Toward Better Assessment of LLMs' Performance in Clinical Error Detection

• URL: http://arxiv.org/abs/2608.16643v1
• Date: Mon, 17 Aug 2026 14:41:53 GMT
• Title: Toward Better Assessment of LLMs' Performance in Clinical Error Detection
• Abstract summary: We evaluate 15 large language models (LLMs) for automated detection of errors in clinical documentation.<n>We find that 13 of 15 models fall below the level of random pairwise discrimination, even while achieving F1 scores that standard practice would read as moderate.<n>We show that F1 and pairwise accuracy are driven in opposite directions by the same underlying bias, so that ranking models by F1 may systematically promote the weakest discriminators.
• Score: 13.451007655769827
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Automated detection of errors in clinical documentation is a promising application of large language models (LLMs), yet decisions to deploy such models rest on benchmarks that evaluate each clinical note in isolation. Error-detection benchmarks are typically constructed by injecting errors into notes, such that each erroneous note has a natural counterpart. Aggregate discriminative metrics (e.g., balanced accuracy or F1) do not exploit this structure. We show that this omission is consequential. In particular, evaluating 15 diverse LLMs on 4 standardized clinical error-detection test sets across 3 languages, we find that 13 of 15 models fall below the level of random pairwise discrimination, even while achieving F1 scores that standard practice would read as moderate. We also observe that the underlying bias patterns differ across languages: the same model can default to "no error" on one language and over-flag errors on another. To diagnose where discrimination breaks down, we further introduce a procedure to score the evidence models cite in their outputs. We find that while models consistently locate error-relevant content, they fail to produce the corresponding correct verdict on the clean counterpart. Finally, we show that F1 and pairwise accuracy are driven in opposite directions by the same underlying bias, so that ranking models by F1 may systematically promote the weakest discriminators. For safety-critical clinical NLP applications, we advocate for supplementing aggregate metrics with paired evaluations in benchmark reporting. Code and analysis scripts are available at https://github.com/healthylaife/paired-clinical-eval.

Related papers

• NovGauge: A Fine-Grained Benchmark for Diagnosing LLMs' Capability in Paper Novelty Assessment [54.4265627247104]
  Large language models (LLMs) are increasingly used in peer review at major AI conferences.<n>Existing benchmarks assess novelty as a single holistic score.<n>We present NovGauge, a human-anchored benchmark for fine-grained novelty assessment diagnosis.
  arXiv  Detail & Related papers  (2026-09-10T08:34:56Z)
• Fault Detection and Explainable Classification in Automotive HIL Validation via Denoising Autoencoders and In-Context Large Language Models [1.9435397960631864]
  Machine and deep learning have advanced fault diagnosis, yet most supervised models require large labelled datasets.<n>We propose a generalisable and explainable two-phase framework for fault detection and classification during real-time validation.
  arXiv  Detail & Related papers  (2026-07-04T06:43:43Z)
• Clinical Term Extraction using Open-Source Small Language Models [2.5316486726943843]
  Open-source small language models were evaluated for ALS-relevant clinical terms in patient documentation.<n>We compared 26 open-source models using aggregate, label-level, and manual-validation multilabel classification metrics.<n>These findings support targeted hybrid extraction rather than replacement of existing rule-based methods.
  arXiv  Detail & Related papers  (2026-06-19T18:59:33Z)
• 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)
• First, Do No Harm (With LLMs): Mitigating Racial Bias via Agentic Workflows [0.0]
  Large language models (LLMs) are increasingly used in clinical settings, raising concerns about racial bias in generated medical text and clinical reasoning.<n>This study uses the EU AI Act as a governance lens to evaluate five widely used LLMs across two tasks.<n>All models deviated from observed racial distributions in the synthetic case generation task, with GPT-4.1 showing the smallest overall deviation.<n>When embedded in an agentic workflow, DeepSeek V3 showed an improvement of 0.0348 in mean p-value, 0.1166 in median p-value, and 0.0949 in mean difference relative to the standalone model.
  arXiv  Detail & Related papers  (2026-04-20T10:02:38Z)
• Fine-Tune, Don't Prompt, Your Language Model to Identify Biased Language in Clinical Notes [31.925112343488976]
  Clinical documentation can contain emotionally charged language with stigmatizing or privileging valences.<n>We present a framework for detecting and classifying such language.<n>Fine-tuning with lexically primed inputs consistently outperformed prompting approaches.
  arXiv  Detail & Related papers  (2026-02-16T22:39:28Z)
• Do Large Language Model Benchmarks Test Reliability? [66.1783478365998]
  We investigate how well current benchmarks quantify model reliability.<n>Motivated by this gap in the evaluation of reliability, we propose the concept of so-called platinum benchmarks.<n>We evaluate a wide range of models on these platinum benchmarks and find that, indeed, frontier LLMs still exhibit failures on simple tasks.
  arXiv  Detail & Related papers  (2025-02-05T18:58:19Z)
• Classification Error Bound for Low Bayes Error Conditions in Machine Learning [50.25063912757367]
  We study the relationship between the error mismatch and the Kullback-Leibler divergence in machine learning.<n>Motivated by recent observations of low model-based classification errors in many machine learning tasks, we propose a linear approximation of the classification error bound for low Bayes error conditions.
  arXiv  Detail & Related papers  (2025-01-27T11:57:21Z)
• Understanding and Mitigating Classification Errors Through Interpretable Token Patterns [58.91023283103762]
  Characterizing errors in easily interpretable terms gives insight into whether a classifier is prone to making systematic errors. We propose to discover those patterns of tokens that distinguish correct and erroneous predictions. We show that our method, Premise, performs well in practice.
  arXiv  Detail & Related papers  (2023-11-18T00:24:26Z)
• LLMs as Factual Reasoners: Insights from Existing Benchmarks and Beyond [135.8013388183257]
  We propose a new protocol for inconsistency detection benchmark creation and implement it in a 10-domain benchmark called SummEdits. Most LLMs struggle on SummEdits, with performance close to random chance. The best-performing model, GPT-4, is still 8% below estimated human performance.
  arXiv  Detail & Related papers  (2023-05-23T21:50:06Z)
• Negation detection in Dutch clinical texts: an evaluation of rule-based and machine learning methods [0.21079694661943607]
  We compare three methods for negation detection in Dutch clinical notes. We found that both the biLSTM and RoBERTa models consistently outperform the rule-based model in terms of F1 score, precision and recall.
  arXiv  Detail & Related papers  (2022-09-01T14:00:13Z)
• LOGAN: Local Group Bias Detection by Clustering [86.38331353310114]
  We argue that evaluating bias at the corpus level is not enough for understanding how biases are embedded in a model. We propose LOGAN, a new bias detection technique based on clustering. Experiments on toxicity classification and object classification tasks show that LOGAN identifies bias in a local region.
  arXiv  Detail & Related papers  (2020-10-06T16:42:51Z)

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.