知识卡片

普通读书笔记卡 · 6500.h · 生成模型:GPT-6

评估数据副本仍要固定版本

理解等级:中级

使用本地副本可降低外部源临时变化影响,但从仓库 main 分支下载仍可能随提交变化。复现应保存数据版本或内容哈希、样本 ID 和抽样方式,不能仅记录“来自同一仓库”或文件名相同。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      The MATH-500 dataset split was originally proposed in the PRM800K repository (https://github.com/openai/prm800k/tree/main?tab=readme-ov-file#math-splits) and is also available on the Hugging Face Hub (https://huggingface.co/datasets/HuggingFaceH4/MATH-500). However, in this book, we load a copy from the code repository to ensure reproducibility in case the external sources change.

    参考来源

    - 位置:《Build a Reasoning Model (From Scratch)》第03章《评估推理模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Build a Reasoning Model (From Scratch)/_epub-src/OEBPS/Text/chapter-3.html) - 结论依据:摘录中关于评估数据副本仍要固定版本的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:The MATH-500 dataset split was originally proposed in the PRM800K repository (https://github.com/openai/prm800k/tree/main?tab=readme-ov-file#math-splits) and is also available on the Hugging Face Hub (https://huggingface.co/datasets/HuggingFaceH4/MATH-500). However, in this book, we load a copy from the code repository to ensure reproducibility in case the external sources change.