colinglab/CLASS_IT
收藏Hugging Face2025-11-07 更新2025-11-15 收录
下载链接：
https://hf-mirror.com/datasets/colinglab/CLASS_IT
下载链接
链接失效反馈官方服务：
资源简介：
CLASS-IT数据集是一个用于BabyLM-scale模型指令调优的小规模数据集，旨在研究小规模语言模型如何从交互驱动和课程化的指令调优中受益。该数据集包含两个互补的组件——Simple Wikipedia（指导性）和Switchboard（对话性），使得可以在结构化的问答式监督和自然的对话式适应之间进行比较。
The CLASS-IT dataset provides instruction-tuning material for BabyLM-scale models, designed to investigate how small-scale language models benefit from interaction-driven and curriculum-based instruction tuning. It contains two complementary components — Simple Wikipedia (instructional) and Switchboard (conversational) — enabling comparison between structured, question–answer style supervision and natural dialogue-based adaptation.
提供机构：
colinglab
