wph6/CS-54k
收藏Hugging Face2025-10-24 更新2025-10-25 收录
下载链接：
https://hf-mirror.com/datasets/wph6/CS-54k
下载链接
链接失效反馈官方服务：
资源简介：
CS-54k是一个大规模的基于论文的数据集，旨在评估和训练用于计算机科学端到端科研工作流程的大型语言模型。它包含54,076个高质量的问答对，这些问答对是从六个顶级计算机科学会议的14,474篇Creative Commons许可的论文中生成的。每个样本都锚定在一篇真实的研究论文上，并分为八个核心研究维度，涵盖了科学探究的完整推理链。
CS-54k is a large-scale, paper-grounded dataset designed to evaluate and train large language models (LLMs) for end-to-end scientific research workflows in computer science. It contains 54,076 high-quality question–answer (Q&A) pairs generated from 14,474 Creative Commons–licensed papers collected from six premier CS conferences, covering the complete reasoning chain of scientific inquiry across eight core research dimensions.
提供机构：
wph6
