ResearchDB
收藏资源简介:
该数据集由浙江大学等机构联合构建,收录了2020至2025年间22个顶级会议(如CVPR、NeurIPS等)的10万篇AI领域论文,形成多维科研文献知识图谱。数据包含论文元信息、摘要、方法、数据集等结构化字段,通过BERTopic聚类和LLM语义解析构建,支持主题演化、方法迁移等分析。其核心价值在于揭示AI研究从模型规模驱动向安全可控、多模态认知的范式转变,为科研趋势分析提供实证基础。
This dataset was jointly constructed by Zhejiang University and other institutions. It contains 100,000 AI-related papers from 22 top conferences (e.g., CVPR, NeurIPS) published between 2020 and 2025, forming a multi-dimensional scientific literature knowledge graph. The dataset includes structured fields such as paper metadata, abstracts, methodologies, and associated datasets, and was constructed via BERTopic clustering and LLM semantic parsing, supporting analyses such as topic evolution and methodology transfer. Its core value lies in revealing the paradigm shift of AI research from model-scale-driven development to secure, controllable and multimodal cognition, providing an empirical foundation for scientific research trend analysis.
Profiling_Scientific_Literature 数据集概述
数据集状态
- 数据集尚未发布。
- 完整的代码与数据集将在未来发布。
数据集主题
- 科学文献分析。
备注
- 请持续关注以获取更新。




