遇见数据集

alibaba-pai/OmniThought

收藏
Hugging Face2025-08-23 更新2025-05-31 收录
官方服务:

资源简介:

OmniThought是一个包含200万个链式思维过程的大规模数据集,为大型推理模型的训练提供了多样化的推理任务和高品质的推理路径。数据集中的每个链式思维过程都标注有推理冗余度(RV)和认知难度(CD),有助于模型更好地理解和执行推理任务。

OmniThought is a large-scale dataset containing 2 million chain-of-thought processes, providing diverse reasoning tasks and high-quality reasoning paths for the training of large reasoning models. Each chain-of-thought process in the dataset is annotated with Reasoning Verbosity (RV) and Cognitive Difficulty (CD), which assist models in better understanding and performing reasoning tasks.

提供机构:
alibaba-pai
搜集汇总
数据集介绍
alibaba-pai/OmniThought 数据集图片
以上内容由遇见数据集搜集并总结生成
二维码
社区交流群
二维码
科研交流群
商业服务