Enregistré dans:
Détails bibliographiques
Auteurs principaux: Hoche, Joseph, Bursuc, Andrei, Brellmann, David, Louppe, Gilles, Izmailov, Pavel, Yao, Angela, Franchi, Gianni
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2512.14177
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866908919913250816
author Hoche, Joseph
Bursuc, Andrei
Brellmann, David
Louppe, Gilles
Izmailov, Pavel
Yao, Angela
Franchi, Gianni
author_facet Hoche, Joseph
Bursuc, Andrei
Brellmann, David
Louppe, Gilles
Izmailov, Pavel
Yao, Angela
Franchi, Gianni
contents Large Vision-Language Models (LVLMs) often produce plausible but unreliable outputs, making robust uncertainty estimation essential. Recent work on semantic uncertainty estimates relies on external models to cluster multiple sampled responses and measure their semantic consistency. However, these clustering methods are often fragile, highly sensitive to minor phrasing variations, and can incorrectly group or separate semantically similar answers, leading to unreliable uncertainty estimates. We propose Semantic Gaussian Process Uncertainty (SGPU), a Bayesian framework that quantifies semantic uncertainty by analyzing the geometric structure of answer embeddings, avoiding brittle clustering. SGPU maps generated answers into a dense semantic space, computes the Gram matrix of their embeddings, and summarizes their semantic configuration via the eigenspectrum. This spectral representation is then fed into a Gaussian Process Classifier that learns to map patterns of semantic consistency to predictive uncertainty, and that can be applied in both black-box and white-box settings. Across six LLMs and LVLMs on eight datasets spanning VQA, image classification, and textual QA, SGPU consistently achieves state-of-the-art calibration (ECE) and discriminative (AUROC, AUARC) performance. We further show that SGPU transfers across models and modalities, indicating that its spectral representation captures general patterns of semantic uncertainty.
format Preprint
id arxiv_https___arxiv_org_abs_2512_14177
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Improving Semantic Uncertainty Quantification in LVLMs with Semantic Gaussian Processes
Hoche, Joseph
Bursuc, Andrei
Brellmann, David
Louppe, Gilles
Izmailov, Pavel
Yao, Angela
Franchi, Gianni
Computer Vision and Pattern Recognition
Large Vision-Language Models (LVLMs) often produce plausible but unreliable outputs, making robust uncertainty estimation essential. Recent work on semantic uncertainty estimates relies on external models to cluster multiple sampled responses and measure their semantic consistency. However, these clustering methods are often fragile, highly sensitive to minor phrasing variations, and can incorrectly group or separate semantically similar answers, leading to unreliable uncertainty estimates. We propose Semantic Gaussian Process Uncertainty (SGPU), a Bayesian framework that quantifies semantic uncertainty by analyzing the geometric structure of answer embeddings, avoiding brittle clustering. SGPU maps generated answers into a dense semantic space, computes the Gram matrix of their embeddings, and summarizes their semantic configuration via the eigenspectrum. This spectral representation is then fed into a Gaussian Process Classifier that learns to map patterns of semantic consistency to predictive uncertainty, and that can be applied in both black-box and white-box settings. Across six LLMs and LVLMs on eight datasets spanning VQA, image classification, and textual QA, SGPU consistently achieves state-of-the-art calibration (ECE) and discriminative (AUROC, AUARC) performance. We further show that SGPU transfers across models and modalities, indicating that its spectral representation captures general patterns of semantic uncertainty.
title Improving Semantic Uncertainty Quantification in LVLMs with Semantic Gaussian Processes
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.14177