codeofpaper/paper-code-graph

codeofpaper: 论文 ↔ 代码图是一个高精度数据集，用于链接研究论文（来自arXiv和OpenReview）与其官方及高置信度的社区GitHub实现，并包含这些仓库的完整分支谱系。该数据集由codeofpaper项目构建和维护，旨在提高研究可重复性。数据集包含论文元数据、GitHub仓库信息、论文与仓库之间的链接边（仅限官方和高置信度社区层级）以及仓库分支谱系，支持研究派生发现（如硬件移植、量化变体等）。数据集采用分层分类（如official和high_confidence_community），排除低置信度边以保持高精度，并包含工程元数据（如框架、许可证和证据字段）以供分类器使用。数据每月更新一次，公开快照有90天延迟以确保稳定性。codeofpaper: paper ↔ code graph is a high-precision dataset linking research papers (from arXiv and OpenReview) to their official and high-confidence community GitHub implementations, plus the full fork lineage for those repositories. Built and maintained by the codeofpaper project, it supports reproducibility research. The dataset includes paper metadata, GitHub repository information, paper-repository links (only official and high-confidence community tiers), and repository fork lineage, enabling research-fork discovery (e.g., hardware ports, quantization variants). It features tiered classification (e.g., official and high_confidence_community), excluding lower-confidence edges for high precision, and contains engineering metadata (e.g., framework, license, and evidence fields) for classifier use. The dataset is refreshed monthly with a 90-day delay in public snapshots for stability.