2026-09-18T19:36:10Zhttps://dspace.mit.edu/server/oai/request

oai:dspace.mit.edu:1721.1/1646022026-01-21T04:07:55Zcom_1721.1_7582com_1721.1_7581col_1721.1_131023

Williams, Michael McGreivy, James C. Massachusetts Institute of Technology. Department of Physics 2026-01-20T19:48:00Z 2026-01-20T19:48:00Z 2025-09 2025-09-02T19:50:38.797Z https://hdl.handle.net/1721.1/164602 Generative Large Language Models (LLMs) are a promising approach to structuring knowledge contained within otherwise unmanageable corpora of research literature produced by large-scale and long-running scientific collaborations. Within experimental particle physics, such structured knowledge bases could expedite methodological and editorial review. Complementarily, within the broader scientific community, generative LLM systems grounded in published work could make for reliable companions allowing non-experts to analyze openaccess data. Techniques such as Retrieval Augmented Generation (RAG) rely on semantically matching localized text chunks, but struggle to maintain coherent context when relevant information spans multiple segments, leading to a fragmented representation devoid of global cross-document information. In this work I utilize the hierarchical organization of experimental physics articles to build a tree representation of the corpus, and present the SciTreeRAG system which leverages this structure with the aim of constructing contexts more focused and contextually rich than a standard RAG. Additionally, I develop methods for using LLMs to transform the unstructured corpus into a structured knowledge graph representation. I then implement SciGraphRAG, a retrieval system that leverages this knowledge graph to access global cross-document relationships eluding standard RAG, with the goal of encapsulating domain-specific connections and expertise. I demonstrate proof-of-concept implementations of both systems using the corpus of the LHCb experiment at CERN. S.M. Massachusetts Institute of Technology In Copyright - Educational Use Permitted Copyright retained by author(s) https://rightsstatements.org/page/InC-EDU/1.0/ Seeing the Forest Through the Trees: Knowledge Retrieval for Streamlining Particle Physics Analysis Thesis application/pdf Master Master of Science in Physics Publication unknown unknown <Publication xmlns="https://www.openaire.eu/cerif-profile/1.1/" id="e27329f4-63f0-4187-a269-fd98ab9f144b"> <Type xmlns="https://www.openaire.eu/cerif-profile/vocab/COAR_Publication_Types">http://purl.org/coar/resource_type/c_1843</Type> <Title>Seeing the Forest Through the Trees: Knowledge Retrieval for Streamlining Particle Physics Analysis</Title> <PublishedIn> <Publication> </Publication> </PublishedIn> <PublicationDate>2025-09</PublicationDate> <Authors> <Author> <DisplayName>McGreivy, James C.</DisplayName> <Affiliation> <OrgUnit> </OrgUnit> </Affiliation> </Author> </Authors> <Editors> </Editors> <Publishers> <Publisher> <DisplayName>Massachusetts Institute of Technology</DisplayName> <OrgUnit /> </Publisher> </Publishers> <License>https://rightsstatements.org/page/InC-EDU/1.0/</License> <Abstract>Generative Large Language Models (LLMs) are a promising approach to structuring knowledge contained within otherwise unmanageable corpora of research literature produced by large-scale and long-running scientific collaborations. Within experimental particle physics, such structured knowledge bases could expedite methodological and editorial review. Complementarily, within the broader scientific community, generative LLM systems grounded in published work could make for reliable companions allowing non-experts to analyze openaccess data. Techniques such as Retrieval Augmented Generation (RAG) rely on semantically matching localized text chunks, but struggle to maintain coherent context when relevant information spans multiple segments, leading to a fragmented representation devoid of global cross-document information. In this work I utilize the hierarchical organization of experimental physics articles to build a tree representation of the corpus, and present the SciTreeRAG system which leverages this structure with the aim of constructing contexts more focused and contextually rich than a standard RAG. Additionally, I develop methods for using LLMs to transform the unstructured corpus into a structured knowledge graph representation. I then implement SciGraphRAG, a retrieval system that leverages this knowledge graph to access global cross-document relationships eluding standard RAG, with the goal of encapsulating domain-specific connections and expertise. I demonstrate proof-of-concept implementations of both systems using the corpus of the LHCb experiment at CERN.</Abstract> <Access xmlns="http://purl.org/coar/access_right" > </Access> </Publication>