HypGen
Infinite Autonomous Research Generator designed to produce and validate scientific hypotheses at scale.
Last updated
Infinite Autonomous Research Generator designed to produce and validate scientific hypotheses at scale.
Last updated
HypGen operates as a continuously evolving research engine, leveraging autonomous agents, large language models, and a dynamically curated knowledge graph. The system is built to autonomously generate scientific hypotheses and research proposals, with an initial focus on longevity science. This project is the result of a collaboration with coordination.network.
The first version of HypGen is inspired by the SciAgents paper (Ghafarollahi & Buehler, 2024), with its approach adapted specifically for longevity research. This initial iteration lays the foundation for continuous refinement, allowing us to gather feedback, expand datasets, and systematically evaluate our methodology.
By analyzing successes and limitations, we iteratively enhance autonomous knowledge generation, validation, and agentic workflows. Each development cycle brings us closer to a more robust, intelligent system—one informed by real-world data and community insights.
To further improve accuracy and relevance, we engage domain experts through coordination.network, integrating their expertise into our generative system. This collaborative approach ensures high-quality, expert-validated improvements, bridging the gap between automated discovery and human insight.
Like in SciAgents, we begin with a knowledge graph built from ~1,000 scientific papers, but we are committed to expansion. Autonomous agents continuously curate and update the dataset, ensuring high-quality, up-to-date research inputs. New papers are added daily, and we periodically publish open-source snapshots of the dataset.
Ghafarollahi, A., & Buehler, M. J. (2024). SciAgents: Automating scientific discovery through multi-agent intelligent graph reasoning (No. arXiv:2409.05556). arXiv. https://doi.org/10.48550/arXiv.2409.05556
Current System Methodology for Generating Scientific Hypothesis
Current Methodology for Constructing Knowledge Graph
