Mechanisms of Object Localization in Vision-Language Models

• URL: http://arxiv.org/abs/2605.19792v1
• Date: Tue, 19 May 2026 12:56:32 GMT
• Title: Mechanisms of Object Localization in Vision-Language Models
• Abstract summary: We study the processes that support object localization in visually-grounded language models (VLMs)<n>We find that localization is driven by a containerization mechanism in which object-aligned tokens define the spatial extent of the object.<n>Only a very small set of attention heads mediates the causal effect for both classification and localization.
• Score: 10.367669666212473
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Visually-grounded language models (VLMs) are highly effective in linking visual and textual information, yet they often struggle with basic classification and localization tasks. While classification mechanisms have been studied more extensively, the processes that support object localization remain poorly understood. In this work, we investigate two representative families, LLaVA-1.5 and InternVL-3.5, using a suite of mechanistic interpretability tools, including token ablations, attention knockout, and causal mediation analysis. We find that localization is driven by a containerization mechanism in which object-aligned tokens define the spatial extent of the object, while the semantic arrangement of tokens within those boundaries is largely irrelevant to the predicted box. Only a very small set of attention heads mediates the causal effect for both classification and localization, concentrating in early-mid layers for LLaVA and mid-late layers for InternVL. The two tasks share some early processing but ultimately depend on largely distinct specialized heads. Overall, we provide the first layer- and head-level account of localization in VLMs, revealing narrow computational pathways that can guide future model design and grounding objectives.

Related papers

• Grounding Isn't Knowing: Do VLMs Need Object Localization for Spatial Reasoning? [24.02857259963021]
  Vision-language models (VLMs) can answer spatial questions, yet the mechanisms connecting object grounding to spatial reasoning remain poorly understood.<n>In this work, we investigate two representative model families, LLaVA-1.5 and Qwen2.5-VL, using a suite of mechanistic interpretability tools.<n>We find that spatial relation prediction follows a staged grounding-to-reasoning process in which object-aligned tokens establish coarse target-reference anchors, while precise bounding-box boundaries are not required.
  arXiv  Detail & Related papers  (2026-08-24T10:26:15Z)
• AwareVLN: Reasoning with Self-awareness for Vision-Language Navigation [78.80536515102305]
  Vision-and-Language Navigation (VLN) requires an agent to ground language instructions to its own movement within a visual environment.<n>We propose AwareVLN, a novel framework that equips the navigation model with a self-aware reasoning mechanism.<n>Our approach features two key innovations: (1) a structural reasoning module that fosters spatial and task-oriented self-awareness, and (2) an automatic data engine with progress division for effective training.
  arXiv  Detail & Related papers  (2026-05-21T17:58:26Z)
• From Human Cognition to Neural Activations: Probing the Computational Primitives of Spatial Reasoning in LLMs [2.3268065525165316]
  spatial intelligence becomes an increasingly important capability for foundation models.<n>It remains unclear whether foundation models' performance on spatial reasoning benchmarks reflects structured internal spatial representations or reliance on language models.
  arXiv  Detail & Related papers  (2026-03-27T11:42:36Z)
• From Drop-off to Recovery: A Mechanistic Analysis of Segmentation in MLLMs [50.185593677108436]
  Multimodal Large Language Models (MLLMs) are increasingly applied to pixel-level vision tasks, yet their intrinsic capacity for spatial understanding remains poorly understood.<n>We investigate segmentation capacity through a layerwise linear probing evaluation across the entire MLLM pipeline: vision encoder, adapter, and LLM.
  arXiv  Detail & Related papers  (2026-03-18T00:22:15Z)
• TagaVLM: Topology-Aware Global Action Reasoning for Vision-Language Navigation [70.23578202012048]
  Vision-Language Navigation (VLN) presents a unique challenge for Large Vision-Language Models (VLMs) due to their inherent architectural mismatch.<n>We propose TagaVLM (Topology-Aware Global Action reasoning), an end-to-end framework that explicitly injects topological structures into the VLM backbone.<n>To enhance topological node information, an Interleaved Navigation Prompt strengthens node-level visual-text alignment.<n>With the embedded topological graph, the model is capable of global action reasoning, allowing for robust path correction.
  arXiv  Detail & Related papers  (2026-03-03T13:28:07Z)
• Spatial Reasoning in Foundation Models: Benchmarking Object-Centric Spatial Understanding [8.202861909913791]
  We present a benchmark for object-centric spatial reasoning in foundation models.<n>We find a stable trade-off: detectors such as GroundingDINO and OWLv2 deliver precise boxes with limited relational reasoning.<n>Our study highlights the gap between localization and true spatial understanding, and pointing toward the need for spatially-aware foundation models in the community.
  arXiv  Detail & Related papers  (2025-09-26T06:06:19Z)
• Unifying Attention Heads and Task Vectors via Hidden State Geometry in In-Context Learning [8.42768790741139]
  In this paper, we analyze two geometric factors that govern performance: the separability and alignment of query hidden states.<n>Previous Token Heads drive separability, while Induction Heads and task vectors enhance alignment.<n>Our findings thus bridge the gap between attention heads and task vectors, offering a unified account of ICL's underlying mechanisms.
  arXiv  Detail & Related papers  (2025-05-24T15:42:20Z)
• How do Large Language Models Understand Relevance? A Mechanistic Interpretability Perspective [64.00022624183781]
  Large language models (LLMs) can assess relevance and support information retrieval (IR) tasks.<n>We investigate how different LLM modules contribute to relevance judgment through the lens of mechanistic interpretability.
  arXiv  Detail & Related papers  (2025-04-10T16:14:55Z)
• Visual-Linguistic Agent: Towards Collaborative Contextual Object Reasoning [26.35257570870916]
  Visual-Linguistic Agent (VLA) is a collaborative framework that combines the relational reasoning strengths of MLLMs with the precise localization capabilities of traditional object detectors. VLA significantly enhances both spatial reasoning and object localization, addressing key challenges in multimodal understanding.
  arXiv  Detail & Related papers  (2024-11-15T15:02:06Z)
• Generalized Robot 3D Vision-Language Model with Fast Rendering and Pre-Training Vision-Language Alignment [55.11291053011696]
  This work presents a framework for dealing with 3D scene understanding when the labeled scenes are quite limited.<n>To extract knowledge for novel categories from the pre-trained vision-language models, we propose a hierarchical feature-aligned pre-training and knowledge distillation strategy.<n>In the limited reconstruction case, our proposed approach, termed WS3D++, ranks 1st on the large-scale ScanNet benchmark.
  arXiv  Detail & Related papers  (2023-12-01T15:47:04Z)
• Unveiling the Potential of Structure-Preserving for Weakly Supervised Object Localization [71.79436685992128]
  We propose a two-stage approach, termed structure-preserving activation (SPA), towards fully leveraging the structure information incorporated in convolutional features for WSOL. In the first stage, a restricted activation module (RAM) is designed to alleviate the structure-missing issue caused by the classification network. In the second stage, we propose a post-process approach, termed self-correlation map generating (SCG) module to obtain structure-preserving localization maps.
  arXiv  Detail & Related papers  (2021-03-08T03:04:14Z)
• PGL: Prior-Guided Local Self-supervised Learning for 3D Medical Image Segmentation [87.50205728818601]
  We propose a PriorGuided Local (PGL) self-supervised model that learns the region-wise local consistency in the latent feature space. Our PGL model learns the distinctive representations of local regions, and hence is able to retain structural information.
  arXiv  Detail & Related papers  (2020-11-25T11:03:11Z)

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.