Original Summary

A small, interpretable transformer language model built substantially from first principles, used to investigate how architecture, data, tokenisation, and computational budget affect learned language behaviour.

  • 情报分类：技术学习与提效
  • 分类依据：内容涉及技术、AI、软件工具或工程实践
  • 信息来源：GitHub · AI 新项目
  • 发布时间：2026/10/4 00:25:25