mlsanityData to 14 Oct 2026

MEDEQUALQA: Evaluating Biases in LLMs with Counterfactual Reasoning

arXiv 2510.12818 ↗First posted 9 Oct 2025cs.CL cs.AI cs.CYcited by 3

On the map: LLM reasoning → Clinical and biomedical applications

Abstract (arXiv)

Large language models (LLMs) are increasingly deployed in clinical decision support, yet subtle demographic cues can influence their reasoning. Prior work has documented disparities in outputs across patient groups, but little is known about how internal reasoning shifts under controlled demographic changes. We introduce MEDEQUALQA, a counterfactual benchmark that perturbs only patient pronouns (he/him, she/her, they/them) while holding critical symptoms and conditions (CSCs) constant. Each clinical vignette is expanded into single-CSC ablations, producing three parallel datasets of approximately 23,000 items each (69,000 total). We evaluate a GPT-4.1 model and compute Semantic Textual Similarity (STS) between reasoning traces to measure stability across pronoun variants. Our results show overall high similarity (mean STS >0.80), but reveal consistent localized divergences in cited risk factors, guideline anchors, and differential ordering, even when final diagnoses remain unchanged. Our error analysis highlights certain cases in which the reasoning shifts, underscoring clinically relevant bias loci that may cascade into inequitable care. MEDEQUALQA offers a controlled diagnostic setting for auditing reasoning stability in medical AI.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Placed in Clinical and biomedical applications, one of 17 directions in LLM reasoning. Builds on 21 of its 26 references that are in the map. First posted on arXiv 9 Oct 2025.

Builds on

21 of 26 references.

References · 21

  1. 2016Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddingscited by 1,312
  2. 2020StereoSet: Measuring stereotypical bias in pretrained language modelscited by 924
  3. 2016Semantics derived automatically from language corpora contain human-like biasescited by 901
  4. 2020CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Modelscited by 710
  5. 2018Gender Bias in Coreference Resolution: Evaluation and Debiasing Methodscited by 659
  6. 2019The Woman Worked as a Babysitter: On Biases in Language Generationcited by 541
  7. 2018Gender Bias in Coreference Resolutioncited by 442
  8. 2021Persistent Anti-Muslim Bias in Large Language Modelscited by 377
  9. 2019Measuring Bias in Contextualized Word Representationscited by 276
  10. 2020GECToR -- Grammatical Error Correction: Tag, Not Rewritecited by 177
  11. 2023Not All Languages Are Created Equal in LLMs: Improving Multilingual Capability by Cross-Lingual-Thought Promptingcited by 165
  12. 2024Linguistic Bias in ChatGPT: Language Models Reinforce Dialect Discriminationcited by 56
9 more
  1. 2024DIALECTBENCH: A NLP Benchmark for Dialects, Varieties, and Closely-Related Languagescited by 50
  2. 2022VALUE: Understanding Dialect Disparity in NLUcited by 47
  3. 2021NeuTral Rewriter: A Rule-Based and Neural Approach to Automatic Rewriting into Gender-Neutral Alternativescited by 46
  4. 2024CoD, Towards an Interpretable Medical Agent using Chain of Diagnosiscited by 43
  5. 2024Dialect prejudice predicts AI decisions about people's character, employability, and criminalitycited by 37
  6. 2024Unmasking and Quantifying Racial Bias of Large Language Models in Medical Report Generationcited by 31
  7. 2024Bias patterns in the application of LLMs for clinical decision support: A comprehensive studycited by 27
  8. 2024AAVENUE: Detecting LLM Biases on NLU Tasks in AAVE via a Novel Benchmarkcited by 11
  9. 2025DeVisE: Behavioral Testing of Medical Large Language Modelscited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026EQUITRIAGE: A Fairness Audit of Gender Bias in LLM-Based Emergency Department Triagecited by 3
  2. 2026MIRA: A Bilingual Benchmark for Medical Information Response Auditcited by 2
  3. 2026Authority Bias in Conversational Search Engines for Academic Paper Recommendationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Evaluating the Presence of Sex Bias in Clinical Reasoning by Large Language Models
  2. 2025Quantifying the Reasoning Abilities of LLMs on Real-world Clinical Cases
  3. 2024Aligning (Medical) LLMs for (Counterfactual) Fairness
  4. 2026First, Do No Harm (With LLMs): Mitigating Racial Bias via Agentic Workflows
  5. 2026MamaBench: Benchmarking LLM Robustness in Maternal and Child Health Diagnosis through Counterfactual Clinical Perturbation
  6. 2026Towards Reliable Medical LLMs: Benchmarking and Enhancing Confidence Estimation of Large Language Models in Medical Consultation
  7. 2026MedArena: Comparing LLMs for Medicine-in-the-Wild Clinician Preferences
  8. 2026Artificial Intolerance: Stigmatizing Language in Clinical Documentation Skews Large Language Model Decision-Making