Finance-LLMs: RAG multi-agent framework for financial analysis

Abstract

In this work, we introduce FIND-MA (Financial Insight via a Network of Distributed Multi Agents), a Retrieval-Augmented Generation (RAG)-based multi-agent framework for funda mental company analysis aimed at enhancing financial decision-making. FIND-MA leverages the reasoning capabilities of state-of-the-art large language models, including DeepSeek-R1 and Qwen3, which integrate structured inference with deep contextual understanding. The frame work orchestrates a network of specialized agents, each assigned to assess a specific aspect of a company’s profile—such as financial health, market positioning, leadership dynamics, and strategic outlook. These agents communicate via a shared memory and dialogue mechanism, enabling collaborative analysis and synthesis. By aggregating diverse insights across modules, FIND-MA produces explainable and data-driven evaluations to support investors, analysts, and stakeholders. This work advances the development of trustworthy AI systems for financial due diligence and corporate valuation.