Research

Paper

AI LLM March 03, 2026

Guideline-Grounded Evidence Accumulation for High-Stakes Agent Verification

Authors

Yichi Zhang, Nabeel Seedat, Yinpeng Dong, Peng Cui, Jun Zhu, Mihaela van de Schaar

Abstract

As LLM-powered agents have been used for high-stakes decision-making, such as clinical diagnosis, it becomes critical to develop reliable verification of their decisions to facilitate trustworthy deployment. Yet, existing verifiers usually underperform owing to a lack of domain knowledge and limited calibration. To address this, we establish GLEAN, an agent verification framework with Guideline-grounded Evidence Accumulation that compiles expert-curated protocols into trajectory-informed, well-calibrated correctness signals. GLEAN evaluates the step-wise alignment with domain guidelines and aggregates multi-guideline ratings into surrogate features, which are accumulated along the trajectory and calibrated into correctness probabilities using Bayesian logistic regression. Moreover, the estimated uncertainty triggers active verification, which selectively collects additional evidence for uncertain cases via expanding guideline coverage and performing differential checks. We empirically validate GLEAN with agentic clinical diagnosis across three diseases from the MIMIC-IV dataset, surpassing the best baseline by 12% in AUROC and 50% in Brier score reduction, which confirms the effectiveness in both discrimination and calibration. In addition, the expert study with clinicians recognizes GLEAN's utility in practice.

Metadata

arXiv ID: 2603.02798
Provider: ARXIV
Primary Category: cs.AI
Published: 2026-03-03
Fetched: 2026-03-04 03:41

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.02798v1</id>\n    <title>Guideline-Grounded Evidence Accumulation for High-Stakes Agent Verification</title>\n    <updated>2026-03-03T09:36:43Z</updated>\n    <link href='https://arxiv.org/abs/2603.02798v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.02798v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>As LLM-powered agents have been used for high-stakes decision-making, such as clinical diagnosis, it becomes critical to develop reliable verification of their decisions to facilitate trustworthy deployment. Yet, existing verifiers usually underperform owing to a lack of domain knowledge and limited calibration. To address this, we establish GLEAN, an agent verification framework with Guideline-grounded Evidence Accumulation that compiles expert-curated protocols into trajectory-informed, well-calibrated correctness signals. GLEAN evaluates the step-wise alignment with domain guidelines and aggregates multi-guideline ratings into surrogate features, which are accumulated along the trajectory and calibrated into correctness probabilities using Bayesian logistic regression. Moreover, the estimated uncertainty triggers active verification, which selectively collects additional evidence for uncertain cases via expanding guideline coverage and performing differential checks. We empirically validate GLEAN with agentic clinical diagnosis across three diseases from the MIMIC-IV dataset, surpassing the best baseline by 12% in AUROC and 50% in Brier score reduction, which confirms the effectiveness in both discrimination and calibration. In addition, the expert study with clinicians recognizes GLEAN's utility in practice.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n    <published>2026-03-03T09:36:43Z</published>\n    <arxiv:primary_category term='cs.AI'/>\n    <author>\n      <name>Yichi Zhang</name>\n    </author>\n    <author>\n      <name>Nabeel Seedat</name>\n    </author>\n    <author>\n      <name>Yinpeng Dong</name>\n    </author>\n    <author>\n      <name>Peng Cui</name>\n    </author>\n    <author>\n      <name>Jun Zhu</name>\n    </author>\n    <author>\n      <name>Mihaela van de Schaar</name>\n    </author>\n  </entry>"
}