investment-banking
收藏资源简介:
Playgent-IB-Bench 是一个用于评估AI代理在现实世界投资银行任务上表现的基准数据集。每个任务将代理置于模拟的交易环境中(包含金融文档、Excel模板和支持数据的ZIP存档),并要求其完成分析师级别的工作产品。数据集包含40个任务,每个任务配有提示和世界分配。验证器目录包含按任务分级的验证检查,世界目录则包含交易环境的ZIP存档。数据集适用于投资银行、强化学习、代理、金融、评估和基准测试等场景。数据集规模为小于1K,语言为英语,许可证为MIT。
Playgent-IB-Bench is a benchmark dataset designed to evaluate the performance of AI Agents on real-world investment banking tasks. Each task places the agent within a simulated trading environment, which encompasses financial documents, Excel templates, and ZIP archives of supporting data, and requires the agent to complete analyst-level work products. The dataset contains 40 tasks, each paired with a prompt and world assignment. The validator directory contains task-level validation checks, while the world directory holds the ZIP archives of the trading environments. This dataset is applicable to scenarios including investment banking, reinforcement learning, AI agents, finance, evaluation, and benchmark testing. The dataset has a size of less than 1K, is in English, and is licensed under the MIT License.



