Reproduce “GRASP: Municipal Budget AI Chatbots for Enhancing Civic Engagement” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jerry Xu, Justin Wang, Joley Leung, Jasmine Gu
The GRASP framework introduces an AI chatbot designed to enhance civic engagement by providing accurate, grounded, and citable answers to municipal budget queries. It achieves 78% accuracy in retrieving information from official documents, outperforming general LLMs like GPT-4o (60%) and Google Gemini (35%) by integrating Retrieval-Augmented Generation (RAG) and an agentic workflow.
Key insights
- Hybrid AI models integrating RAG and agentic workflows (like ReAct) are highly effective in overcoming LLM limitations such as hallucinations and knowledge cutoffs in specialized, dynamic knowledge domains.
- Strategic prompt engineering and the incorporation of domain-specific knowledge are critical for enabling AI systems to accurately interpret complex documents and deliver contextually relevant information.
- Providing precise, page-level citations directly from source documents significantly enhances the transparency, trustworthiness, and verifiability of AI-generated responses in public sector applications.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.