OpenScholar: a retrieval-augmented language model for synthesizing scientific literature

nature.com

A new retrieval-augmented language model, OpenScholar, synthesizes scientific literature by retrieving information from 45 million papers and providing citation-backed answers. OpenScholar outperforms GPT-4o in correctness on a new benchmark, ScholarQABench, and achieves human-expert level citation accuracy, with experts preferring its responses. The system is open-sourced, including its code, models, and datasets, aiming to advance research in scientific literature synthesis.


With a significance score of 5.7, this news ranks in the top 0.5% of today's 30649 analyzed articles.

Get summaries of news with significance over 5.5 (usually ~10 stories per week). Read by 10,000+ subscribers: