Paper2Agent turns scientific papers into AI research tools

Scientific papers could soon do more than sit on a shelf. A system called Paper2Agent takes a study's code, methods and data and bundles them into a tool set an AI assistant can use to work with new data, researchers report in Nature.

James Zou, a computer scientist at Stanford University, calls it essentially "a virtual author of the paper" that is available around the clock.

In tests, Zou and colleagues tried to convert 100 computational biology papers into working tool sets and succeeded with 74. The rest ran into obstacles such as missing code or data, or software that wouldn't run. For the converted papers, the tool set could run analyses and answer questions about the results - more accurately and more than twice as fast as the same agent working from the original papers and code.

Peter Clark, a senior research director at the Allen Institute for Artificial Intelligence in Seattle who was not involved in the work, says making another scientist's code usable can take real effort; he often spends half a day figuring out how to get shared code running. Paper2Agent instead packages a paper's code into reusable tools, tests them against reference outputs and writes step-by-step instructions for an AI assistant. He calls it "an exciting step" toward papers that actively help scientists.

Zou frames the shift as the next stage in how knowledge is passed along - from stone tablets and papyrus to paper, then digital formats, and now from passive to active. He says his team has already built over 100 paper-agents from its own work, and that these have helped him find collaborators.

Clark notes the system specializes in helping scientists reuse methods and code, not in answering general questions as a true virtual author should. He also worries about millions of low-quality papers becoming agents that pester scientists, potentially making things "worse rather than better." For open-ended questions, he says he would still write to the researchers themselves - "for precisely the things they didn't put in the paper."