LongFinanceQA
收藏资源简介:
LongFinanceQA是一个针对金融领域的长文本理解任务构建的合成数据集,包含46,457个长文本问答对,这些问答对具有高质量的中间推理步骤(CoT)。数据集基于6,911份中英文双语年度财务报告构建,每份样本包括一个实际的长文本问题以及对应的增强答案和中间的推理步骤。该数据集旨在提高长文本上下文推理能力,并为长文本大型语言模型提供高质量的训练数据。
LongFinanceQA is a synthetic dataset constructed for the long-text understanding task in the financial domain. It contains 46,457 high-quality long-text question-answer pairs, all equipped with complete intermediate Chain-of-Thought (CoT) reasoning steps. The dataset is built upon 6,911 Chinese-English bilingual annual financial reports. Each sample comprises a real-world long-text question, paired with the corresponding enhanced answer and intermediate reasoning steps. This dataset is designed to enhance long-text contextual reasoning capabilities and provide high-quality training data for large language models targeting long-text processing tasks.

- 1Facilitating Long Context Understanding via Supervised Chain-of-Thought Reasoning美国罗切斯特大学 · 2025年



