FinGLM

FinGLM: 致力于构建一个开放的、公益的、持久的金融大模型项目，利用开源开放来促进「AI+金融」.目前开源数据有 70G/1w+份年报数据、10000条人工标注评测数据等。

FinGLM: A project dedicated to developing an open, public-welfare-oriented and sustainable financial large language model (LLM), which advocates open-source and open-access principles to facilitate the integration of 'AI + Finance'. Currently, its open-sourced datasets consist of 70 gigabytes of annual report data with over 10,000 documents, as well as 10,000 manually annotated evaluation data entries, and so on.

搜集汇总

背景与挑战

背景概述

FinGLM是一个专注于金融领域的开源大模型数据集，包含约70GB的上市公司年报数据（涵盖2019-2021年共11588份报告）和10000条人工标注的评测数据，旨在支持文本生成和对话任务。该数据集致力于通过AI技术深度解析金融文本，促进“AI+金融”的应用，并提供公益性的学习资源和项目框架。