2026-10-15T04:05:11Z https://www.scidb.cn/oai

10.57760/sciencedb.37740 2026-09-14T15:09:16Z

2026-09-14 Tibetan-Chinese Bilingual Legal Consultation and Multi-turn Dialogue Dataset for LLM Fine-tuning doi:10.57760/sciencedb.37740 en This dataset aims to provide high-quality and structured corpus support for fine-tuning the Large Language Model (LLM) in the vertical legal domain of specific ethnic minority areas. The data content is based on the actual scenes of grass-roots governance and agricultural and pastoral areas/urban life in the Xizang Autonomous Region, and covers 10 common legal fields in depth, including tourism service disputes, grassland circulation, cordyceps harvesting, house leasing, labor disputes, marriage and family. The dataset consists of three core subsets: stage one data consisting of a single round of basic legal Q&A, high-quality advanced stage two data that has been deeply cleaned, and stage three data consisting of multiple rounds of bilingual conversations with complex character emotions (such as anxiety and anger) and specific practical difficulties. All dialogues are based on China's current national laws and local regulations in Xizang, and are constructed using large model oriented synthesis technology, with a strict bilingual bilingual format between China and Tibet. This dataset effectively fills the gap in Tibetan legal AI training corpus, aiming to enhance the logical reasoning, empathy soothing, and practical guidance abilities of large models in complex grassroots legal consultation scenarios. All dispute scenarios in this dataset are synthesized by AI logic and do not involve any real cases or personal privacy. They are purely for academic research and model development.  Tibetan-Chinese Bilingual; Legal Consultation; Large Language Model (LLM); Instruction Fine-tuning; Synthetic Data; Grassroots Governance; Agricultural and Pastoral Areas of Tibet Zhang Jingxuan Hui Cao Dou Liyang Xie Xiaowei PUBLIC https://creativecommons.org/licenses/by-nc/4.0/ dataset Science Data Bank