Cross-lingual consistency of large language models in ICD-11 coding: benchmark code and data (six languages, six models)
收藏资源简介:
Reproducible benchmark accompanying the manuscript "Cross-lingual consistency of large language models in ICD-11 coding: a six-language benchmark." Includes the scoring and figure-generation code, the per-call results log (results_prod.csv) for six LLMs across six languages, and an item manifest (entities_manifest.json) listing the 300 sampled ICD-11 entities as code, official ICD-API URL, and English title. To respect the ICD-11 license (CC BY-ND 3.0 IGO), the WHO definition text is not redistributed; the official multilingual titles and definitions can be retrieved from the WHO ICD-API using each entity's URL, or by re-running the pipeline with the fixed seed. Measurement date: June 2026; ICD-11 release 2025-01.Author, affiliation and venue details are withheld to preserve review anonymity and will be restored upon acceptance.



