ViLU: Learning Vision-Language Uncertainties for Failure Prediction

TL;DR

ViLU introduces a novel framework for uncertainty quantification in vision-language models, leveraging multi-modal representations and a loss-agnostic predictor to improve failure prediction across diverse datasets.

Contribution

The paper presents ViLU, a new post-hoc uncertainty quantification method that integrates visual and textual features for better failure prediction in vision-language models.

Findings

Significant improvements over state-of-the-art failure prediction methods.

Effective on both classification and large-scale caption datasets.

Ablation studies confirm architecture and training effectiveness.

Abstract

Reliable Uncertainty Quantification (UQ) and failure prediction remain open challenges for Vision-Language Models (VLMs). We introduce ViLU, a new Vision-Language Uncertainty quantification framework that contextualizes uncertainty estimates by leveraging all task-relevant textual representations. ViLU constructs an uncertainty-aware multi-modal representation by integrating the visual embedding, the predicted textual embedding, and an image-conditioned textual representation via cross-attention. Unlike traditional UQ methods based on loss prediction, ViLU trains an uncertainty predictor as a binary classifier to distinguish correct from incorrect predictions using a weighted binary cross-entropy loss, making it loss-agnostic. In particular, our proposed approach is well-suited for post-hoc settings, where only vision and text embeddings are available without direct access to the model itself. Extensive experiments on diverse datasets show the significant gains of our method compared to state-of-the-art failure prediction methods. We apply our method to standard classification datasets, such as ImageNet-1k, as well as large-scale image-caption datasets like CC12M and LAION-400M. Ablation studies highlight the critical role of our architecture and training in achieving effective uncertainty quantification. Our code is publicly available and can be found here: https://github.com/ykrmm/ViLU.