LexChain: Modeling Legal Reasoning Chains for Chinese Tort Case Analysis

TL;DR

The paper tackles the challenge of domain-specific legal reasoning in Chinese civil tort law, showing that existing LLMs struggle with structured tort analysis. It introduces LexChain, a three-module framework (legal element identification, liability analysis, judgment summarization) and builds LexChain_eval and LexChain_train benchmarks to evaluate and train models on tort reasoning. Through prompting (Prompt_LC) and reasoning-enhanced training (SFT_LC, DPO_LC), the authors demonstrate consistent gains over zero-shot baselines and reveal improved generalization to related legal tasks like Legal NER and Criminal Damages Calculation. The work highlights the value of explicitly modeling legal reasoning chains to improve interpretability and performance in domain-sensitive legal analysis.

Abstract

Legal reasoning is a fundamental component of legal analysis and decision-making. Existing computational approaches to legal reasoning predominantly rely on generic reasoning frameworks such as syllogism, which do not comprehensively examine the nuanced process of legal reasoning. Moreover, current research has largely focused on criminal cases, with insufficient modeling for civil cases. In this work, we present a novel framework to explicitly model legal reasoning in the analysis of Chinese tort-related civil cases. We first operationalize the legal reasoning process in tort analysis into the three-module LexChain framework, with each module consisting of multiple finer-grained sub-steps. Informed by the LexChain framework, we introduce the task of tort legal reasoning and construct an evaluation benchmark to systematically assess the critical steps within analytical reasoning chains for tort analysis. Leveraging this benchmark, we evaluate existing large language models for their legal reasoning ability in civil tort contexts. Our results indicate that current models still fall short in accurately handling crucial elements of tort legal reasoning. Furthermore, we introduce several baseline approaches that explicitly incorporate LexChain-style reasoning through prompting or post-training. The proposed baselines achieve significant improvements in tort-related legal reasoning and generalize well to related legal analysis tasks, demonstrating the value of explicitly modeling legal reasoning chains to enhance the reasoning capabilities of language models.

Figures (3)

• Figure 1: Illustration of the case information (English translation) in a Chinese tort case and the three-module LexChain reasoning framework.
• Figure 2: An example of the constructed reasoning text for identifying plaintiff(s) and defendant(s) based on legal elements extracted from the original judicial document.
• Figure 3: Evaluation results of GPT-4o and DeepSeek-R1 on the four most common types of tort liability, including liability for motor vehicle traffic accident (TA), liability for injury to providers of labor services (LS), liability for product (P) and liability for medical damage (MD).