Data and analysis outputs for: A systematic evaluation and benchmarking of text summarization methods for biomedical literature

Résumé fourni par la source

Data and analysis outputs supporting the systematic benchmarking of 62 automatic text summarization methods on 1,000 biomedical abstracts from ScienceDirect and Cell Press, using author-written highlights as reference summaries. Includes the gold-standard dataset (1,000 titles, abstracts, and author-written highlights), all generated summaries with per-paper and per-model metric scores across lexical, semantic, and factual dimensions, raw ratings from eight expert evaluators, LLM-as-a-judge panel scores, the data-leakage analyses, and the aggregated ranking and agreement tables reported in the paper. A snapshot of the analysis code is included as a tarball; the code is archived separately at DOI 10.5281/zenodo.21839980 and maintained at github.com/Delta4AI/LLMTextSummarizationBenchmark.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.