TheFinAI/Fin-o1-8B

Overview

Fin-o1-8B is an 8 billion parameter financial reasoning model developed by TheFinAI, built upon the Qwen3-8B base model. It is specifically fine-tuned for financial numerical reasoning and question answering over financial text and tables. The model's training methodology involves supervised fine-tuning (SFT) on financial chain-of-thought data, followed by GRPO (Generative Reinforcement Learning with Policy Optimization) reinforcement learning, incorporating accuracy and format rewards.

Key Capabilities

• Financial Numerical Reasoning: Excels at tasks requiring calculations and logical deduction within financial contexts.
• Question Answering: Designed to answer questions based on financial text and tables.
• Chain-of-Thought Processing: Benefits from training on reasoning paths derived from datasets like FinQA, TAT-QA, DocMath-Eval, Econ-Logic, and BizBench-QA.
• Extended Context: Supports a substantial context length of 40,960 tokens, enabling processing of longer financial documents.

Good for

• Research in Financial AI: Ideal for academic and industrial research into financial language models and reasoning.
• Baseline for Financial Reasoning Models: Serves as a strong foundation for developing and evaluating new financial AI systems.
• Analyzing Financial Documents: Useful for tasks involving extracting and reasoning over information from financial reports and data.

It is important to note that while powerful, Fin-o1-8B is intended for research and should not be used for investment advice or automated financial decisions without human oversight, as it can produce incorrect calculations or hallucinations.