arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.04690cs.IR
收藏

多源检索与推理用于法律判决预测

Multi-Source Retrieval and Reasoning for Legal Sentencing Prediction

Tsinghua University(清华大学);Quancheng Laboratory(泉城实验室);MegaTech.AI Inc(MegaTech.AI公司)

Junjie Chen, Haitao Li, Qilei Zhang, Zhenghua Li, Ya Zhang, Quan Zhou, Cheng Luo, Yiqun Liu, Min Zhang, Yueyue Wu, Dongsheng Guo, Qingyao Ai

AI总结:

本文提出MSR²框架,通过多源检索与强化学习提升法律判决预测的准确性和可解释性。

摘要:

法律判决预测(LJP)旨在从案件事实中预测司法结果,通常包括法律条文、指控和判决预测。尽管近期方法在前两个子任务上表现良好,但法律判决预测(LSP)仍然困难,因为其需要精细的客观知识和灵活的主观推理。为了解决这些限制,我们提出了$MSR^2$,一个在大语言模型中集成多源检索和推理的框架,结合强化学习。$MSR^2$使大语言模型能够基于推理需求进行多源检索,并应用过程级奖励来指导中间的主观推理步骤。在两个真实世界数据集上的实验表明,$MSR^2$在LSP中提高了准确性和可解释性,为实用法律AI提供了一步前景。我们的代码可在https://anonymous.4open.science/r/MSR2-FC3B获取。

英文摘要:

Legal judgment prediction (LJP) aims to predict judicial outcomes from case facts and typically includes law article, charge, and sentencing prediction. While recent methods perform well on the first two subtasks, legal sentencing prediction (LSP) remains difficult due to its need for fine-grained objective knowledge and flexible subjective reasoning. To address these limitations, we propose $MSR^2$, a framework that integrates multi-source retrieval and reasoning in LLMs with reinforcement learning. $MSR^2$ enables LLMs to perform multi-source retrieval based on reasoning needs and applies a process-level reward to guide intermediate subjective reasoning steps. Experiments on two real-world datasets show that $MSR^2$ improves both accuracy and interpretability in LSP, providing a promising step toward practical legal AI. Our code is available at https://github.com/cjj826/MSR2.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑