[arXiv]score: 0.19

Counterfactual Audit of Open-Source LLMs for Clinical Triage Bias

October 2, 2026

A comparative audit of ten open-source LLMs, including MedGemma and Qwen2.5 variants, evaluates bias in pediatric Emergency Severity Index prediction. The study uses paired counterfactual vignettes to test how changing demographic or socioeconomic variables affects clinical acuity assignments.

HOW THIS AFFECTS YOU

●

researcherThis provides a framework for conducting counterfactual audits on domain-specific LLMs.

●

policyThis research underscores the need for rigorous bias auditing in models used for high-stakes medical triage.

●

healthYou should be aware of potential demographic biases when using open-source models for clinical decision support.

read original ↗arxiv.org

DAILY DIGEST

you don't check 9 sources — we do. one email every morning, read in 2 min. free. unsubscribe anytime. privacy

← back to feed