Author(s)
Songwei Dong, Bingyan Lu, Makayla Kienlen, Nick Laneman, Caleb Reinking, and Cong Shen
Abstract
Spectrum policy making requires reviewing large volumes of information from different sources, including technical reports, stakeholder filings and incumbent licenses, which makes the process slow and resource-intensive. To address this problem, we propose a multi-modal agentic retrieval-augmented generation (RAG) system for analyzing spectrum proceeding documents and license data. Through customized prompts and tailored tools, the system demonstrates the ability to handle complex queries and perform multi-step retrieval and analysis. For evaluation, we built a benchmark dataset in the form of question–answer pairs and used it to compare our system with others. Results show that the customized RAG system significantly improves the understanding of spectrum data, providing more reliable and verifiable support for spectrum policy analysis.