Information-seeking failures of large language models in agentic clinical reasoning

2026-07-11

AI Summary

This study addresses a critical limitation of large language models (LLMs) in clinical reasoning: their frequent failure to proactively and effectively seek essential information under uncertainty, leading to erroneous decisions. The authors introduce a multi-turn interactive agent evaluation framework focused on hematologic oncology, requiring models to request clinical data over three rounds before rendering diagnostic and therapeutic judgments. Their analysis reveals that the primary bottleneck for current LLMs is not insufficient medical knowledge but a systematic failure in information seeking, with cognitive bias patterns resembling those of novice human clinicians. Evaluation across 32 state-of-the-art models shows a peak accuracy of only 68%; information utilization strongly correlates with diagnostic accuracy (R = 0.69), yet plummets to 26% in the final round, and locally coherent reasoning trajectories often decouple from overall correctness.

Abstract

Large language models achieve high scores on medical knowledge assessments, yet clinical reasoning requires actively deciding what to investigate under uncertainty. We developed an agentic evaluation framework in hematologic oncology in which models must proactively request clinical data across three sequential rounds before committing to a diagnosis and treatment plan. Across 32 frontier models, the best achieved only 68% overall accuracy. Information utilization, the fraction of available data actually requested, was the strongest predictor of diagnostic accuracy (R = 0.69, P < 0.001), yet utilization collapsed from 57% to 26% in the final round, leaving molecular and cytogenetic data critical for treatment selection unexamined. Reasoning traces scored high on a clinical reasoning rubric (91% above threshold) but decorrelated from accuracy, revealing a gap between locally coherent rationales and globally correct conclusions. Error analysis identified search satisficing, anchoring and premature closure as the dominant failure modes, the same cognitive biases that characterize novice clinicians under dual-process models of diagnostic reasoning. These findings demonstrate that the primary limitation of current models in clinical oncology is not insufficient medical knowledge but a systematic failure of information-seeking under uncertainty.

Problem

• information-seeking
• clinical reasoning
• large language models
• diagnostic accuracy
• cognitive biases

Innovation

• agentic evaluation
• information-seeking
• clinical reasoning
• cognitive biases
• large language models