70亿参数模型用强化学习提升法律推理能力，效果媲美更大模型。
Unilaw-R1: A Large Language Model for Legal Reasoning with Reinforcement Learning and Iterative Inference
- 基于17000条法律推理链数据，分两阶段训练增强逻辑与知识。
- 在权威评测中表现超越同规模模型，接近540亿参数大模型水平。
- 适合法律AI研发者、司法科技从业者快速部署高可靠推理系统。
当前聚焦推理的大语言模型在多个领域快速发展，但在复杂法律问题上的能力仍待探索。本文提出面向法律推理的轻量级模型Unilaw-R1，仅含70亿参数，显著降低部署成本，有效应对法律知识不足、推理逻辑不可靠和业务泛化能力弱三大挑战。我们构建了包含1.7万条提炼筛选后的思维链（CoT）样本的高质量数据集Unilaw-R1-Data。在此基础上，采用监督微调（SFT）与强化学习（RL）相结合的两阶段训练策略，大幅提升了复杂法律推理任务表现，并支持可解释的决策过程。为评估法律推理能力，我们设计了专用基准Unilaw-R1-Eval，涵盖单选与多选任务。实验表明，Unilaw-R1在主流基准上表现优异，超越所有同规模模型，性能接近540亿参数的DeepSeek-R1-Distill-Qwen-32B（54.9%）。经领域特定训练后，在LawBench和LexEval上平均超过Qwen-2.5-7B-Instruct（46.6%）6.6个百分点。
原文摘要 · Abstract (English)
Reasoning-focused large language models (LLMs) are rapidly evolving across various domains, yet their capabilities in handling complex legal problems remains underexplored. In this paper, we introduce Unilaw-R1, a large language model tailored for legal reasoning. With a lightweight 7-billion parameter scale, Unilaw-R1 significantly reduces deployment cost while effectively tackling three core challenges in the legal domain: insufficient legal knowledge, unreliable reasoning logic, and weak business generalization. To address these issues, we first construct Unilaw-R1-Data, a high-quality dataset containing 17K distilled and screened chain-of-thought (CoT) samples. Based on this, we adopt a two-stage training strategy combining Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL), which significantly boosts the performance on complex legal reasoning tasks and supports interpretable decision-making in legal AI applications. To assess legal reasoning ability, we also introduce Unilaw-R1-Eval, a dedicated benchmark designed to evaluate models across single- and multi-choice legal tasks. Unilaw-R1 demonstrates strong results on authoritative benchmarks, outperforming all models of similar scale and achieving performance on par with the much larger DeepSeek-R1-Distill-Qwen-32B (54.9%). Following domain-specific training, it also showed significant gains on LawBench and LexEval, exceeding Qwen-2.5-7B-Instruct (46.6%) by an average margin of 6.6%.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。