M&A Agent: Generating Strategic Insight Through Role-Based Agent Simulation

ACL ARR 2025 May Submission3308 Authors

Abstract: Large language models (LLMs) and LLM-based agent systems have shown considerable potential in investment and stock trading. However, their use in corporate finance, particularly in strategic decision-making tasks such as merger and acquisition (M\&A) evaluation, remains underexplored. Traditional analysis and machine learning methods often struggle in this domain due to the limited availability of target company data. To address this challenge, we propose M\&A Agent, a multi-agent framework built on LLMs that simulates the M\&A process and assesses deal value. The framework consists of two stages: M\&A simulation and value evaluation. Following real-world procedures, the simulation includes financial analysis, negotiation, board decision-making, and regulatory review. Through structured interactions among agents, the system transforms static financial data into richer, more dynamic information. This simulation is then reviewed by an evaluation committee of agents, which assigns a score and provides justification. Experiments on real-world M\&A cases demonstrate that our method produces significantly better deal value rankings compared to baselines, as measured by NDCG. The code is publicly available at https://anonymous.4open.science/r/2AB73965 to support reproducibility.

Paper Type: Long

Research Area: Special Theme (conference specific)

Research Area Keywords: LLM/AI agents, applications, financial/business NLP

Contribution Types: NLP engineering experiment

Languages Studied: Chinese

Keywords: LLM/AI agents, applications, financial/business NLP

Submission Number: 3308