m-a-p/CriticLeanInstruct
收藏Hugging Face2025-07-10 更新2025-08-09 收录
下载链接：
https://hf-mirror.com/datasets/m-a-p/CriticLeanInstruct
下载链接
链接失效反馈官方服务：
资源简介：
CriticLeanInstruct数据集是一系列JSONL文件，包含用于监督微调(SFT)和强化学习(RL)过程的对齐大型语言模型(LLM)的数据。数据集包括纯批评性数据以及与数学和代码相关的混合数据，旨在帮助模型在数学和编程任务上的表现。数据集有不同的版本，如CriticLean_12K、CriticLean_4K、CriticLean_Mix_16K和CriticLean_Mix_48K，每个版本都有不同的样本大小和领域专注度。
The CriticLeanInstruct dataset is a collection of JSONL files containing data for aligning large language models (LLMs) through Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) processes. The dataset includes pure critical data as well as mixed data related to math and code, designed to help models perform better on mathematical and programming tasks. There are different versions of the dataset, such as CriticLean_12K, CriticLean_4K, CriticLean_Mix_16K, and CriticLean_Mix_48K, each with different sample sizes and domain focuses.
提供机构：
m-a-p
