Fin-R1: A Large Language Model for Financial Reasoning through Reinforcement Learning

2025-03-20

AI Summary

To address the limitations of large language models (LLMs) in complex financial tasks—such as multi-step calculation, logical reasoning, and decision understanding—this paper introduces Fin-R1, a finance-specialized reasoning LLM. Methodologically, Fin-R1 employs a novel two-stage domain-specific training framework: Stage I constructs a high-quality, structured financial reasoning dataset via knowledge distillation from DeepSeek-R1; Stage II integrates supervised fine-tuning (SFT) and Proximal Policy Optimization (PPO)-based reinforcement learning to inject domain knowledge and enhance reasoning robustness. Empirical results demonstrate that Fin-R1 (7B) achieves state-of-the-art performance on FinQA and ConvFinQA, outperforming larger models. It also delivers substantial accuracy improvements in real-world financial tasks—including financial statement analysis and investment decision simulation—validating both the efficacy of domain-customized reasoning architectures and their generalization capability across diverse financial reasoning scenarios.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Reasoning large language models are rapidly evolving across various domains. However, their capabilities in handling complex financial tasks still require in-depth exploration. In this paper, we introduce Fin-R1, a reasoning large language model specifically designed for the financial sector. Fin-R1 is built using a two-stage architecture, leveraging a financial reasoning dataset distilled and processed based on DeepSeek-R1. Through supervised fine-tuning (SFT) and reinforcement learning (RL) training, it demonstrates performance close to DeepSeek-R1 with a parameter size of 7 billion across a range of financial reasoning tasks. It achieves the state-of-the-art (SOTA) in the FinQA and ConvFinQA tasks between those LLMs in our evaluation, surpassing larger models in other tasks as well. Fin-R1 showcases strong reasoning and decision-making capabilities, providing solutions to various problems encountered in the financial domain. Our code is available at https://github.com/SUFE-AIFLM-Lab/Fin-R1.

Problem

• Develops Fin-R1 for financial reasoning tasks
• Enhances financial decision-making using reinforcement learning
• Achieves SOTA in FinQA and ConvFinQA benchmarks

Innovation

• Two-stage architecture for financial reasoning
• Supervised fine-tuning and reinforcement learning
• State-of-the-art performance in financial tasks