LexGenius

LexGenius: An Expert-Level Benchmark for Large Language Models in Legal General Intelligence

<div align="center">

</div>

---

Overview

<div align="center"> <img src="image/compare.png" width="60%"/> </div>

LexGenius addresses a fundamental challenge in applying large language models (LLMs) to the legal domain—the absence of a professional, systematic, and trustworthy evaluation framework for legal intelligence. LexGenius is an expert-level benchmarking suite designed for Chinese legal scenarios, assessing LLMs across diverse legal tasks and capability dimensions to measure their understanding, reasoning, and normative application skills. Through rigorous dataset construction, realistic legal problem design, and human–LLM collaborative validation, LexGenius significantly enhances the objectivity, discriminability, and reliability of legal intelligence evaluation, offering actionable insights for model development, optimization, and deployment—without requiring users to build separate assessment systems.

<div align="center"> <img src="image/Framwork.png" width="90%"/> </div>

By integrating the three-level structure of seven legal dimensions, eleven tasks, and twenty abilities, LexGenius provides a structured legal intelligence evaluation framework that supports systematic capability assessment and cross-model comparative analysis across diverse large-scale LLMs.

Experimental Results

Comparison of the 12 SOTA LLMs with human experts on 7 core dimensions of legal intelligence:

<div align="center"> <img src="image/7 core dimensions.png" width="50%"/> </div>

Performance of 12 LLMs and human experts on 11 legal tasks, showing a significant gap between LLMs and humans:

<div align="center"> <img src="image/11 legal tasks.png" width="100%"/> </div>

Average ranking and average score ranking of the 12 SOTA LLMs in the 20 legal intelligence abilities.:

<div align="center"> <img src="image/20 legal intelligence abilities.png" width="80%"/> </div>

BibTex

If you find this work is helpful for your research, please cite:

@misc{liu2025lexgeniusbenchmark,
      title={LexGenius: An Expert-Level Benchmark for Large Language Models in Chinese Legal General Intelligence},
      author={Wenjin Liu and Haoran Luo and Xin Feng and Xiang Ji and Lijuan Zhou and Rui Mao and Jiapu Wang and Shirui Pan and Erik Cambria},
      year={2025},
      eprint={2512.04578},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2512.04578}
}