Multimodal Multi-Agent Empowered Legal Judgment Prediction

TL;DR

This work tackles Legal Judgment Prediction (LJP) under multimodal evidence and procedural realism by introducing JurisMMA, a multi-agent framework that simulates court proceedings across six stages and leverages a knowledge-retrieval module. It also presents JurisMM, a large-scale dataset with over 100k up-to-date Chinese criminal judgments and 83 multimodal video-text samples for comprehensive evaluation. Empirical results on JurisMM and LawBench show JurisMMA outperforms strong baselines, with ablations confirming the essential roles of the knowledge base and multi-agent collaboration. The approach demonstrates improved reasoning over complex legal tasks and lays a practical foundation for future multimodal, legally grounded AI systems in real-world judicial contexts.

Abstract

Legal Judgment Prediction (LJP) aims to predict the outcomes of legal cases based on factual descriptions, serving as a fundamental task to advance the development of legal systems. Traditional methods often rely on statistical analyses or role-based simulations but face challenges with multiple allegations, diverse evidence, and lack adaptability. In this paper, we introduce JurisMMA, a novel framework for LJP that effectively decomposes trial tasks, standardizes processes, and organizes them into distinct stages. Furthermore, we build JurisMM, a large dataset with over 100,000 recent Chinese judicial records, including both text and multimodal video-text data, enabling comprehensive evaluation. Experiments on JurisMM and the benchmark LawBench validate our framework's effectiveness. These results indicate that our framework is effective not only for LJP but also for a broader range of legal applications, offering new perspectives for the development of future legal methods and datasets.

Figures (2)

• Figure 1: Overview of our framework JurisMMA.
• Figure 2: Overall performance on LawBench.