INVESTORBENCH
收藏资源简介:
INVESTORBENCH是由斯蒂文斯理工学院等机构开发的一个开源基准,旨在评估基于大型语言模型的金融决策代理。该数据集包含三个主要部分:股票市场环境、加密货币市场环境和ETF市场环境。每个环境都整合了来自多个来源的数据,如Yahoo Finance的股票数据、SEC EDGAR的公司报告以及CoinMarketCap的加密货币数据。数据集通过整合多模态数据,提供了一个全面的市场环境,用于评估金融代理的决策能力。数据集的应用领域主要集中在金融投资决策,旨在通过模拟真实市场环境,提升金融代理的决策质量和适应性。
INVESTORBENCH is an open-source benchmark developed by Stevens Institute of Technology and other institutions, designed to evaluate large language model-powered financial decision-making agents. This dataset comprises three core modules: stock market environment, cryptocurrency market environment, and ETF market environment. Each module integrates multi-source data, including stock data from Yahoo Finance, corporate reports from SEC EDGAR, and cryptocurrency data from CoinMarketCap. By consolidating multimodal data, the dataset delivers a holistic market simulation environment for assessing the decision-making proficiency of financial agents. The primary application domain of this benchmark is financial investment decision-making, where it aims to enhance the decision quality and adaptive capacity of financial agents via simulated real-world market scenarios.

- 1INVESTORBENCH: A Benchmark for Financial Decision-Making Tasks with LLM-based Agent斯蒂文斯理工学院, 哥伦比亚大学, 哈佛大学, The Fin AI · 2024年



