ThinkChain-20M
收藏资源简介:
SVECTOR-CORPORATION/ThinkChain-20M是一个合成的推理数据集,包含超过2200万个通用推理问题和使用Spec-T1生成的回答。该数据集填补了涵盖社会科学、自然科学、教育、创意写作和一般对话等非代码/数学领域的开源推理数据集的空白。数据集的推理轨迹和答案未经过个别准确性验证,可用于微调更小、更高效的模型,以模仿大型模型如Spec-T1的推理能力。
SVECTOR-CORPORATION/ThinkChain-20M is a synthetic reasoning dataset containing over 22 million general reasoning questions and responses generated by Spec-T1. This dataset fills the gap in open-source reasoning datasets covering non-code/mathematical domains including social sciences, natural sciences, education, creative writing, and general conversation. The reasoning trajectories and answers in this dataset have not been individually verified for accuracy, and can be used to fine-tune smaller, more efficient models to replicate the reasoning capabilities of large models such as Spec-T1.




