harvardairobotics/MedConclusion

--- license: mit task_categories: - text-generation language: - en tags: - medical - biomedical - abstract - conclusion-generation pretty_name: MedConclusion size_categories: - 1M<n<10M --- # MedConclusion **MedConclusion** is a large-scale dataset of 5.7M PubMed structured abstracts for biomedical conclusion generation. Each instance pairs the non-conclusion sections of an abstract with the original author-written conclusion, providing naturally occurring supervision for evidence-to-conclusion reasoning. MedConclusion also includes journal-level metadata such as biomedical category and SJR, enabling subgroup analysis across biomedical domains. This repository contains the full version of the dataset. For faster evaluation and model prototyping, please check out the [**Compact Version**](https://huggingface.co/datasets/harvardairobotics/MedConclusion-Compact). - **Train**: 70% - **Validation**: 10% - **Test**: 20% ## Benchmark Information This dataset is introduced in the paper [MedConclusion: A Benchmark for Biomedical Conclusion Generation from Structured Abstracts](https://arxiv.org/abs/2604.06505). ## Citation ```bibtex @article{li2026medconclusion, title={MedConclusion: A Benchmark for Biomedical Conclusion Generation from Structured Abstracts}, author={Li, Weiyue and Qian, Ruizhi and Li, Yi and Li, Yongce and Long, Yunfan and Cai, Jiahui and Luo, Yan and Wang, Mengyu}, journal={arXiv preprint arXiv:2604.06505}, year={2026} } ```