Fino1: On the Transferability of Reasoning Enhanced LLMs to Finance
Lingfei Qian, Weipeng Zhou, Yan Wang, Xueqing Peng, Jimin Huang, Qianqian Xie
2025-02-13
Summary
This paper talks about Fino1, a new AI model designed specifically to understand and solve complex financial problems. The researchers tested many existing AI models on financial tasks and then created Fino1 to address the shortcomings they found.
What's the problem?
While AI models have gotten really good at general reasoning, they struggle with financial tasks that involve understanding specialized terms, working with tables of data, and solving equations. Even AI models that are good at other types of reasoning don't always do well with financial problems, especially when they involve long texts or multiple tables of information.
What's the solution?
The researchers created Fino1, which is based on an existing AI model called Llama-3.1-8B-Instruct. They trained Fino1 using special techniques that help it understand financial concepts and solve problems step-by-step. They also used something called reinforcement learning, which is like giving the AI practice problems and rewarding it for getting the right answers. Even though Fino1 is smaller than some other AI models, it performed 10% better on financial tasks than its original version and even outperformed some much larger models.
Why it matters?
This matters because as we rely more on AI in the financial world, we need models that can accurately handle complex financial information. Fino1 shows that by focusing on specific areas like finance, we can create AI that's better at specialized tasks. This could lead to more reliable AI tools for financial analysis, which could help make better investment decisions, understand economic trends, and even prevent financial mistakes or fraud. The researchers also made all their work public, which means other scientists can build on it to create even better financial AI tools in the future.
Abstract
Recent advancements in large language models (LLMs) have shown strong general reasoning abilities, yet their effectiveness in financial reasoning remains underexplored. In this study, we comprehensively evaluate 16 powerful reasoning and general LLMs on three complex financial tasks involving financial text, tabular data, and equations, assessing numerical reasoning, tabular interpretation, financial terminology comprehension, long-context processing, and equation-based problem solving. Our results show that while better datasets and pretraining improve financial reasoning, general enhancements like CoT fine-tuning do not always yield consistent gains. Moreover, all reasoning strategies face challenges in improving performance on long-context and multi-table tasks. To address these limitations, we develop a financial reasoning-enhanced model based on Llama-3.1-8B-Instruct, by CoT fine-tuning and reinforcement learning with domain-specific reasoning paths. Even with simple fine-tuning with one financial dataset, our model achieves a consistent 10% performance improvement across tasks, surpassing all 8B models and even Llama3-70B-Instruct and Llama3.1-70B-Instruct on average. Our results highlight the need for domain-specific adaptations in financial tasks, emphasizing future directions such as multi-table reasoning, long-context processing, and financial terminology comprehension. All our datasets, models, and codes are publicly available. Furthermore, we introduce a leaderboard for benchmarking future datasets and models.