Researchers from the Carnegie Mellon University (#CMU) just published information on how to create an #OpenSource literature review system competitive with commercial #LLMs.

#OpenScholar leverages a database of 45 million #OpenAccess papers, directly linking claims to source articles to reduce hallucinated references. OpenScholar is freely available and can be run locally or adapted to enhance other LLMs.

https://openscilm.allen.ai/

#Research #Science #Academia

Can language models synthesize scientific literature? by Ai2

Demo of fully open, retrieval-augmented language model. Synthesize 8M+ open access research papers to answer scientific questions. By the researchers at the University of Washington and Semantic Scholar.

@mvugt

Interesting that #OpenScholar relies on #OpenAccess only. So all literature behind paywalls is ignored.

#ScientificPublishing

@SciMag @news-from-science-SciMag

"Although GPT-4o hallucinates citations 78–90% of the time, OpenScholar achieves citation accuracy on par with human experts."

Now if only it wasn't boiling the oceans over.

"Synthesizing scientific literature with retrieval-augmented language models", Asai et al. 2026 https://www.nature.com/articles/s41586-025-10072-4

#OpenScholar #LLMs #ScientificLiterature #science

Synthesizing scientific literature with retrieval-augmented language models - Nature

A 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.

Nature

"an artificial intelligence (AI) program designed specifically to analyze the scientific literature. ...#OpenScholar..."

"OpenScholar’s creators at the Allen Institute"

#science

https://www.science.org/content/article/open-source-ai-program-can-answer-science-questions-better-humans

"Transforming Science with Large Language Models: A Survey on AI-assisted Scientific Discovery, Experimentation, Content Generation, and Evaluation"

Wonderful paper I found from my IArxiv.org feed. It gives nice comparison between features of the leading AI search engines.

#research #Elicit #SciSpace #Undermind #OpenScholar #Perplexity #OpenAlex #Scilit
#AceMap #ConnectedPapers #CiteSpace #Consensus #SemanticScholar #Lens
#ResearchRabbit #Baidu
#notesumAI #ResearchTrend
#ORKG

#開源分享 一款輔助閱讀和理解科研論文的AI工具,一個科研助手:OpenScholar

它可以基於提問自動尋找相關的論文,並把論文內容消化後用通俗易懂的方式回,且會標註資訊來源,解決每年新發表的論文太多看不過來的問題

專業工具,專門理解和處理專業學術內容

支持標準RAG流程,包含檢索器+重排序器管道,具備自反思生成能力

專案地址: github.com/AkariAsai/OpenScholar

#RAG #論文閱讀工具 #OpenScholar

Update. "The #OpenScholar team has released not only the code for the language model but also the entire retrieval pipeline, a specialized 8-billion-parameter model fine-tuned for scientific tasks, and a datastore of [#OpenAccess] scientific papers. 'To our knowledge, this is the first open release of a complete pipeline for a scientific assistant LM —from data to training recipes to model checkpoints,' the researchers wrote in their blog post announcing the system."
https://venturebeat.com/ai/openscholar-the-open-source-a-i-thats-outperforming-gpt-4o-in-scientific-research/

#LLM #OpenSource

OpenScholar: The open-source A.I. that’s outperforming GPT-4o in scientific research

OpenScholar, an innovative AI system by Allen Institute for AI and University of Washington, revolutionizes scientific research by processing 45 million papers instantly, offering researchers citation-backed answers and challenging proprietary AI systems.

VentureBeat

A new open-source AI system, OpenScholar, is promising to rewrite the rules for how researchers access, evaluate, and synthesize scientific literature. It even outperforms much larger proprietary models like GPT-4o.

#openscholar #opensource #ai #gpt
https://venturebeat.com/ai/openscholar-the-open-source-a-i-thats-outperforming-gpt-4o-in-scientific-research/

OpenScholar: The open-source A.I. that’s outperforming GPT-4o in scientific research

OpenScholar, an innovative AI system by Allen Institute for AI and University of Washington, revolutionizes scientific research by processing 45 million papers instantly, offering researchers citation-backed answers and challenging proprietary AI systems.

VentureBeat
I want to publish in open journals. I am also limited in my budget. Even the lowest APC is unaffordable. I have the work, now what? #PhDLife #OpenScholar #AcademicChatter
@brembs Time to stop publishing with them #OpenScholar