Guardado en:
Detalles Bibliográficos
Autores principales: Lombardo, Gabriele, Maiorana, Luigi, Presti, Liliana Lo, La Cascia, Marco
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:https://arxiv.org/abs/2605.09090
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913107286163456
author Lombardo, Gabriele
Maiorana, Luigi
Presti, Liliana Lo
La Cascia, Marco
author_facet Lombardo, Gabriele
Maiorana, Luigi
Presti, Liliana Lo
La Cascia, Marco
contents Visual Grounding benchmarks assume that the object described by a referring expression is always present in the image, and grounding models are therefore rarely evaluated under semantically mismatched captions. In such cases, models frequently exhibit approximation behavior, producing a plausible bounding box that satisfies only part of the expression (\eg, preserving the original object while ignoring modified contextual cues). Because mismatched captions represent realistic edge cases, this behavior compromises reliability and raises concerns from an explainability perspective. Identifying its underlying causes is thus essential for improving model faithfulness and interpretability. Adopting a mechanistic interpretability viewpoint, this work examines whether embedding anisotropy contributes to counterfactual failures. A similarity-controlled counterfactual caption generation protocol is introduced to systematically perturb object or contextual components within predefined embedding similarity intervals, enabling a fine-grained analysis of grounding behavior as a function of alignment. Experiments on two Transformer-based models with markedly different embedding geometries (BERT-based TransVG and CLIP-based SwimVG) reveal no meaningful correlation between cosine similarity and approximation. These findings suggest that anisotropy alone does not account for counterfactual errors, and that robustness requires investigating finer-grained geometric properties of the embedding space.
format Preprint
id arxiv_https___arxiv_org_abs_2605_09090
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Investigating Anisotropy in Visual Grounding under Controlled Counterfactual Perturbations
Lombardo, Gabriele
Maiorana, Luigi
Presti, Liliana Lo
La Cascia, Marco
Computer Vision and Pattern Recognition
Artificial Intelligence
Visual Grounding benchmarks assume that the object described by a referring expression is always present in the image, and grounding models are therefore rarely evaluated under semantically mismatched captions. In such cases, models frequently exhibit approximation behavior, producing a plausible bounding box that satisfies only part of the expression (\eg, preserving the original object while ignoring modified contextual cues). Because mismatched captions represent realistic edge cases, this behavior compromises reliability and raises concerns from an explainability perspective. Identifying its underlying causes is thus essential for improving model faithfulness and interpretability. Adopting a mechanistic interpretability viewpoint, this work examines whether embedding anisotropy contributes to counterfactual failures. A similarity-controlled counterfactual caption generation protocol is introduced to systematically perturb object or contextual components within predefined embedding similarity intervals, enabling a fine-grained analysis of grounding behavior as a function of alignment. Experiments on two Transformer-based models with markedly different embedding geometries (BERT-based TransVG and CLIP-based SwimVG) reveal no meaningful correlation between cosine similarity and approximation. These findings suggest that anisotropy alone does not account for counterfactual errors, and that robustness requires investigating finer-grained geometric properties of the embedding space.
title Investigating Anisotropy in Visual Grounding under Controlled Counterfactual Perturbations
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2605.09090