alibaba-pai/OmniThought-0528

OmniThought-0528是一个高质量链式思维（CoT）数据集，旨在通过高质量的推理链蒸馏来增强语言模型的推理能力。该数据集包含365,000个推理链，涵盖数学、编程和科学等多个领域，这些推理链由先进的教师模型生成并经过严格验证。每个推理链都根据其推理详细程度和认知难度进行了评分。

OmniThought-0528 is a high-quality Chain-of-Thought (CoT) dataset designed to enhance the reasoning capabilities of language models through high-quality CoT distillation. The dataset consists of 365,000 reasoning chains across various domains including mathematics, coding, and science, generated and rigorously validated using state-of-the-art teacher models. Each reasoning chain is scored and annotated based on its reasoning verbosity and cognitive difficulty.