likhitjuttada/finance-reasoning-sft-dataset

Personal Finance Reasoning Dataset是一个合成的指令调优数据集，旨在通过经典书籍中的心理框架，教导语言模型进行个人财务和投资决策的推理。数据集的目标不是回忆书籍内容，而是原则性推理：模型应该将框架应用于它从未见过的新情况。数据来源于五本经典财务书籍，包括《金钱心理学》、《富爸爸穷爸爸》、《巴比伦最富有的人》、《聪明的投资者》和《思考致富》。数据集包含三种类型的示例：应用问答、思维链和跨书对比。每个示例都是一个JSON对象，包含用户和助手的对话，其中助手的回答包含显式的逐步推理。数据集是通过三个阶段生成的：原则提取、示例生成和质量过滤。数据集适用于监督微调（SFT）的语言模型，特别是那些需要推理个人财务和财富构建框架的模型。数据集有一些局限性，如数据偏差、覆盖不均、仅限英语等。

The Personal Finance Reasoning Dataset is a synthetic instruction-tuning dataset designed to teach language models to reason through personal finance and investing decisions using the mental frameworks from classic books in the genre. The goal is not recall of book content but principled reasoning: the model should apply frameworks to novel situations it has never seen. Data is sourced from five classic finance books, including *The Psychology of Money*, *Rich Dad Poor Dad*, *The Richest Man in Babylon*, *The Intelligent Investor*, and *Think and Grow Rich*. The dataset contains three types of examples: applied Q&A, chain-of-thought, and cross-book contrast. Each example is a JSON object containing a conversation between a user and an assistant, where the assistants response includes explicit step-by-step reasoning. The dataset was generated in three stages: principle extraction, example generation, and quality filtering. It is intended for supervised fine-tuning (SFT) of language models, particularly those that require reasoning about personal finance and wealth-building frameworks. The dataset has some limitations, such as synthetic data bias, uneven coverage, and English-only content.