LePREC: Reasoning as Classification over Structured Factors for Assessing Relevance of Legal Issues
Summary: arXiv:2604.19464v1 Announce Type: cross
Abstract: More than half of the global population struggles to meet their civil justice needs due to limited legal resources. While Large Language Models (LLMs) have demonstrated impressive reasoning capabilities, significant challenges remain even at the foundational step of legal issue identification. To investigate LLMs’ capabilities in this task, we constructed a dataset from 769 real-world Malaysian Contract Act court cases, using GPT-4o to extract facts and generate candidate legal issues, annotated by senior legal experts, which reveals a critical limitation: while LLMs generate diverse issue candidates, their precision remains inadequate (GPT-4o achieves only 62%).
To address this gap, we propose LePREC (Legal Professional-inspired Reasoning Elicitation and Classification), a neuro-symbolic framework combining neural generation with structured statistical reasoning. LePREC consists of:
- Neuro Component: This component leverages LLMs to transform legal descriptions into question-answer pairs representing diverse analytical factors.
- Symbolic Component: This applies sparse linear models over these discrete features, learning explicit algebraic weights that identify the most informative reasoning factors.
Unlike end-to-end neural approaches, LePREC achieves interpretability through transparent feature weighting while maintaining data efficiency through correlation-based statistical classification. Experiments show a 30-40% improvement over advanced LLM baselines, including GPT-4o and Claude, confirming that correlation-based factor-issue analysis offers a more data-efficient solution for relevance decisions.
The Need for Improved Legal Issue Identification
The complexities of legal systems worldwide leave many individuals without adequate access to justice. Traditional legal resources are often not enough to meet the varying needs of the population. In this context, the role of LLMs has garnered attention, yet their application in legal issue identification has not been thoroughly successful. The findings from our dataset underscore the necessity for more reliable methodologies in this area.
LePREC Framework Overview
LePREC aims to bridge the gap in legal issue identification by utilizing a dual-component system:
- Neuro Component: This part of the framework utilizes advanced LLM capabilities to generate a rich set of question-answer pairs, thereby capturing diverse analytical factors that could influence legal reasoning.
- Symbolic Component: By employing sparse linear models, this component systematically evaluates the discrete features generated by the neuro component. It assigns algebraic weights to these features, enabling the identification of the most significant reasoning factors influencing legal issues.
Results and Implications
Our experiments indicate that LePREC significantly outperforms existing LLM-based methods. The 30-40% improvement in performance signifies a notable advancement in the precision of legal issue identification. This not only enhances the reliability of legal resources but also contributes to a more equitable legal landscape by ensuring that individuals have better access to relevant legal information.
In conclusion, LePREC stands as a pivotal advancement towards improving the efficacy of legal decision-making processes. By merging neural and symbolic approaches, it offers a promising direction for future research and application in legal technology.