{
"$type": "site.standard.document",
"coverImage": {
"$type": "blob",
"ref": {
"$link": "bafkreiego3p5hjbkacjyhccmbehk54d4zdxsqhitzleizanbbq6hlpkh3q"
},
"mimeType": "image/png",
"size": 125444
},
"path": "/articles/s41586-025-10072-4",
"publishedAt": "2026-02-05T05:04:54.644Z",
"site": "https://www.nature.com",
"tags": [
"doi:10.1038/s41586-025-10072-4"
],
"textContent": "Nature, Published online: 04 February 2026; doi:10.1038/s41586-025-10072-4\n\nA specialized, open-source, retrieval-augmented language model is introduced for answering scientific queries and synthesizing literature, the responses of which are shown to be preferred by human evaluations over expert-written answers.",
"title": "Synthesizing scientific literature with retrieval-augmented language models",
"updatedAt": "2026-02-04T00:00:00.000Z"
}