Verification Mirage: Mapping the Reliability Boundary of Self-Verification in Medical VQA

TL;DR

This paper critically examines the reliability of self-verification in medical VQA models, revealing fundamental flaws and task-dependent limitations that compromise safety and accuracy.

Contribution

It introduces a diagnostic framework to map the reliability boundary of self-verification, highlighting the phenomenon of verification mirage and its implications.

Findings

Verification mirage causes high error and bias, especially in knowledge-intensive tasks.

Verifiers tend to under-attend to image evidence, leading to lazy verification.

Cross-verification does not fully mitigate the mirage, and multi-turn loops can lock in wrong answers.

Abstract

Self-verification, re-invoking the same vision language model (VLM) in a fresh context to check its own generated answer, is increasingly used as a default safety layer for medical visual question answering (VQA). We argue that this practice is fundamentally unreliable. We introduce [METHOD NAME], a diagnostic framework for mapping the reliability boundary of medical VLM self-verification by decomposing verifier behavior into discrimination capability and agreement bias. Because the verifier and answer generator are capacity-coupled, the verifier can overly agree with the generator, creating a verification mirage: a regime with both high verifier error and high agreement bias, driven by false acceptance of incorrect answers. Evaluating six open-weight VLMs across five medical VQA datasets and seven medical tasks, we find that this boundary is strongly task-conditioned. Knowledge-intensive clinical tasks fall deepest into the mirage, simpler tasks are more resistant, and perceptual tasks lie in between. Verification also fails to provide an independent safety signal: logistic mixed-effects analysis shows that verifier error and agreement bias become more likely when the generator is wrong, while saliency analyses show that verifiers under-attend to image evidence relative to generators, a phenomenon we call the lazy verifier. Cross-verification reduces but does not eliminate the mirage. Moreover, when verification is reused in multi-turn actor-verifier loops, most initially wrong answers become locked in by false verification. Since our experiments use clean benchmarks, the observed reliability boundary likely underestimates failures in real clinical deployment.