Legal Mathematical Reasoning with LLMs: Procedural Alignment through Two-Stage Reinforcement Learning

• URL: http://arxiv.org/abs/2504.02590v2
• Date: Sat, 07 Jun 2025 07:22:31 GMT
• Title: Legal Mathematical Reasoning with LLMs: Procedural Alignment through Two-Stage Reinforcement Learning
• Authors: Kepu Zhang, Guofu Xie, Weijie Yu, Mingyue Xu, Xu Tang, Yaxin Li, Jun Xu,
• Abstract summary: Legal mathematical reasoning is essential for applying large language models (LLMs) in high-stakes legal contexts.<n>We present LexNum, the first Chinese legal mathematical reasoning benchmark.<n>We also propose LexPam, a two-stage reinforcement learning framework for efficient legal reasoning training.
• Score: 12.90492832643565
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Legal mathematical reasoning is essential for applying large language models (LLMs) in high-stakes legal contexts, where outputs must be both mathematically accurate and procedurally compliant. However, existing legal LLMs lack structured numerical reasoning, and open-domain models, though capable of calculations, often overlook mandatory legal steps. To address this, we present LexNum, the first Chinese legal mathematical reasoning benchmark, covering three representative scenarios where each instance reflects legally grounded procedural flows. We further propose LexPam, a two-stage reinforcement learning framework for efficient legal reasoning training. Leveraging curriculum learning, we use a stronger teacher model to partition data into basic and challenging subsets. A lightweight 1.5B student model is then fine-tuned with Group Relative Policy Optimization, which avoids costly value networks and enables stable training from sparse, end-of-sequence rewards. The first stage improves accuracy and format; the second introduces a novel reward to guide procedural alignment via task-specific legal elements. Experiments show that existing models perform poorly on LexNum, while LexPam enhances both mathematical accuracy and legal coherence, and generalizes effectively across tasks and domains.

Related papers

• LEXam: Benchmarking Legal Reasoning on 340 Law Exams [61.344330783528015]
  LEXam is a novel benchmark derived from 340 law exams spanning 116 law school courses across a range of subjects and degree levels.<n>The dataset comprises 4,886 law exam questions in English and German, including 2,841 long-form, open-ended questions and 2,045 multiple-choice questions.
  arXiv  Detail & Related papers  (2025-05-19T08:48:12Z)
• SynLexLM: Scaling Legal LLMs with Synthetic Data and Curriculum Learning [0.0]
  We introduce SynLexLM, a novel approach to efficiently pre-train a legal LLM.<n>Our method employs curriculum learning, progressing from simple to complex legal texts and queries, combined with synthetic data augmentation.<n>Preliminary work involves generating synthetic QA pairs reflecting legal reasoning.
  arXiv  Detail & Related papers  (2025-04-26T01:42:22Z)
• Engineering the Law-Machine Learning Translation Problem: Developing Legally Aligned Models [0.0]
  We introduce a five-stage interdisciplinary framework that integrates legal and ML-technical analysis during machine learning model development.<n>This framework facilitates designing ML models in a legally aligned way and identifying high-performing models that are legally justifiable.
  arXiv  Detail & Related papers  (2025-04-23T13:41:17Z)
• Aplicação de Large Language Models na Análise e Síntese de Documentos Jurídicos: Uma Revisão de Literatura [0.0]
  Large Language Models (LLMs) have been increasingly used to optimize the analysis and synthesis of legal documents.<n>This study aims to conduct a systematic literature review to identify the state of the art in prompt engineering applied to LLMs in the legal context.
  arXiv  Detail & Related papers  (2025-04-01T12:34:00Z)
• LexPro-1.0 Technical Report [19.83460019437367]
  We introduce our first-generation reasoning model, LexPro-1.0, a large language model designed for the highly specialized Chinese legal domain.<n>To address this, we first compile millions of legal documents covering over 20 types of crimes from 31 provinces in China for model training.<n>The model further undergoes large-scale reinforcement learning without additional supervision, emphasizing the enhancement of its reasoning capabilities and explainability.
  arXiv  Detail & Related papers  (2025-03-10T05:54:23Z)
• Artificial Intelligence and Legal Analysis: Implications for Legal Education and the Profession [0.0]
  This article reports the results of a study examining the ability of legal and nonlegal Large Language Models to perform legal analysis.<n>The results show that LLMs can conduct basic IRAC analysis, but are limited by brief responses lacking detail, an inability to commit to answers, false confidence, and hallucinations.
  arXiv  Detail & Related papers  (2025-02-04T19:50:48Z)
• LegalAgentBench: Evaluating LLM Agents in Legal Domain [53.70993264644004]
  LegalAgentBench is a benchmark specifically designed to evaluate LLM Agents in the Chinese legal domain.<n>LegalAgentBench includes 17 corpora from real-world legal scenarios and provides 37 tools for interacting with external knowledge.
  arXiv  Detail & Related papers  (2024-12-23T04:02:46Z)
• RuleArena: A Benchmark for Rule-Guided Reasoning with LLMs in Real-World Scenarios [58.90106984375913]
  RuleArena is a novel and challenging benchmark designed to evaluate the ability of large language models (LLMs) to follow complex, real-world rules in reasoning.<n> Covering three practical domains -- airline baggage fees, NBA transactions, and tax regulations -- RuleArena assesses LLMs' proficiency in handling intricate natural language instructions.
  arXiv  Detail & Related papers  (2024-12-12T06:08:46Z)
• Can Large Language Models Grasp Legal Theories? Enhance Legal Reasoning with Insights from Multi-Agent Collaboration [27.047809869136458]
  Large Language Models (LLMs) could struggle to fully understand legal theories and perform legal reasoning tasks. We introduce a challenging task (confusing charge prediction) to better evaluate LLMs' understanding of legal theories and reasoning capabilities. We also propose a novel framework: Multi-Agent framework for improving complex Legal Reasoning capability.
  arXiv  Detail & Related papers  (2024-10-03T14:15:00Z)
• Revisiting the Superficial Alignment Hypothesis [0.9831489366502302]
  The Superficial Alignment Hypothesis posits that almost all of a language model's abilities and knowledge are learned during pre-training. We re-examine these claims by studying the scaling behavior of post-training with increasing finetuning examples.
  arXiv  Detail & Related papers  (2024-09-27T22:14:10Z)
• InternLM-Law: An Open Source Chinese Legal Large Language Model [72.2589401309848]
  InternLM-Law is a specialized LLM tailored for addressing diverse legal queries related to Chinese laws. We meticulously construct a dataset in the Chinese legal domain, encompassing over 1 million queries. InternLM-Law achieves the highest average performance on LawBench, outperforming state-of-the-art models, including GPT-4, on 13 out of 20 subtasks.
  arXiv  Detail & Related papers  (2024-06-21T06:19:03Z)
• Can LLMs Reason with Rules? Logic Scaffolding for Stress-Testing and Improving LLMs [87.34281749422756]
  Large language models (LLMs) have achieved impressive human-like performance across various reasoning tasks. However, their mastery of underlying inferential rules still falls short of human capabilities. We propose a logic scaffolding inferential rule generation framework, to construct an inferential rule base, ULogic.
  arXiv  Detail & Related papers  (2024-02-18T03:38:51Z)
• InternLM-Math: Open Math Large Language Models Toward Verifiable Reasoning [98.53491178426492]
  We open-source our math reasoning LLMs InternLM-Math which is continue pre-trained from InternLM2. We unify chain-of-thought reasoning, reward modeling, formal reasoning, data augmentation, and code interpreter in a unified seq2seq format. Our pre-trained model achieves 30.3 on the MiniF2F test set without fine-tuning.
  arXiv  Detail & Related papers  (2024-02-09T11:22:08Z)
• BLT: Can Large Language Models Handle Basic Legal Text? [44.89873147675516]
  GPT-4 and Claude perform poorly on basic legal text handling. Poor performance on benchmark casts into doubt their reliability as-is for legal practice. Fine-tuning on training set brings even a small model to near-perfect performance.
  arXiv  Detail & Related papers  (2023-11-16T09:09:22Z)
• A Comprehensive Evaluation of Large Language Models on Legal Judgment Prediction [60.70089334782383]
  Large language models (LLMs) have demonstrated great potential for domain-specific applications. Recent disputes over GPT-4's law evaluation raise questions concerning their performance in real-world legal tasks. We design practical baseline solutions based on LLMs and test on the task of legal judgment prediction.
  arXiv  Detail & Related papers  (2023-10-18T07:38:04Z)
• Precedent-Enhanced Legal Judgment Prediction with LLM and Domain-Model Collaboration [52.57055162778548]
  Legal Judgment Prediction (LJP) has become an increasingly crucial task in Legal AI. Precedents are the previous legal cases with similar facts, which are the basis for the judgment of the subsequent case in national legal systems. Recent advances in deep learning have enabled a variety of techniques to be used to solve the LJP task.
  arXiv  Detail & Related papers  (2023-10-13T16:47:20Z)
• LAiW: A Chinese Legal Large Language Models Benchmark [17.66376880475554]
  General and legal domain LLMs have demonstrated strong performance in various tasks of LegalAI. We are the first to build the Chinese legal LLMs benchmark LAiW, based on the logic of legal practice.
  arXiv  Detail & Related papers  (2023-10-09T11:19:55Z)
• Towards LogiGLUE: A Brief Survey and A Benchmark for Analyzing Logical Reasoning Capabilities of Language Models [56.34029644009297]
  Large language models (LLMs) have demonstrated the ability to overcome various limitations of formal Knowledge Representation (KR) systems. LLMs excel most in abductive reasoning, followed by deductive reasoning, while they are least effective at inductive reasoning. We study single-task training, multi-task training, and "chain-of-thought" knowledge distillation fine-tuning technique to assess the performance of model.
  arXiv  Detail & Related papers  (2023-10-02T01:00:50Z)
• LawBench: Benchmarking Legal Knowledge of Large Language Models [35.2812008533622]
  Large language models (LLMs) have demonstrated strong capabilities in various aspects. It is unclear how much legal knowledge they possess and whether they can reliably perform legal-related tasks. LawBench has been meticulously crafted to have precise assessment of the LLMs' legal capabilities from three cognitive levels.
  arXiv  Detail & Related papers  (2023-09-28T09:35:59Z)
• Large Language Models as Tax Attorneys: A Case Study in Legal Capabilities Emergence [5.07013500385659]
  This paper explores Large Language Models' (LLMs) capabilities in applying tax law. Our experiments demonstrate emerging legal understanding capabilities, with improved performance in each subsequent OpenAI model release. Findings indicate that LLMs, particularly when combined with prompting enhancements and the correct legal texts, can perform at high levels of accuracy but not yet at expert tax lawyer levels.
  arXiv  Detail & Related papers  (2023-06-12T12:40:48Z)
• SatLM: Satisfiability-Aided Language Models Using Declarative Prompting [68.40726892904286]
  We propose a new satisfiability-aided language modeling (SatLM) approach for improving the reasoning capabilities of large language models (LLMs) We use an LLM to generate a declarative task specification rather than an imperative program and leverage an off-the-shelf automated theorem prover to derive the final answer. We evaluate SATLM on 8 different datasets and show that it consistently outperforms program-aided LMs in the imperative paradigm.
  arXiv  Detail & Related papers  (2023-05-16T17:55:51Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.