aiqbench analysis
This research critically evaluates confidence estimation methods for vision-language models in medical imaging, revealing that no single estimator reliably predicts when a model should abstain across different domains. The study concludes that safe deployment relies on calibrated triage under clinical oversight rather than autonomous deferral, as base-model competence ultimately limits what can be safely delegated.
Calibrated Triage, Not Autonomy: Confidence Estimation for Medical Vision-Language Models
arXiv:2606.15910v2 Announce Type: replace Abstract: A vision-language model can answer a question about a chest radiograph or a pathology slide fluently and confidently while barely using the image, relying instead on language priors. In medicine this is the failure that matters most: the answer looks trustworthy and is not, and the natural safeguard is a confidence score reliable enough to say when the model should abstain. We ask a deployment question rather than an accuracy one: how much medical imaging work a vision-language model can safely defer on its own, and which confidence signal makes that possible. We evaluate nine confidence estimators, spanning training-free logit baselines, prompt-based self-reports, and trained internal probes, across five open-weight LVLMs and three medical VQA datasets covering broad clinical imaging, radiology, and pathology, every probe trained only on natural images and applied to medicine without adaptation. Recast as bounded selective prediction, the comparison is cautionary. Standard metrics mislead: discrimination barely separates the estimators, and a fixed high-confidence cutoff separates them far less than it appears, because their scores sit on incomparable scales; no estimator is reliably best across domains or models. What can be safely deferred is set at two levels: base-model competence fixes a ceiling, and the confidence layer determines how much of it is reachable. At a 20% error tolerance the strongest estimator defers about a quarter of radiology cases under a distribution-free guarantee and a third under a held-out threshold, and little to none of pathology. The usable role is calibrated triage under clinical oversight, not autonomous deferral: a good estimator makes a competent model defer safely where it is competent, but none manufactures reliability where the base model lacks it. We release all outputs, correctness judgments, and confidence scores, with code.
Abstract:A vision-language model can answer a question about a chest radiograph or a pathology slide fluently and confidently while barely using the image, relying instead on language priors. In medicine this is the failure that matters most: the answer looks trustworthy and is not, and the natural safeguard is a confidence score reliable enough to say when the model should abstain. We ask a deployment question rather than an accuracy one: how much medical imaging work a vision-language model can safely defer on its own, and which confidence signal makes that possible. We evaluate nine confidence estimators, spanning training-free logit baselines, prompt-based self-reports, and trained internal probes, across five open-weight LVLMs and three medical VQA datasets covering broad clinical imaging, radiology, and pathology, every probe trained only on natural images and applied to medicine without adaptation. Recast as bounded selective prediction, the comparison is cautionary. Standard metrics mislead: discrimination barely separates the estimators, and a fixed high-confidence cutoff separates them far less than it appears, because their scores sit on incomparable scales; no estimator is reliably best across domains or models. What can be safely deferred is set at two levels: base-model competence fixes a ceiling, and the confidence layer determines how much of it is reachable. At a 20% error tolerance the strongest estimator defers about a quarter of radiology cases under a distribution-free guarantee and a third under a held-out threshold, and little to none of pathology. The usable role is calibrated triage under clinical oversight, not autonomous deferral: a good estimator makes a competent model defer safely where it is competent, but none manufactures reliability where the base model lacks it. We release all outputs, correctness judgments, and confidence scores, with code.
Computation and Language (cs.CL)
arXiv:2606.15910 [cs.CL]
(or arXiv:2606.15910v2 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2606.15910 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 378 words
Related coverage
Calibrated Triage, Not Autonomy: Confidence Estimation for Medical Vision-Language Models
arXiv · Jun 17, 2026
Overconfidence and Calibration in Medical VQA: Empirical Findings and Hallucination-Aware Mitigation
arXiv · Oct 1, 2026
AI in healthcare: applications and best practices
Databricks · Jul 24, 2026
Separating Capability from Confidence: Grounded Dual-State Calibration for GRPO-Trained Medical Vision-Language Models
arXiv · Sep 9, 2026