Clinician-Level Agreement Without Clinical Caution: LLM Evaluator Limits in Medical AI Benchmarking

• URL: http://arxiv.org/abs/2607.01103v1
• Date: Wed, 01 Jul 2026 15:55:31 GMT
• Title: Clinician-Level Agreement Without Clinical Caution: LLM Evaluator Limits in Medical AI Benchmarking
• Abstract summary: We introduce MedQADE, the first standardised open-response clinical benchmark for German.<n>The top-performing evaluator model, Gemini 3 Flash, reached alignment consistent with the physician ceiling.<n>Despite this statistical alignment, automated evaluators exhibited near-absent clinical metacognition.
• Score: 27.829988205900055
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Open-response evaluation provides stronger clinical validity than multiple-choice benchmarks but creates a scoring bottleneck that motivates automated LLM-asa-Judge approaches. Whether such evaluators replicate clinical calibration and caution, however, remains untested. We introduce MedQADE, the first standardised open-response clinical benchmark for German, a major clinical language lacking native evaluation infrastructure, comprising 3,800 items annotated by ten practising physicians and nine Large Language Model (LLM) evaluators. The top-performing evaluator model, Gemini 3 Flash, reached alignment consistent with the physician ceiling (\k{appa} = 0.694 vs. \k{appa} = 0.709), though wide confidence intervals limit interpretation. Despite this statistical alignment, automated evaluators exhibited near-absent clinical metacognition: physicians scaled abstention with item difficulty, while frontier models assigned definitive scores in every case. We additionally quantified systematic lineage-dependent biases, where models preferentially scored architectural siblings, an effect independent of language. These results show that statistical alignment does not ensure clinical caution, and that evaluator independence requires explicit verification.

Related papers

• Beyond Scalar Scores: Exploring LLM-based Metrics for Clinical Significance Evaluation in Radiology Reports [49.5225801722164]
  Existing metrics obscure this requirement by reducing report quality to a medically ungrounded scalar.<n>We study this boundary using ReEvalMed benchmark as testbed and evaluate metric-level clinical significance.
  arXiv  Detail & Related papers  (2026-06-17T08:10:30Z)
• Auditing Multimodal LLM Raters: Central Tendency Bias in Clinical Ordinal Scoring [2.835587889687748]
  Multimodal large language models (LLMs) are increasingly explored as automated evaluators in clinical settings.<n>We benchmark three frontier LLM families against supervised deep learning models for scoring Clock Drawing Test (CDT) images.<n>Our findings extend the LLM-as-a-judge bias literature from NLP evaluation to clinical assessment.
  arXiv  Detail & Related papers  (2026-05-11T15:37:24Z)
• Detecting Stealth Sycophancy in Mental-Health Dialogue with Dynamic Emotional Signature Graphs [4.002154823241671]
  We study multi-domain support-dialogue evaluation without relying on large language models as final judges.<n>We find that direct LLM judges and symmetric text-similarity metrics are poorly aligned with therapeutic quality.<n>To address this issue, we propose Dynamic Emotional Signature Graphs (DESG)<n>DESG represents dialogue windows with decoupled clinical states and scores them using asymmetric clinical geometry.
  arXiv  Detail & Related papers  (2026-05-05T07:56:20Z)
• Case-Specific Rubrics for Clinical AI Evaluation: Methodology, Validation, and LLM-Clinician Agreement Across 823 Encounters [3.018184429993625]
  Methods requiring expert review per scoring instance are too slow and expensive for safe, iterative deployment.<n>Twenty clinicians authored 1,646 rubrics for 823 clinical cases across primary care, psychiatry, oncology, and behavioral health.<n>Case-specific rubrics offer a path for clinical AI evaluation that preserves expert judgment while enabling automation at three orders lower cost.
  arXiv  Detail & Related papers  (2026-04-27T17:17:56Z)
• Same Verdict, Different Reasons: LLM-as-a-Judge and Clinician Disagreement on Medical Chatbot Completeness [49.2667937337333]
  We stress-test this assumption for detecting incomplete patient-facing medical responses.<n>We evaluate three granularities (General-Likert, Analytical-Rubric, Dynamic-Checklist) and three backbone models across two clinician-annotated datasets.
  arXiv  Detail & Related papers  (2026-03-26T19:01:55Z)
• Cerebra: A Multidisciplinary AI Board for Multimodal Dementia Characterization and Risk Assessment [56.62016795093786]
  Cerebra is an interactive multi-agent AI team that coordinates specialized agents for EHR, clinical notes, and medical imaging analysis.<n>It supports privacy-preserving deployment by operating on structured representations and remains robust when modalities are incomplete.<n>Cerebra significantly improved expert performance, increasing accuracy by 17.5 percentage points in prospective dementia risk estimation.
  arXiv  Detail & Related papers  (2026-03-23T05:46:45Z)
• QuarkMedBench: A Real-World Scenario Driven Benchmark for Evaluating Large Language Models [8.050059911826338]
  Large Language Models (LLMs) excel on standardized medical exams, but high scores often fail to translate to high-quality responses for real-world medical queries.<n>We introduce QuarkMedBench, an ecologically valid benchmark tailored for real-world medical LLM assessment.
  arXiv  Detail & Related papers  (2026-03-14T01:51:43Z)
• MedArena: Comparing LLMs for Medicine-in-the-Wild Clinician Preferences [50.71326426975699]
  We present MedArena, an interactive evaluation platform for medical large language models (LLMs)<n>Given a clinician-provided query, MedArena presents responses from two randomly selected models and asks the user to select the preferred response.<n>Out of 1571 preferences collected across 12 LLMs up to November 1, 2025, Gemini 2.0 Flash Thinking, Gemini 2.5 Pro, and GPT-4o were the top three models by Bradley-Terry rating.
  arXiv  Detail & Related papers  (2026-03-13T22:30:26Z)
• Suppressing Prior-Comparison Hallucinations in Radiology Report Generation via Semantically Decoupled Latent Steering [94.37535002230504]
  We develop a training-free, inference-time control framework termed Semantically Decoupled Latent Steering.<n>Our approach constructs a semantic-free intervention vector via large language model (LLM)-driven semantic decomposition.<n>We show that our approach significantly reduces the probability of historical hallucinations.
  arXiv  Detail & Related papers  (2026-02-27T04:49:01Z)
• WER is Unaware: Assessing How ASR Errors Distort Clinical Understanding in Patient Facing Dialogue [3.468314243424983]
  Automatic Speech Recognition (ASR) is increasingly deployed in clinical dialogue.<n>Standard evaluations still rely heavily on Error Error Rate (WER)<n>This paper challenges that standard, investigating whether WER or other common metrics correlate with clinical impact of transcription errors.
  arXiv  Detail & Related papers  (2025-11-20T16:59:20Z)
• Quantifying the Reasoning Abilities of LLMs on Real-world Clinical Cases [48.87360916431396]
  We introduce MedR-Bench, a benchmarking dataset of 1,453 structured patient cases, annotated with reasoning references.<n>We propose a framework encompassing three critical examination recommendation, diagnostic decision-making, and treatment planning, simulating the entire patient care journey.<n>Using this benchmark, we evaluate five state-of-the-art reasoning LLMs, including DeepSeek-R1, OpenAI-o3-mini, and Gemini-2.0-Flash Thinking, etc.
  arXiv  Detail & Related papers  (2025-03-06T18:35:39Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.