@article{cai2025lexrelbenchmarkinglegalrelation,
title = {LexRel: Benchmarking Legal Relation Extraction for Chinese Civil Cases},
author = {Yida Cai and Ranjuexiao Hu and Huiyuan Xie and Chenyang Li and Yun Liu and Yuxiao Ye and Zhenghao Liu and Weixing Shen and Zhiyuan Liu},
journal = {arXiv preprint arXiv:2512.12643},
year = {2025}}
@misc{LAWGPT-zh,
author = {Hongcheng Liu and Yusheng Liao and Yutong Meng and Yuhao Wang},
title = {XieZhi: Chinese Law Large Language Model},
year = {2023},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/LiuHC0428/LAW_GPT}}}
LeCaRDv2's tar.gz archive is verified and automatically extracted to
knowledge/legal_cases/LeCaRDv2/candidate_55192/ by the current LexVerse downloader.
Prebuilt indexes
We provide prebuilt hybrid indexes, one per collection, combining BM25 and
vector retrieval.
lexverse resources download indexes
For LeCaRDv2, BM25 indexes only qw text; vectors cover qw + fact + reason.
The License column lists licenses stated by the upstream sources. Refer to each original source for the applicable scope and access conditions. A - indicates that no license or note is listed. Organization by LexVerse does not change the licenses of the original materials.