| Issue |
EPJ Web Conf.
Volume 337, 2025
27th International Conference on Computing in High Energy and Nuclear Physics (CHEP 2024)
|
|
|---|---|---|
| Article Number | 01279 | |
| Number of page(s) | 8 | |
| DOI | https://doi.org/10.1051/epjconf/202533701279 | |
| Published online | 07 October 2025 | |
https://doi.org/10.1051/epjconf/202533701279
AccGPT: A CERN Knowledge Retrieval Chatbot
CERN
1 Corresponding author: juan.manuel.guijarro@cern.ch
Published online: 7 October 2025
In the vast landscape of CERN’s internal documentation, finding and accessing relevant detailed information remains a complex and timeconsuming task. To address this challenge, the AccGPT project (Accelerating Science GPT) aims for the development of an intelligent chatbot leveraging Natural Language Processing (NLP) technologies. We utilize open-source Large Language Models (LLMs) to create a specialized chatbot for CERN internal text-based knowledge retrieval, with the potential future extensions to code assistance and other functionalities.
A promising first prototype utilising a Retrieval Augmented Generation (RAG) pipeline has already been developed and deployed. Ongoing improvements focus on enhancing the retrieval accuracy, integrating more powerful and larger LLMs, or fine-tuning with domain-specific data to improve domain accuracy and relevance.
The chatbot’s user interface design and overall experience are being iteratively improved, and efforts are underway to prepare AccGPT for community-wide testing. Additionally, automated data scraping and preprocessing pipelines are being implemented to ensure an up-to-date, selfsustaining knowledge base.
© The Authors, published by EDP Sciences, 2025
This is an Open Access article distributed under the terms of the Creative Commons Attribution License 4.0, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.