Compositional Reasoning Depth Predicts Clinical AI Failure: Empirical Evidence Consistent with Transformer Compositionality Limits in Electronic Health Record Question Answering

• URL: http://arxiv.org/abs/2606.16890v1
• Date: Mon, 15 Jun 2026 16:00:27 GMT
• Title: Compositional Reasoning Depth Predicts Clinical AI Failure: Empirical Evidence Consistent with Transformer Compositionality Limits in Electronic Health Record Question Answering
• Abstract summary: We show monotone decline with hop count in EHR question-answer pairs across four hop levels.<n>Hop count is a theory-motivated, cross-architecture predictor of large-model error on EHR question answering.
• Score: 0.6768558752130311
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Aggregate accuracy benchmarks conceal a systematic structure in how large language models fail at electronic health record (EHR) question answering: questions requiring more inferential steps produce disproportionately more errors. Motivated by theoretical results on transformer compositionality limits, we introduce a pre-specified hop-count taxonomy -- the number of distinct reasoning steps required to answer a clinical question from an EHR -- as a principled predictor of model failure. We annotate 313 clinician-generated MedAlign EHR question-answer pairs across four hop levels and evaluate 301 questions in a within-model ablation (claude-sonnet-4-6, zero-shot vs. extended thinking) and cross-architecture replications (gpt-4o and gpt-5.4-2026-03-05, zero-shot). All three models, spanning two providers and two OpenAI generations (GPT-4 and GPT-5), show monotone accuracy decline with hop count: Claude Sonnet zero-shot falls from 30.6% (hop=1) to 17.6% (hop=4) (Cochran-Armitage z=-2.30, p=0.011; OR per hop 0.72, 95% CI [0.56,0.92], p=0.008); GPT-4o replicates this (37.8% to 14.7%; OR 0.58 [0.45,0.75], p<0.001); and gpt-5.4-2026-03-05 confirms it (37.8% to 23.5%; OR 0.80 [0.66,0.98], p=0.027). A pre-specified context-sufficiency audit shows higher-hop questions are not differentially disadvantaged by EHR truncation (answerability 93-95% at hops 2-4 vs. 79% at hop=1), so the decline reflects compositional reasoning difficulty. Extended thinking did not significantly flatten the accuracy-depth curve across three reasoning conditions, and thinking-token usage scaled with hop count (r=0.31, p<0.0001), consistent with the predicted O(k) computational requirement. Hop count is thus a theory-motivated, cross-architecture predictor of large-language-model error on EHR question answering, with direct implications for deployment risk stratification of clinical AI.

Related papers

• Not All Relations Are Equal: Relation-Balanced and Calibrated Graph Learning for Provenance-Based Intrusion Detection [66.44527094471619]
  We present RECAL, an unsupervised framework using relation-balanced masked graph learning to better capture rare interaction patterns.<n>On three DARPA E3 datasets, RECAL achieves F1 scores of 99.99%, 99.93%, and 99.99%, outperforming the best baseline on each dataset by 0.88, 0.82, and 0.42 percentage points, respectively.
  arXiv  Detail & Related papers  (2026-09-15T00:31:01Z)
• Accurate Ensembles, Fragile Narratives: Multi-Scale Stacking and a Fidelity Audit of LLM-Generated Explanations for Credit Risk [0.0]
  Credit scoring increasingly relies on models whose decision logic cannot be read off their parameters.<n>A common proposal closes that gap with a language model: compute feature attributions, hand them to an LLM, and let it write the rationale.<n>We build such a system end to end and test whether the second half of the promise holds.
  arXiv  Detail & Related papers  (2026-08-08T13:22:14Z)
• Passing Coarse Marginal Checks Can Be Cheap: Persona Mixtures and Imprecise Treatment-Response Estimates in an LLM Persona Panel [0.0]
  We study a fixed panel of sixteen lightweight persona-conditioned GPT-4.1 configurations in repeated strategic games.<n>Variation was strongly prompt-indexed, but its share depended on uncertainty assumptions.<n>Results concern one fixed model-prompt panel and do not establish human substitutability.
  arXiv  Detail & Related papers  (2026-08-02T04:10:11Z)
• Automated Proving of Shannon-Type Entropy Inequalities via Fine-Tuned Language Models and Guided Tree Search [50.16356451328644]
  Proving Shannon-type entropy inequalities is a fundamental task in information theory.<n>We investigate whether small-scale large language models, fine-tuned on atomic proof steps, can automate this process.<n>We show that GPT-5.5 solves 1.7% samples under zero-shot prompting while Psitip solves 33.3% samples.
  arXiv  Detail & Related papers  (2026-06-04T05:43:12Z)
• Amplifying, Not Learning: Fine-Tuned AI Text Detectors Amplify a Pretrained Direction [51.56484100374058]
  Text detectors amplify a pretrained typicality axis; they do not construct an AI-vs-human boundary.<n>On raw encoders before any task supervision, projecting onto centroid(AI)-centroid(HC3) achieves NYT-vs-HC3 AUROC 0.806/0.944/0.834 across three architectures.<n>On RoBERTa-base, raw projection exceeds fine-tuning; on RoBERTa-base, full fine-tuning reduces discrimination below raw on both fluent-formal populations tested.
  arXiv  Detail & Related papers  (2026-05-20T19:08:38Z)
• ClinicalBench: Stress-Testing Assertion-Aware Retrieval for Cross-Admission Clinical QA on MIMIC-IV [0.0]
  Reasoning benchmarks measure clinical performance on clean inputs.<n>We evaluate the step before reasoning: retrieval over real EHR notes, where negation, temporality, and family-versus-patient attribution can flip a correct answer to a wrong one.<n>EpiKG carries assertion label and a temporality tag with every fact in a patient knowledge graph, then routes retrieval by question intent.
  arXiv  Detail & Related papers  (2026-05-11T18:47:52Z)
• The Alignment Tax: Response Homogenization in Aligned LLMs and Its Implications for Uncertainty Estimation [1.8345614451086532]
  RLHF-aligned language models exhibit response homogenization on TruthfulQA.<n>40-79% of questions produce a single semantic cluster across 10 i.i.d. samples.
  arXiv  Detail & Related papers  (2026-03-25T09:35:15Z)
• ThReadMed-QA: A Multi-Turn Medical Dialogue Benchmark from Real Patient Questions [5.63130104359934]
  ThReadMed-QA is a benchmark of 2,437 fully-answered patient-seeking conversation threads extracted from r/AskDocs.<n>We evaluate five state-of-the-art LLMs on a stratified test split of 238 conversations.<n>Even the strongest model, GPT-5, achieves only 41.2% fully-correct responses.
  arXiv  Detail & Related papers  (2026-03-11T20:17:57Z)
• PanCanBench: A Comprehensive Benchmark for Evaluating Large Language Models in Pancreatic Oncology [48.732366302949515]
  Large language models (LLMs) have achieved expert-level performance on standardized examinations, yet multiple-choice accuracy poorly reflects real-world clinical utility and safety.<n>We developed a human-in-the-loop pipeline to create expert rubrics for de-identified patient questions.<n>We evaluated 22 proprietary and open-source LLMs using an LLM-as-a-judge framework, measuring clinical completeness, factual accuracy, and web-search integration.
  arXiv  Detail & Related papers  (2026-03-02T00:50:39Z)
• Causal Understanding by LLMs: The Role of Uncertainty [43.87879175532034]
  Recent papers show LLMs achieve near-random accuracy in causal relation classification.<n>We investigate whether pretraining exposure to causal examples improves causal understanding.
  arXiv  Detail & Related papers  (2025-09-24T13:06:35Z)
• Benchmarking Reasoning Robustness in Large Language Models [76.79744000300363]
  We find significant performance degradation on novel or incomplete data.<n>These findings highlight the reliance on recall over rigorous logical inference.<n>This paper introduces a novel benchmark, termed as Math-RoB, that exploits hallucinations triggered by missing information to expose reasoning gaps.
  arXiv  Detail & Related papers  (2025-03-06T15:36:06Z)
• CRTRE: Causal Rule Generation with Target Trial Emulation Framework [47.2836994469923]
  We introduce a novel method called causal rule generation with target trial emulation framework (CRTRE) CRTRE applies randomize trial design principles to estimate the causal effect of association rules. We then incorporate such association rules for the downstream applications such as prediction of disease onsets.
  arXiv  Detail & Related papers  (2024-11-10T02:40:06Z)

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.