---
license: mit
task_categories:
- text-generation
language:
- en
tags:
- medical
- biomedical
- abstract
- conclusion-generation
pretty_name: MedConclusion Compact
size_categories:
- 100K<n<1M
---

# MedConclusion-Compact

**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 **Compact** version of the dataset, designed for faster evaluation and model prototyping. For the full dataset (5.7M instances), please check out the [**Full Version**](https://huggingface.co/datasets/harvardairobotics/MedConclusion).

- **Train**: 100,000 instances
- **Validation**: 10,000 instances
- **Test**: 30,000 instances

## 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}
}
```