Metonymy in vision models undermines attention-based interpretability

• URL: http://arxiv.org/abs/2605.06095v1
• Date: Thu, 07 May 2026 12:14:46 GMT
• Title: Metonymy in vision models undermines attention-based interpretability
• Abstract summary: Part-based reasoning is a strategy to make a computer vision model directly focus on the object parts that are relevant to the downstream task.<n>We show that modern pretrained vision transformers violate the locality assumption and exhibit a strong intra-object leakage.<n>We then show that this inherently disentangled feature extraction improves attribute-driven part discovery on a variety of tasks.
• Score: 31.26551387218204
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Part-based reasoning is a classical strategy to make a computer vision model directly focus on the object parts that are relevant to the downstream task. In the context of deep learning, this also serves to improve by-design interpretability, often by using part-centric attention mechanisms on top of a latent image representation provided by a standard, black-box model. This approach is based on a locality assumption: that the latent representation of an object part encodes primarily information about the corresponding image region. In this work, we test this basic assumption, measuring intra-object leakage in vision models using part-based attribute annotations. Through a comprehensive experimental evaluation, we show that modern pretrained vision transformers violate the locality assumption and exhibit a strong intra-object leakage, in which each part encodes information from the whole object, a visual metonymy that compromises the faithfulness of attention-based interpretable-by-design methods for part-based reasoning, ultimately rendering them uninterpretable. In addition, we establish an upper bound using a two-stage approach that prevents leakage by design. We then show that this inherently disentangled feature extraction improves attribute-driven part discovery on a variety of tasks, confirming the practical impact of intra-object leakage. Our results uncover a neglected issue affecting the interpretability of part-based representations, such as those in CBMs relying on part-centric concepts, highlighting that two-stage approaches offer a promising way to mitigate it.

Related papers

• Multi-Part Object Representations via Graph Structures and Co-Part Discovery [24.418060973308908]
  We propose a novel method that leverages on explicit graph representations for parts and present a co-part object discovery algorithm.<n> Experimental results on simulated, realistic, and real-world images show marked improvements in the quality of discovered objects compared to state-of-the-art methods.<n>We also show that the discovered object-centric representations can more accurately predict key object properties in a downstream task.
  arXiv  Detail & Related papers  (2025-12-20T03:38:41Z)
• Lost in Embeddings: Information Loss in Vision-Language Models [38.8742043080561]
  Vision--language models often process visual inputs through a pretrained vision encoder, followed by a projection into the language model's embedding space via a connector component.<n>While crucial for modality fusion, the potential information loss induced by this projection step and its direct impact on model capabilities remain understudied.<n>We introduce two complementary approaches to examine and quantify this loss by analyzing the latent representation space.
  arXiv  Detail & Related papers  (2025-09-15T14:38:06Z)
• Oh-A-DINO: Understanding and Enhancing Attribute-Level Information in Self-Supervised Object-Centric Representations [9.949149600332836]
  Self-supervised vision models and slot-based representations excel at identifying edge-derived geometry but fail to preserve non-geometric surface-level cues.<n>We show that learning an auxiliary latent space over segmented patches, where VAE regularisation enforces compact, disentangled object-centric representations, recovers these missing attributes.
  arXiv  Detail & Related papers  (2025-03-12T21:57:41Z)
• An Investigation on The Position Encoding in Vision-Based Dynamics Prediction [19.700374722227107]
  Vision-based dynamics prediction models, which predict object states by utilizing RGB images and simple object descriptions, were challenged by environment misalignments. This paper provides studies to investigate the process and necessary conditions for encoding position information via using the bounding box as the object abstract into output features.
  arXiv  Detail & Related papers  (2024-08-27T17:02:03Z)
• Zero-Shot Object-Centric Representation Learning [72.43369950684057]
  We study current object-centric methods through the lens of zero-shot generalization. We introduce a benchmark comprising eight different synthetic and real-world datasets. We find that training on diverse real-world images improves transferability to unseen scenarios.
  arXiv  Detail & Related papers  (2024-08-17T10:37:07Z)
• Weakly-supervised Contrastive Learning for Unsupervised Object Discovery [52.696041556640516]
  Unsupervised object discovery is promising due to its ability to discover objects in a generic manner. We design a semantic-guided self-supervised learning model to extract high-level semantic features from images. We introduce Principal Component Analysis (PCA) to localize object regions.
  arXiv  Detail & Related papers  (2023-07-07T04:03:48Z)
• Spotlight Attention: Robust Object-Centric Learning With a Spatial Locality Prior [88.9319150230121]
  Object-centric vision aims to construct an explicit representation of the objects in a scene. We incorporate a spatial-locality prior into state-of-the-art object-centric vision models. We obtain significant improvements in segmenting objects in both synthetic and real-world datasets.
  arXiv  Detail & Related papers  (2023-05-31T04:35:50Z)
• Towards Unsupervised Visual Reasoning: Do Off-The-Shelf Features Know How to Reason? [30.16956370267339]
  We introduce a protocol to evaluate visual representations for the task of Visual Question Answering. In order to decouple visual feature extraction from reasoning, we design a specific attention-based reasoning module. We compare two types of visual representations, densely extracted local features and object-centric ones, against the performances of a perfect image representation using ground truth.
  arXiv  Detail & Related papers  (2022-12-20T14:36:45Z)
• Unsupervised Part Discovery from Contrastive Reconstruction [90.88501867321573]
  The goal of self-supervised visual representation learning is to learn strong, transferable image representations. We propose an unsupervised approach to object part discovery and segmentation. Our method yields semantic parts consistent across fine-grained but visually distinct categories.
  arXiv  Detail & Related papers  (2021-11-11T17:59:42Z)
• Revisiting spatio-temporal layouts for compositional action recognition [63.04778884595353]
  We take an object-centric approach to action recognition. The main focus of this paper is compositional/few-shot action recognition. We demonstrate how to improve the performance of appearance-based models by fusion with layout-based models.
  arXiv  Detail & Related papers  (2021-11-02T23:04:39Z)
• Self-Supervision by Prediction for Object Discovery in Videos [62.87145010885044]
  In this paper, we use the prediction task as self-supervision and build a novel object-centric model for image sequence representation. Our framework can be trained without the help of any manual annotation or pretrained network. Initial experiments confirm that the proposed pipeline is a promising step towards object-centric video prediction.
  arXiv  Detail & Related papers  (2021-03-09T19:14:33Z)

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.