LLM Hackathon Outcomes: From Knowledge to Action in Materials Science and Chemistry
In 2025, a hackathon focused on large language models (LLMs) in materials science and chemistry yielded various applications developed by the community. These initiatives are divided into two main categories: Knowledge Infrastructure, which involves systems for organizing, retrieving, synthesizing, and validating scientific data, and Action Systems, which automate or coordinate scientific tasks. A significant trend observed is the transition from tools designed for specific purposes to comprehensive, multi-agent workflows that integrate retrieval, reasoning, tool utilization, and domain-specific validation. Notable themes include retrieval-augmented generation serving as foundational infrastructure and the ongoing representation of structured knowledge.
Key facts
- Hackathon focused on LLM applications in materials science and chemistry.
- Projects categorized as Knowledge Infrastructure and Action Systems.
- Shift from single-purpose tools to multi-agent workflows.
- Retrieval-augmented generation used as grounding infrastructure.
- Persistent structured knowledge representation is a prominent theme.
- Applications cover the full scientific research lifecycle.
- Community-developed applications analyzed for emerging patterns.
- Event took place in 2025.
Entities
Institutions
- arXiv