Debate-Feedback: A Multi-Agent Framework for Efficient Legal Judgment Prediction

2025-04-07

AI Summary

To address the challenges of heavy reliance on large-scale labeled data, poor interpretability, and insufficient robustness in legal judgment prediction, this paper proposes the Debate-Feedback multi-agent framework. Departing from conventional fine-tuning paradigms, it pioneers the integration of authentic courtroom debate mechanisms into LegalAI: large language model (LLM) agents engage in dynamic adversarial debate, cross-examination, and reliability-aware feedback calibration to enable reasoning-driven prediction under zero- or few-shot settings. Crucially, the framework eliminates the need for historical case fine-tuning, substantially reducing data dependency; its multi-step collaborative reasoning enhances decision transparency and auditability. Evaluated across multiple legal judgment benchmarks, our approach outperforms both general-purpose and domain-specific models, achieving up to 37% improvement in reasoning efficiency. Results demonstrate the effectiveness, parameter efficiency, and cross-task generalizability of debate-based dynamic reasoning in LegalAI.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Safety and Robustness
• Multiagent Systems: Adversarial Agents

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

The use of AI in legal analysis and prediction (LegalAI) has gained widespread attention, with past research focusing on retrieval-based methods and fine-tuning large models. However, these approaches often require large datasets and underutilize the capabilities of modern large language models (LLMs). In this paper, inspired by the debate phase of real courtroom trials, we propose a novel legal judgment prediction model based on the Debate-Feedback architecture, which integrates LLM multi-agent debate and reliability evaluation models. Unlike traditional methods, our model achieves significant improvements in efficiency by minimizing the need for large historical datasets, thus offering a lightweight yet robust solution. Comparative experiments show that it outperforms several general-purpose and domain-specific legal models, offering a dynamic reasoning process and a promising direction for future LegalAI research.

Problem

• Improving legal judgment prediction using multi-agent debate
• Reducing reliance on large historical datasets for LegalAI
• Enhancing efficiency and robustness in legal analysis models

Innovation

• Debate-Feedback architecture for legal prediction
• LLM multi-agent debate enhances reasoning
• Lightweight solution minimizes historical data need