufficient information to perform several perception-heavy tasks zero-shot: segmentation, semantic correspondence, temporal correspondence, and referring expression detection. We find that while the language model does augment the visual information received from the projection of input visual encodings-which we reveal correlates with overall MLM perception capability-it contains less visual information on several tasks than the equivalent visual encoder (SigLIP) that has not undergone MLM finetuning. Further, we find that the visual information corresponding to input-agnostic image key tokens in later layers of language models contains artifacts which reduce perception capability of the overall MLM. Next, we discuss controlling visual information in the language model, showing that adding a text prefix to the image input improves perception capabilities of visual representations. Finally, we reveal that if language models were able to better control their visual information, their perception would significantly improve; e.g., in 33.3% of Art Style questions in the BLINK benchmark, perception information present in the language model is not surfaced to the output! Our findings reveal insights into the role of key-value tokens in multimodal systems, paving the way for deeper mechanistic interpretability of MLMs and suggesting new directions for training their visual encoder and language model components.

Related papers

• 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)
• Perceiving Beyond Language Priors: Enhancing Visual Comprehension and Attention in Multimodal Models [1.9253106218929117]
  Multimodal Large Language Models (MLLMs) often fail to fully leverage visual input, defaulting to strong language priors.<n>Our approach first provides insights into how MLLMs internally build visual understanding of image regions and then introduces techniques to amplify this capability.<n>We demonstrate the superior multimodal understanding of our resultant model through a detailed upstream analysis quantifying its ability to predict visually-dependent tokens as well as 10 pt boost on visually challenging tasks.
  arXiv  Detail & Related papers  (2025-05-08T20:04:27Z)
• Lifting the Veil on Visual Information Flow in MLLMs: Unlocking Pathways to Faster Inference [28.24397677839652]
  Multimodal large language models (MLLMs) improve performance on vision-language tasks by integrating visual features from pre-trained vision encoders into large language models.<n>How MLLMs process and utilize visual information remains unclear.<n>We propose Hierarchical Modality-Aware Pruning (HiMAP), a plug-and-play inference acceleration method that dynamically prunes image tokens at specific layers, reducing computational costs by approximately 65% without sacrificing performance.
  arXiv  Detail & Related papers  (2025-03-17T12:31:23Z)
• Semantic-Clipping: Efficient Vision-Language Modeling with Semantic-Guidedd Visual Selection [53.558449071113245]
  Vision-Language Models (VLMs) leverage aligned visual encoders to transform images into visual tokens, allowing them to be processed similarly to text by the backbone large language model (LLM)<n>Recent advancements in vision-language modeling introduce image cropping techniques that feed all encoded sub-images into the model.<n>We propose a lightweight, universal framework that seamlessly integrates with existing VLMs to enhance their ability to process finegrained details.
  arXiv  Detail & Related papers  (2025-03-14T18:33:31Z)
• Towards Interpreting Visual Information Processing in Vision-Language Models [24.51408101801313]
  Vision-Language Models (VLMs) are powerful tools for processing and understanding text and images.<n>We study the processing of visual tokens in the language model component of LLaVA, a prominent VLM.
  arXiv  Detail & Related papers  (2024-10-09T17:55:02Z)
• ControlMLLM: Training-Free Visual Prompt Learning for Multimodal Large Language Models [73.34709921061928]
  We propose a training-free method to inject visual prompts into Multimodal Large Language Models (MLLMs)<n>We optimize a learnable latent variable based on an energy function, enhancing the strength of referring regions in the attention map.<n>Our method offers a promising direction for integrating referring abilities into MLLMs, and supports referring with box, mask, scribble and point.
  arXiv  Detail & Related papers  (2024-07-31T11:40:29Z)
• 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)
• Draw-and-Understand: Leveraging Visual Prompts to Enable MLLMs to Comprehend What You Want [58.091825321168514]
  We present the Draw-and-Understand framework, exploring how to integrate visual prompting understanding capabilities into Multimodal Large Language Models (MLLMs)<n>Visual prompts allow users to interact through multi-modal instructions, enhancing the models' interactivity and fine-grained image comprehension.<n>In this framework, we propose a general architecture adaptable to different pre-trained MLLMs, enabling it to recognize various types of visual prompts.
  arXiv  Detail & Related papers  (2024-03-29T16:26:20Z)
• Beyond Text: Frozen Large Language Models in Visual Signal Comprehension [34.398976855955404]
  Vision-to-Language Tokenizer, abbreviated as V2T Tokenizer, transforms an image into a foreign language'' with the combined aid of an encoder-decoder, the LLM vocabulary, and a CLIP model. We undertake rigorous experiments to validate our method, encompassing understanding tasks like image recognition, image captioning, and visual question answering, as well as image denoising tasks like inpainting, outpainting, deblurring, and shift restoration.
  arXiv  Detail & Related papers  (2024-03-12T17:59:51Z)
• Incorporating Visual Experts to Resolve the Information Loss in Multimodal Large Language Models [121.83413400686139]
  This paper proposes to improve the visual perception ability of MLLMs through a mixture-of-experts knowledge enhancement mechanism. We introduce a novel method that incorporates multi-task encoders and visual tools into the existing MLLMs training and inference pipeline.
  arXiv  Detail & Related papers  (2024-01-06T02:02:34Z)
• 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)
• Contextual Object Detection with Multimodal Large Language Models [66.15566719178327]
  We introduce a novel research problem of contextual object detection. Three representative scenarios are investigated, including the language cloze test, visual captioning, and question answering. We present ContextDET, a unified multimodal model that is capable of end-to-end differentiable modeling of visual-language contexts.
  arXiv  Detail & Related papers  (2023-05-29T17:50:33Z)
• VLMAE: Vision-Language Masked Autoencoder [21.97700040013084]
  We propose a vision-language masked autoencoder framework (VLMAE) for vision-language pre-training. VLMAE employs visual generative learning, facilitating the model to acquire fine-grained and unbiased features.
  arXiv  Detail & Related papers  (2022-08-19T14:39:18Z)
• Visually-Augmented Language Modeling [137.36789885105642]
  We propose a novel pre-training framework, named VaLM, to Visually-augment text tokens with retrieved relevant images for Language Modeling. With the visually-augmented context, VaLM uses a visual knowledge fusion layer to enable multimodal grounded language modeling. We evaluate the proposed model on various multimodal commonsense reasoning tasks, which require visual information to excel.
  arXiv  Detail & Related papers  (2022-05-20T13:41:12Z)

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.