Reading Between the Lines: Abstaining from VLM-Generated OCR Errors via Latent Representation Probes

TL;DR

This paper introduces Latent Representation Probing (LRP), a method to improve VLMs' ability to abstain from uncertain predictions by analyzing internal representations, enhancing reliability in safety-critical scene text tasks.

Contribution

The paper proposes LRP, a novel approach that leverages internal hidden states and attention patterns in VLMs to better detect uncertainty and abstain from unreliable answers.

Findings

LRP improves abstention accuracy by 7.6% over baselines.

Probes generalize across datasets and uncertainty sources.

Optimal signals are found in intermediate layers.

Abstract

As VLMs are deployed in safety-critical applications, their ability to abstain from answering when uncertain becomes crucial for reliability, especially in Scene Text Visual Question Answering (STVQA) tasks. For example, OCR errors like misreading "50 mph" as "60 mph" could cause severe traffic accidents. This leads us to ask: Can VLMs know when they can't see? Existing abstention methods suggest pessimistic answers: they either rely on miscalibrated output probabilities or require semantic agreement unsuitable for OCR tasks. However, this failure may indicate we are looking in the wrong place: uncertainty signals could be hidden in VLMs' internal representations. Building on this insight, we propose Latent Representation Probing (LRP): training lightweight probes on hidden states or attention patterns. We explore three probe designs: concatenating representations across all layers, aggregating attention over visual tokens, and ensembling single layer probes by majority vote. Experiments on four benchmarks across image and video modalities show LRP improves abstention accuracy by 7.6\% over best baselines. Our analysis reveals: probes generalize across various uncertainty sources and datasets, and optimal signals emerge from intermediate rather than final layers. This establishes a principled framework for building deployment-ready AI systems by detecting confidence signals from internal states rather than unreliable outputs.