Automated scientific discovery with large language models is transforming the research lifecycle from ideation to experimentation, yet existing agents struggle to autonomously process raw data collected from scientific experiments. We introduce SciDER, a data-centric end-to-end system that automates the research lifecycle. Unlike traditional frameworks, our specialized agents collaboratively parse and analyze raw scientific data, generate hypotheses and experimental designs grounded in specific data characteristics, and write and execute corresponding code. Evaluation on three benchmarks shows SciDER excels in specialized data-driven scientific discovery and outperforms general-purpose agents and state-of-the-art models through its self-evolving memory and critic-led feedback loop. Distributed as a modular Python package, we also provide easy-to-use PyPI packages with a lightweight web interface to accelerate autonomous, data-driven research and aim to be accessible to all researchers and developers.
@article{arxiv.2603.01421,
title = {SciDER: Scientific Data-centric End-to-end Researcher},
author = {Ke Lin and Yilin Lu and Shreyas Bhat and Xuehang Guo and Junier Oliva and Qingyun Wang},
journal= {arXiv preprint arXiv:2603.01421},
year = {2026}
}
Comments
The submission was made prematurely, and the authors need to perform additional analysis and resolve the issues related to authorship before making the work public