Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Large Language Model Integration for Knowledge Retrieval and Interaction for the DUNE Experiment

A. Rafique, A. Singh, R. Srinivas

TL;DR

This paper presents DUNE-GPT, an internal retrieval-augmented generation system enabling natural-language querying of DUNE's scattered documentation while preserving data privacy within Fermilab infrastructure. It describes a three-component pipeline—document processing and embeddings, a FAISS-based vector retriever, and an LLM interface—that grounds responses in verified DUNE sources using embeddings from $multi ext{-}qa ext{mpnet-base-dot-v1}$ and LLMs hosted on Argo and Ollama. Preliminary benchmarks report ~70% retrieval accuracy across detector, reconstruction, and data-analysis queries, with plans to expand the corpus, support multi-modal content, and deploy collaboration-wide. The work aims to boost collaboration productivity and onboarding by providing explainable, citation-grounded answers without exposing data externally.

Abstract

The Deep Underground Neutrino Experiment (DUNE) is a next-generation neutrino experiment that will generate an unprecedented volume of heterogeneous information-from documentation and technical notes to experimental data and reconstruction pipelines. Efficient knowledge retrieval and contextual understanding are increasingly critical for collaboration-wide productivity and onboarding. In this work, we present DUNE-GPT, a prototype framework that leverages large language models (LLMs) and retrieval-augmented generation (RAG) to enable natural-language querying of DUNE's internal documentation and technical resources. The system provides an intelligent interface for DUNE collaborators to interact with experiment-specific knowledge while maintaining data privacy and infrastructure compliance within Fermilab computing resources.

Large Language Model Integration for Knowledge Retrieval and Interaction for the DUNE Experiment

TL;DR

This paper presents DUNE-GPT, an internal retrieval-augmented generation system enabling natural-language querying of DUNE's scattered documentation while preserving data privacy within Fermilab infrastructure. It describes a three-component pipeline—document processing and embeddings, a FAISS-based vector retriever, and an LLM interface—that grounds responses in verified DUNE sources using embeddings from and LLMs hosted on Argo and Ollama. Preliminary benchmarks report ~70% retrieval accuracy across detector, reconstruction, and data-analysis queries, with plans to expand the corpus, support multi-modal content, and deploy collaboration-wide. The work aims to boost collaboration productivity and onboarding by providing explainable, citation-grounded answers without exposing data externally.

Abstract

The Deep Underground Neutrino Experiment (DUNE) is a next-generation neutrino experiment that will generate an unprecedented volume of heterogeneous information-from documentation and technical notes to experimental data and reconstruction pipelines. Efficient knowledge retrieval and contextual understanding are increasingly critical for collaboration-wide productivity and onboarding. In this work, we present DUNE-GPT, a prototype framework that leverages large language models (LLMs) and retrieval-augmented generation (RAG) to enable natural-language querying of DUNE's internal documentation and technical resources. The system provides an intelligent interface for DUNE collaborators to interact with experiment-specific knowledge while maintaining data privacy and infrastructure compliance within Fermilab computing resources.
Paper Structure (10 sections, 3 figures)

This paper contains 10 sections, 3 figures.

Table of Contents

  1. Introduction
  2. System Overview
  3. Data Sources
  4. Embedding and Retrieval Layer
  5. Large Language Model Integration
  6. Implementation and Deployment
  7. Preliminary Results
  8. Future Work
  9. Acknowledgments
  10. References

Figures (3)

  • Figure 1: Example of input and output types for a Large Language Model (LLM).
  • Figure 2: Overview of the Retrieval-Augmented Generation (RAG) pipeline. Embeddings are created first, followed by retrieval of the most relevant documents upon user query, which are then passed to the LLM for final response generation.
  • Figure 3: Frontend web interface showing a sample question, response, and the top three retrieved references used in response generation.