The 3rd Place Solution of CCIR CUP 2025: A Framework for Retrieval-Augmented Generation in Multi-Turn Legal Conversation

2025-10-17

AI Summary

Legal multi-turn dialogue poses two key challenges for retrieval-augmented generation (RAG): (1) difficulty in dynamically aligning statutory provisions with evolving conversational context, and (2) weak traceability of supporting legal evidence. To address these, this paper proposes a domain-specific RAG framework for legal multi-turn dialogue. The framework integrates three core components: (1) dynamic query rewriting to adapt user intent across turns; (2) context-aware dense paragraph retrieval followed by re-ranking to enhance relevance; and (3) explicit dialogue state tracking to maintain legal grounding and enable interpretable citation. Compared to conventional single-turn RAG, our approach significantly improves both the relevance of retrieved legal provisions and case law, and the factual accuracy of generated responses. Empirical evaluation on the CCIR CUP 2025 Legal Question Answering track demonstrates its effectiveness and practicality in complex legal reasoning tasks, where it ranked third among all participants.

Technology Category

• Natural Language Processing: Question Answering
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Statistical Relational/Logic Learning

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: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Retrieval-Augmented Generation has made significant progress in the field of natural language processing. By combining the advantages of information retrieval and large language models, RAG can generate relevant and contextually appropriate responses based on items retrieved from reliable sources. This technology has demonstrated outstanding performance across multiple domains, but its application in the legal field remains in its exploratory phase. In this paper, we introduce our approach for "Legal Knowledge Retrieval and Generation" in CCIR CUP 2025, which leverages large language models and information retrieval systems to provide responses based on laws in response to user questions.

Problem

• Developing retrieval-augmented generation for multi-turn legal conversations
• Applying RAG technology to legal domain exploration
• Generating law-based responses using retrieval and language models

Innovation

• Retrieval-augmented generation framework for legal conversations
• Combines information retrieval with large language models
• Generates responses based on retrieved legal knowledge