marin-community/open-thoughts-4-11k-math-qwen3-32b-agreed-answers

该数据集包含来自OpenThoughts4数学数据集的10,953个样本，其中Qwen3-32B和Qwen3-235B-A22B都产生了完全匹配的有效答案（使用`oxed{}`格式）。数据集包含了Qwen3-32B的推理过程。这些样本是从两个父数据集（每个包含29,963个样本）中筛选出的，属于两个模型都有答案且答案一致的子类别。数据集可用于消融研究（比较答案相同情况下的推理风格）、模型蒸馏（在已验证正确的推理上训练学生模型）和推理分析（研究不同规模模型如何解决相同问题）等用途。

This dataset contains 10,953 samples from the OpenThoughts4 math dataset where both Qwen3-32B and Qwen3-235B-A22B produced valid `oxed{}` answers that match exactly. This dataset contains the Qwen3-32B reasoning traces. These samples are derived from two parent datasets (each with 29,963 samples) and belong to the both have answer and agree subcategory. The dataset can be used for ablation studies (comparing reasoning styles when answers are identical), model distillation (training student models on verified correct reasoning), and reasoning analysis (studying how different model sizes approach the same problems).