A Benchmark Dataset of Chinese Development Finance with Climate Relevance and SDG Annotations from 2000–2021

China has become a major provider of climate finance to the Global South, yet systematic and transparent data capturing the full scope of its efforts remains limited. This study presents a refined project-level dataset of Chinese development finance from 2000 to 2021, comprising 20,985 records from the AidData Global Chinese Development Finance Dataset. Using a multi-stage approach, we identify 1,383 climate-related projects, amounting to $421 billion, and assign SDG targets to all entries. Climate relevance and SDG alignment were determined through extensive manual annotation following the Rio Marker and SDG frameworks. To ensure consistency and scalability, we conducted technical validation using pretrained BERT models for climate and SDG classification. The resulting dataset not only enables systematic analysis of China’s evolving international climate finance portfolio and sustainable development commitments, but also serves as a benchmark resource for researchers in climate and sustainable development fields to train or refine advanced language models.

中国已成为全球南方气候融资的主要供给方之一，但能够全面覆盖其相关行动的系统性、透明化数据仍较为匮乏。本研究基于AidData全球中国发展融资数据集（AidData Global Chinese Development Finance Dataset）中的20985条记录，构建了2000年至2021年中国发展融资的精细化项目级数据集。本研究采用多阶段方法，从中识别出1383个气候相关项目，总金额达4210亿美元，并为所有条目标注了可持续发展目标（SDGs, Sustainable Development Goals）对应指标。气候相关性与SDG匹配度的判定，依托里约标记（Rio Marker）与SDG框架，通过大规模人工标注完成。为确保数据集的一致性与可扩展性，本研究采用预训练BERT模型开展气候与SDG分类的技术验证。本研究所生成的数据集，不仅可用于系统性分析中国不断演变的国际气候融资组合与可持续发展承诺，同时也可为气候与可持续发展领域的研究人员训练或优化先进语言模型提供基准参考资源。