TeleQnA Subset
收藏arXiv2025-09-30 收录
下载链接：
https://zindi.africa/competitions/specializing-large-language-models-for-telecom-networks-by-itu-ai-ml-in-5g-challenge
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是从TeleQnA数据集中筛选出的一个子集，专注于两个类别：标准规范和标准概述，共包含366个问题供公众测试使用。此外，该数据集被用于评估包括Phi-2和Falcon-7B在内的多种模型在检索增强生成（RAG）框架下的性能表现。数据集的规模如下：训练集包含1461个问题，公共测试集包含366个问题，私有测试集包含2000个问题。该任务针对的是电信领域的问答。
This dataset is a curated subset derived from the TeleQnA dataset, focusing on two categories: standard specifications and standard overviews, with a total of 366 questions intended for public testing. Additionally, this dataset is employed to evaluate the performance of multiple models including Phi-2 and Falcon-7B under the Retrieval-Augmented Generation (RAG) framework. The dataset is structured as follows: the training set contains 1,461 questions, the public test set contains 366 questions, and the private test set contains 2,000 questions. This task targets question answering within the telecommunications domain.
提供机构：
Zindi
