A Multi-Agent Retrieval-Augmented Framework for Personalized Government Scheme Recommendation

Abstract: Accessing appropriate government welfare schemes remains a significant challenge due to
fragmented information, complex eligibility criteria, and lack of ersonalization. Citizens often struggle
to identify schemes relevant to their socio-economic profile, resulting in underutilization of public
benefits. This paper proposes a Multi-Agent Retrieval-Augmented Generation (MA-RAG) Framework
designed to deliver personalized government scheme recommendations. The system integrates multiple
intelligent agents with a retrieval-augmented architecture to analyze user profiles, retrieve relevant
policy documents, validate eligibility conditions, and generate explainable recommendations. By
combining semantic search, agent-based reasoning, and large language models, the framework ensures
accurate, context-aware, and citizen-centric scheme discovery. The proposed model enhances
accessibility, transparency, and efficiency in public service delivery.