Seeing is Believing, but How Much? A Comprehensive Analysis of Verbalized Calibration in Vision-Language Models

TL;DR

This paper evaluates how well vision-language models express their confidence through natural language, revealing prevalent miscalibration and proposing a prompting method to improve confidence alignment across tasks and models.

Contribution

It provides a comprehensive analysis of verbalized confidence in VLMs, identifies calibration issues, and introduces Visual Confidence-Aware Prompting to enhance uncertainty estimation.

Findings

Visual reasoning models show better calibration.

Current VLMs often exhibit notable miscalibration.

Modality-specific reasoning improves confidence reliability.

Abstract

Uncertainty quantification is essential for assessing the reliability and trustworthiness of modern AI systems. Among existing approaches, verbalized uncertainty, where models express their confidence through natural language, has emerged as a lightweight and interpretable solution in large language models (LLMs). However, its effectiveness in vision-language models (VLMs) remains insufficiently studied. In this work, we conduct a comprehensive evaluation of verbalized confidence in VLMs, spanning three model categories, four task domains, and three evaluation scenarios. Our results show that current VLMs often display notable miscalibration across diverse tasks and settings. Notably, visual reasoning models (i.e., thinking with images) consistently exhibit better calibration, suggesting that modality-specific reasoning is critical for reliable uncertainty estimation. To further address calibration challenges, we introduce Visual Confidence-Aware Prompting, a two-stage prompting strategy that improves confidence alignment in multimodal settings. Overall, our study highlights the inherent miscalibration in VLMs across modalities. More broadly, our findings underscore the fundamental importance of modality alignment and model faithfulness in advancing reliable multimodal systems.