Large Language Model Agents for Investment Management: Foundations, Benchmarks, and Research Frontiers

Abstract

Recent advances in Large Language Models (LLMs) have triggered a new wave of intelligent financial agents capable of complex reasoning, tool use, and autonomous decision-making. This survey presents a comprehensive review of LLM-based agents in the context of investment and trading, focusing on applications such as portfolio optimization, risk management, information retrieval, and automated strategy generation. We systematically categorize the literature by use case and architectural innovations including multiagent collaborations, reflection mechanisms, and tool-augmented pipelines. Additionally, we review emerging evaluation frameworks and benchmark datasets tailored to finance-specific agent tasks. The survey identifies current trends, technical limitations, and open challenges related to robustness, explainability, and real-world deployment. We conclude with emerging directions for building more capable, adaptive, and trustworthy financial AI agents aligned with the demands of modern investment ecosystems.

Keywords: LLM Agents, Investment and Trading Strategy, Portfolio Optimization, Risk Management