原书出处
Build a Reasoning Model (From Scratch)
We use a non-overlapping training subset derived from the original MATH dataset, which means that all MATH-500 evaluation problems are explicitly excluded from the training data. This prevents data leakage and ensures a clean separation between training and evaluation, as illustrated in figure 6.11.
原书定位详情
来源:a19c7933-4eda-4176-aebd-30fe7e1ad292
正文位置:extracted/Build a Reasoning Model (From Scratch)/_epub-src/OEBPS/Text/chapter-6.html
EPUB SHA-256:52d4d07986020be70f5aac21762dfcf60fcba08db0ba29d4f63957caf12aa377