A Stanford system called Paper2Agent turns static research papers into interactive AI agents that can answer questions and work with other papers' agents.
For centuries, scientific discoveries have lived quietly in published pages, waiting for someone to read, understand and connect them to other work. Stanford researchers are testing a different approach with Paper2Agent, a system that transforms a scientific paper, its code and its data into an interactive AI agent.
The most striking part happens when agents built from different papers interact directly. Researchers demonstrated this by combining methods and data from separate studies to prioritize a candidate gene linked to a genetic variant associated with psoriasis risk, something that previously would have required a human to manually find the connection.
The team has already built more than 100 scientific agents and successfully converted 74 of 100 computational biology papers tested, each one validated to reproduce reference results. The article is clear that humans still drive hypothesis generation and interpretation, but the technology hints at a future where scientific knowledge is no longer isolated on separate pages.
- Paper2Agent converts research papers into interactive AI agents using their code and data
- Agents from different papers can collaborate to find connections humans might miss
- Tested on a genetic variant linked to psoriasis risk, published in Nature
- 74 of 100 computational biology papers were successfully converted and validated
- Over 100 agents already exist, with human validation still essential
