Pith

open record

sign in

arxiv: 2411.17491 · v1 · pith:FUMMT2QM · submitted 2024-11-26 · cs.CV · cs.AI

What's in the Image? A Deep-Dive into the Vision of Vision Language Models

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel pith:FUMMT2QMrecord.jsonopen to challenge →

classification cs.CV cs.AI
keywords imagevisualtokenslayersmodelsvlmsinformationcomplex
0
0 comments X
read the original abstract

Vision-Language Models (VLMs) have recently demonstrated remarkable capabilities in comprehending complex visual content. However, the mechanisms underlying how VLMs process visual information remain largely unexplored. In this paper, we conduct a thorough empirical analysis, focusing on attention modules across layers. We reveal several key insights about how these models process visual data: (i) the internal representation of the query tokens (e.g., representations of "describe the image"), is utilized by VLMs to store global image information; we demonstrate that these models generate surprisingly descriptive responses solely from these tokens, without direct access to image tokens. (ii) Cross-modal information flow is predominantly influenced by the middle layers (approximately 25% of all layers), while early and late layers contribute only marginally.(iii) Fine-grained visual attributes and object details are directly extracted from image tokens in a spatially localized manner, i.e., the generated tokens associated with a specific object or attribute attend strongly to their corresponding regions in the image. We propose novel quantitative evaluation to validate our observations, leveraging real-world complex visual scenes. Finally, we demonstrate the potential of our findings in facilitating efficient visual processing in state-of-the-art VLMs.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Vision-Default, Prior-Override: Causal Mechanisms of Perception-Knowledge Conflict in Vision-Language Models

    cs.CL 2026-06 conditional novelty 7.0

    VLMs default to visual grounding but a sparse circuit of 2.5-4.8% attention heads in later layers mediates prior-knowledge overrides, identified causally via patching and ablation across three model families.

  2. Analysis-by-Proxy: Localization Signals in VLMs Operating as Condition Encoders

    cs.CV 2026-07 conditional novelty 6.0

    A lightweight Q-Former proxy trained on VLM hidden states reveals that localization signals peak in input-dependent intermediate layers, not the final layers used by standard editing pipelines.

  3. Counting to Four is still a Chore for VLMs

    cs.CV 2026-04 unverdicted novelty 6.0

    VLMs fail at counting because visual evidence degrades in later language layers, and a lightweight Modality Attention Share intervention can encourage better use of image information during answer generation.