iamwillferguson/StockSensei_Ground_Truth

该数据集旨在为开发者提供一个基准评估套件，用于在金融背景下微调大型语言模型。数据集包含两个文件：一个是问题和答案对的地面真值数据集，另一个是在SEC 10-K文件上微调多个LLM后的评估结果。

This is a ground-truth dataset for benchmarking evaluation of fine-tuned Large Language Models (LLMs) in the context of financial advice, containing question and answer pairs. The dataset was created to address the rise in day trading and social media financial advice, and the lack of open-source resources for training models for such tasks. It includes question and answer pairs created using Gemini-1.5-Flash and Langchain, as well as evaluation results of several LLMs fine-tuned on SEC 10-K filings. Future work includes developing a larger, more in-depth ground-truth dataset, and researching bespoke numerical metrics for applications related to financial advice.