BBAI Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/chrisisking/black-box-multi-agent-integation/tree/main/data
下载链接
链接失效反馈官方服务：
资源简介：
该数据集旨在大规模地模拟现实世界的对话代理，涵盖了众包语音输入的广泛领域。具体来说，该数据集包含了37个广泛领域类别的语音输入，其中3700条用于训练，1850条用于测试。规模上，数据集共有5550条语音输入，每个问题对应19个问题-回答对，总计105,450个回答。此外，该数据集的任务是黑箱代理集成（Bbai）。
This dataset aims to simulate real-world conversational agents at scale, covering a wide range of domains with crowdsourced speech inputs. Specifically, the dataset includes speech inputs across 37 broad domain categories, with 3,700 samples allocated for training and 1,850 samples for testing. In terms of overall scale, the dataset contains a total of 5,550 speech inputs, where each question corresponds to 19 question-answer pairs, leading to a grand total of 105,450 responses. Additionally, the core task of this dataset is Black-box Agent Integration (Bbai).
提供机构：
Crowdsourced via Amazon Mechanical Turk
