Evaluating large language models for diagnostic reasoning from unstructured clinical narratives in epilepsy

Meghal Dani, Muthu Jeyanthi Prakash, Filip Rosa, Zeynep Akata, Stefanie Liebe · 2026

Read Evaluating large language models for diagnostic reasoning from unstructured clinical narratives in epilepsy by Meghal Dani, Muthu Jeyanthi Prakash, Filip Rosa, Zeynep Akata, Stefanie Liebe (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13197429.txt.

Description

Meghal Dani and four co-authors test eight large language models on real diagnostic reasoning in epilepsy, using messy clinical stories rather than exam questions. Most model evaluations use tidy question banks, which flatter the models. Their 2026 study gives the models seizure descriptions drawn from unstructured patient narratives, filtered and standardized. Each model ranks seven possible seizure onset zones by likelihood. Two of the eight models are medical specialists; the rest are general systems, from GPT-3.5 and GPT-4 to open models.

Localizing where seizures start guides surgery, so errors here are not academic. Narrative text is how patients actually present, rambling and incomplete. The authors benchmark reasoning under that realism. Their study supplies neurologists with evidence about which models reason safely. It is an evaluation paper, and its verdicts are meant to be rerun as models change.

Catalog record

CREATORMeghal Dani, Muthu Jeyanthi Prakash, Filip Rosa, Zeynep Akata, Stefanie Liebe
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE52 KB
VERSION1.0
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC13197429

Browse the full library · About the Tower of Babel