PL-CA: A Parametric Legal Case Augmentation Framework

2025-09-08

AI Summary

To address two key limitations in judicial-domain RAG—performance degradation caused by long-context inputs and the lack of expert-annotated, multi-task benchmarks—this paper proposes Parameterized Retrieval-Augmented Generation (P-RAG). P-RAG encodes retrieved legal cases as learnable parameter vectors and injects them into the feed-forward networks of large language models via LoRA-based low-rank adaptation, thereby alleviating contextual burden. Concurrently, we introduce a novel multi-task legal evaluation benchmark comprising over 2,000 expert-annotated samples. Experimental results demonstrate that P-RAG maintains or improves downstream task performance while significantly reducing computational overhead. Moreover, it enhances model stability and generalization in complex legal reasoning. This work establishes a lightweight, efficient, and scalable paradigm for retrieval-augmented generation tailored to judicial large language models.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Conventional RAG is considered one of the most effective methods for addressing model knowledge insufficiency and hallucination, particularly in the judicial domain that requires high levels of knowledge rigor, logical consistency, and content integrity. However, the conventional RAG method only injects retrieved documents directly into the model's context, which severely constrains models due to their limited context windows and introduces additional computational overhead through excessively long contexts, thereby disrupting models' attention and degrading performance on downstream tasks. Moreover, many existing benchmarks lack expert annotation and focus solely on individual downstream tasks while real-world legal scenarios consist of multiple mixed legal tasks, indicating conventional benchmarks' inadequacy for reflecting models' true capabilities. To address these limitations, we propose PL-CA, which introduces a parametric RAG (P-RAG) framework to perform data augmentation on corpus knowledge and encode this legal knowledge into parametric vectors, and then integrates this parametric knowledge into the LLM's feed-forward networks (FFN) via LoRA, thereby alleviating models' context pressure. Additionally, we also construct a multi-task legal dataset comprising more than 2000 training and test instances, which are all expert-annotated and manually verified. We conduct our experiments on our dataset, and the experimental results demonstrate that our method reduces the overhead associated with excessively long contexts while maintaining competitive performance on downstream tasks compared to conventional RAG. Our code and dataset are provided in the appendix.

Problem

• Conventional RAG constrains models with limited context windows and computational overhead
• Existing legal benchmarks lack expert annotation and multi-task complexity
• Judicial domain requires high knowledge rigor, consistency, and content integrity

Innovation

• Parametric RAG encodes legal knowledge into vectors
• Integrates knowledge via LoRA into FFN networks
• Constructs expert-annotated multi-task legal dataset