Learning to See Before Seeing: Demystifying LLM Visual Priors from Language Pre-training

• URL: http://arxiv.org/abs/2509.26625v1
• Date: Tue, 30 Sep 2025 17:57:44 GMT
• Title: Learning to See Before Seeing: Demystifying LLM Visual Priors from Language Pre-training
• Authors: Junlin Han, Shengbang Tong, David Fan, Yufan Ren, Koustuv Sinha, Philip Torr, Filippos Kokkinos,
• Abstract summary: Large Language Models (LLMs) develop rich visual priors despite being trained on text alone.<n>These priors allow latent visual capabilities to be unlocked for vision tasks with a relatively small amount of multimodal data.<n>We show that visual priors are composed of separable perception and reasoning priors with unique scaling trends and origins.
• Score: 37.93241751782069
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large Language Models (LLMs), despite being trained on text alone, surprisingly develop rich visual priors. These priors allow latent visual capabilities to be unlocked for vision tasks with a relatively small amount of multimodal data, and in some cases, to perform visual tasks without ever having seen an image. Through systematic analysis, we reveal that visual priors-the implicit, emergent knowledge about the visual world acquired during language pre-training-are composed of separable perception and reasoning priors with unique scaling trends and origins. We show that an LLM's latent visual reasoning ability is predominantly developed by pre-training on reasoning-centric data (e.g., code, math, academia) and scales progressively. This reasoning prior acquired from language pre-training is transferable and universally applicable to visual reasoning. In contrast, a perception prior emerges more diffusely from broad corpora, and perception ability is more sensitive to the vision encoder and visual instruction tuning data. In parallel, text describing the visual world proves crucial, though its performance impact saturates rapidly. Leveraging these insights, we propose a data-centric recipe for pre-training vision-aware LLMs and verify it in 1T token scale pre-training. Our findings are grounded in over 100 controlled experiments consuming 500,000 GPU-hours, spanning the full MLLM construction pipeline-from LLM pre-training to visual alignment and supervised multimodal fine-tuning-across five model scales, a wide range of data categories and mixtures, and multiple adaptation setups. Along with our main findings, we propose and investigate several hypotheses, and introduce the Multi-Level Existence Bench (MLE-Bench). Together, this work provides a new way of deliberately cultivating visual priors from language pre-training, paving the way for the next generation of multimodal LLMs.

Related papers

• 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)
• Visual Jigsaw Post-Training Improves MLLMs [58.29961336087896]
  We introduce Visual Jigsaw, a generic self-supervised post-training framework designed to strengthen visual understanding in large language models (MLLMs)<n>Visual Jigsaw is formulated as a general ordering task: visual inputs are partitioned, shuffled, and the model must reconstruct the visual information by producing the correct permutation in natural language.<n>Extensive experiments demonstrate substantial improvements in fine-grained perception, temporal reasoning, and 3D spatial understanding.
  arXiv  Detail & Related papers  (2025-09-29T17:59:57Z)
• Visual Representation Alignment for Multimodal Large Language Models [38.319869213758686]
  Multimodal large language models (MLLMs) trained with visual instruction tuning have achieved strong performance across diverse tasks.<n>But they remain limited in vision-centric tasks such as object counting or spatial reasoning.<n>We present VIsual Representation ALignment (VIRAL), a simple yet effective regularization strategy that aligns the internal visual representations of MLLMs with those of pre-trained vision foundation models.
  arXiv  Detail & Related papers  (2025-09-09T17:59:14Z)
• 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)
• TWIST & SCOUT: Grounding Multimodal LLM-Experts by Forget-Free Tuning [54.033346088090674]
  We introduce TWIST & SCOUT, a framework that equips pre-trained MLLMs with visual grounding ability.<n>To fine-tune the model effectively, we generate a high-quality synthetic dataset we call SCOUT.<n>This dataset provides rich supervision signals, describing a step-by-step multimodal reasoning process.
  arXiv  Detail & Related papers  (2024-10-14T13:35:47Z)
• Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs [61.143381152739046]
  We introduce Cambrian-1, a family of multimodal LLMs (MLLMs) designed with a vision-centric approach.<n>Our study uses LLMs and visual instruction tuning as an interface to evaluate various visual representations.<n>We provide model weights, code, supporting tools, datasets, and detailed instruction-tuning and evaluation recipes.
  arXiv  Detail & Related papers  (2024-06-24T17:59:42Z)
• Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs [50.77984109941538]
  Our research reveals that the visual capabilities in recent multimodal LLMs still exhibit systematic shortcomings. We identify ''CLIP-blind pairs'' - images that CLIP perceives as similar despite their clear visual differences. We evaluate various CLIP-based vision-and-language models and found a notable correlation between visual patterns that challenge CLIP models and those problematic for multimodal LLMs.
  arXiv  Detail & Related papers  (2024-01-11T18:58:36Z)
• 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)

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.