Loading...
Loading...
Contributors:
Dataset URI:
10.5281/zenodo.23154472Data and code for the article Excess drift: separating model change from repeat noise when hosted large language models supply species conservation facts (Ji Sun Hong and Yuyong Kim).The article asks whether a change in what a hosted large language model says about a species, between two dates, reflects a change of the model behind an unchanged name or only the variation between repeated calls. It defines excess drift (within-date repeat agreement minus between-date agreement) and applies it to 2,620 stored responses from three companion studies and 540 fresh responses (27 September 2026) from five model names on Korean endangered species, scoring free text and the extracted statutory endangered-species class and IUCN Red List category.The archive contains every stored and fresh response with the requested model name, the returned model identifier, fingerprint and token usage where logged; every result table and figure reported in the article; the numbers printed in the article; the analysis scripts; and the companion-study files the analysis reads (frozen fact extractor, reference values, judge label sets). The analysis can be reproduced from the archive alone (see README.md). Data are released under CC BY 4.0 and code under the MIT License.
Keyword(s):
endangered species; environmental software; IUCN Red List; Korean fauna; large language models.
MeSH Term(s):
Grants and Funding:
No Grants or Funding Data Available
Associated Publication(s):
No Associated Publication Data Available
Supplementary Data:
No Supplementary Data Available
Related Dataset(s):
No Related Dataset Data Available
The National Library of Medicine (NLM) provides information about biomedical datasets. Inclusion on an NLM Web page does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: NLM Web Policies (nih.gov)