Embodied AI Papers
Papers / 论文流
A field-indexed reading stream for embodied AI and robotics papers, with compact cards and per-domain pagination.
论文阅读如果只是倒序列表，很快会失去领域结构，也很难判断每篇论文为什么值得看。
PaperItems 区分自动导入字段和人工覆盖字段，前台按 effective domain 分组，卡片同时展示英文标题、中文标题、理由、标签和链接。
论文页可按领域浏览，默认每组 6 篇，展开后按 12 篇分页，适合作为持续阅读入口。
Reading flow
The Papers page is organized around domains instead of one endless reverse-chronological feed. A sticky directory provides a map of the collection, and every domain starts with six papers before expanding into twelve-item pages.
Each card keeps the decision-making information close together: date, original and translated title, technical tags, and direct Paper or Code links.
Data pipeline
Paper records are imported through the Payload local API and deduplicated by stable external identity plus import hash. Public pages read published Payload content only, so seed/import must initialize the database before the site is considered restored.