FinBen
收藏资源简介:
FinBen是由武汉大学开发的首个全面开源评估基准,专门设计用于彻底评估大型语言模型(LLMs)在金融领域的能力。该数据集包含35个子数据集,覆盖23个金融任务,旨在评估LLMs在归纳推理、联想记忆、定量推理等方面的认知能力。FinBen不仅涵盖了传统的NLP任务,如文本分析和知识提取,还创新性地评估了LLMs在实际金融应用中的表现,如股票预测和信用评分,首次直接评估了LLMs的交易性能。此数据集的应用领域广泛,旨在解决金融领域中复杂的信息处理和决策问题。
FinBen is the first comprehensive open-source evaluation benchmark developed by Wuhan University, tailored to thoroughly evaluate the capabilities of Large Language Models (LLMs) in the financial domain. This dataset includes 35 sub-datasets spanning 23 financial tasks, designed to assess the cognitive abilities of LLMs such as inductive reasoning, associative memory, and quantitative reasoning. Beyond covering traditional NLP tasks including text analysis and knowledge extraction, FinBen also innovatively evaluates LLMs' performance in practical financial applications such as stock prediction and credit scoring, and marks the first attempt to directly assess the trading performance of LLMs. With broad application scenarios, this benchmark aims to address complex information processing and decision-making issues in the financial field.




