pkuHaowei/kumo-easy
收藏Hugging Face2025-04-10 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/pkuHaowei/kumo-easy
下载链接
链接失效反馈官方服务：
资源简介：
KUMO是一个评估大型语言模型复杂推理能力的基准测试数据集，通过程序生成的推理游戏进行。数据集包含五个环境：MedicalEnv、EducationEnv、MusicEnv、FantasyEnv和ChemicalEnv，每个环境有50个示例。还包括100个自动生成的示例领域，涉及计算机科学、生物学、艺术等多个类别，每个领域包含约50个真相和30个动作。
KUMO is a benchmark dataset for evaluating the complex reasoning capabilities of large language models, consisting of procedurally generated reasoning games. The dataset includes five environments: MedicalEnv, EducationEnv, MusicEnv, FantasyEnv, and ChemicalEnv, each with 50 examples. It also features 100 autogenerated exemplar domains across categories such as Computer Science, Biology, and Art, with each domain containing around 50 truths and 30 actions.
提供机构：
pkuHaowei
