arXiv 2605.259202026-09-22cs.CLcs.AI

LLM 能时间旅行吗？通过强化学习增强法律智能搜索中的时间一致性

Can LLMs Time Travel? Enhancing Temporal Consistency in Legal Agentic Search through Reinforcement Learning

Department of Computer Science and Engineering, HKUST, Hong Kong SAR, China（香港科技大学计算机科学与工程系）；School of Law, Tsinghua University, Beijing, China（清华大学法学院）；Cheriton School of Computer Science, University of Waterloo, Waterloo, Canada（滑铁卢大学丘成桐计算机科学系）

Wei Fan, Yining Zhou, Mufan Zhang, Yanbing Weng, Yiran HU, Tianshi Zheng, Baixuan Xu, Chunyang Li, Jianhui Yang, Haoran Li, Yangqiu Song

备注：: Accepted by EMNLP NLLP 2026

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

提出 LegalSearch-R1 框架，结合本地 statute RAG 和在线搜索，通过强化学习在跨修订期数据上训练，以解决法律 LLM 的时间偏差和搜索代理缺乏时间约束的问题，在13项法律任务上超越现有方法。

摘要：

虽然增强智能搜索能力的大型语言模型在法律推理方面显示出前景，但它们忽略了一个基本约束：适用法律必须与每个案件的时间背景相匹配，因为法条的事后追溯适用违反了核心法律原则并导致错误结论。我们的观察表明，当前的法律 LLM 存在锚定于其训练截止日期的时间偏差，而搜索代理很少将时间约束纳入查询，并且仅靠网络搜索无法提供法律推理所需的精确法条和先例引用。为应对这些挑战，我们提出 LegalSearch-R1，一个端到端的强化学习框架，它将本地 statute RAG 用于精确条文匹配，与在线网络搜索用于更广泛的法律知识相结合，并在涵盖多个修订期的按时间索引的数据上训练以强制执行时间一致性。在我们涵盖13项法律任务的基准上的大量实验表明，我们的7B参数代理在时间一致性上以12.9%至29.8%的优势超越最先进的深度研究框架和专门的法律 LLM，以57.7%至80.3%的优势超越基线，并展现出强大的域外泛化能力。代码和数据可在 https://github.com/AlexFanw/LegalSearch-R1 获取。

英文摘要：

While large language models (LLMs) augmented with agentic search capabilities show promise for legal reasoning, they overlook a fundamental constraint that applicable law must match the temporal context of each case, as retroactive application of statutes violates core legal principles and leads to erroneous conclusions. Our observations reveal that current legal LLMs suffer from temporal bias anchored to their training cutoff, while search agents rarely incorporate temporal constraints into queries, and that web search alone cannot provide the precise statute and precedent citations that legal reasoning demands. To address these challenges, we propose LegalSearch-R1, an end-to-end reinforcement learning framework that pairs local statute RAG for precise article matching with online web search for broader legal knowledge, trained on temporally-indexed data spanning multiple amendment periods to enforce temporal consistency. Extensive experiments on our benchmark covering 13 legal tasks demonstrate that our 7B-parameter agent outperforms state-of-the-art deep research frameworks and specialized legal LLMs by 12.9% to 29.8%, surpasses baselines by 57.7% to 80.3% on temporal consistency, and exhibits robust out-of-domain generalization. The code and data are available at https://github.com/AlexFanw/LegalSearch-R1.

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。