gtfintechlab/KG-MuLQA-D
收藏Hugging Face2026-01-13 更新2026-02-07 收录
下载链接：
https://hf-mirror.com/datasets/gtfintechlab/KG-MuLQA-D
下载链接
链接失效反馈官方服务：
资源简介：
KG-MuLQA-D数据集是一个基于SEC信用协议（2013-2022年）的问答对集合，包含20,139个问答对，分为五个复杂度级别。每个问答对都标有复合复杂度级别（L=#hops+#set-ops+plurality），并分为简单、中等和困难三类。数据集分为开发集和测试集，开发集包含40个文档和4,418个问答对，测试集包含130个文档和15,721个问答对（未发布）。数据集旨在评估长上下文LLM的性能，并支持模型开发和验证。
The KG-MuLQA-D dataset is a collection of 20,139 QA pairs derived from 170 SEC credit agreements (2013–2022) and categorized by five complexity levels. Each QA pair is tagged with a composite complexity level (L=#hops+#set-ops+plurality), split into Easy, Medium, and Hard. The dataset is divided into development and test sets, with the development set containing 40 documents and 4,418 QA pairs, and the test set containing 130 documents and 15,721 QA pairs (not released). The dataset is designed for evaluating long-context LLMs and supports model development and validation.
提供机构：
gtfintechlab
