In a study, AI model OpenScholar synthesizes scientific research and cites sources as accurately as human experts

OpenScholar, an open-source AI model developed by a UW and Ai2 research team, synthesizes scientific research and cites sources as accurately as human experts. It outperformed other AI models, including GPT-4o, on a benchmark test and was preferred by scientists 51% of the time. The team is working on a follow-up model, DR Tulu, to improve on OpenScholar’s findings.