VLM-UQBench: A Benchmark for Modality-Specific and Cross-Modality Uncertainties in Vision Language Models

TL;DR

This paper introduces VLM-UQBench, a comprehensive benchmark for evaluating modality-specific and cross-modal uncertainty in vision-language models, revealing current UQ methods' limitations in detecting subtle, instance-level ambiguities.

Contribution

The paper presents VLM-UQBench, a new benchmark with perturbation-based evaluation metrics for modality-aware uncertainty in VLMs, highlighting gaps in current UQ methods.

Findings

Existing UQ methods show modality-specific strengths and weaknesses.

UQ scores weakly correlate with hallucinations and often fail to detect subtle ambiguities.

UQ methods perform comparably to reasoning-based baselines on overt ambiguities but struggle with fine-grained uncertainty.

Abstract

Uncertainty quantification (UQ) is vital for ensuring that vision-language models (VLMs) behave safely and reliably. A central challenge is to localize uncertainty to its source, determining whether it arises from the image, the text, or misalignment between the two. We introduce VLM-UQBench, a benchmark for modality-specific and cross-modal data uncertainty in VLMs, It consists of 600 real-world samples drawn from the VizWiz dataset, curated into clean, image-, text-, and cross-modal uncertainty subsets, and a scalable perturbation pipeline with 8 visual, 5 textual, and 3 cross-modal perturbations. We further propose two simple metrics that quantify the sensitivity of UQ scores to these perturbations and their correlation with hallucinations, and use them to evaluate a range of UQ methods across four VLMs and three datasets. Empirically, we find that: (i) existing UQ methods exhibit strong modality-specific specialization and substantial dependence on the underlying VLM, (ii) modality-specific uncertainty frequently co-occurs with hallucinations while current UQ scores provide only weak and inconsistent risk signals, and (iii) although UQ methods can rival reasoning-based chain-of-thought baselines on overt, group-level ambiguity, they largely fail to detect the subtle, instance-level ambiguity introduced by our perturbation pipeline. These results highlight a significant gap between current UQ practices and the fine-grained, modality-aware uncertainty required for reliable VLM deployment.