Improving Multilingual Retrieval-Augmented Language Models through Dialectic Reasoning Argumentations

2025-04-07

AI Summary

This work addresses knowledge conflicts arising from cross-lingual retrieval in multilingual RAG. We propose DRAG, the first RAG framework integrating dialectical reasoning. DRAG employs a modular argumentation architecture to achieve multilingual semantic alignment, fine-grained conflict detection and resolution, and lightweight contextual argument generation. Instead of simple retrieval fusion, it delivers structured argumentative explanations for critical knowledge integration. Compared to conventional approaches, DRAG significantly improves factual accuracy and robustness in multilingual RAG, enhances resilience against noisy and contradictory knowledge, and supports low-overhead deployment. Crucially, it provides a scalable, analytical reasoning paradigm tailored for resource-constrained language models.

Technology Category

• Knowledge Representation and Reasoning: Argumentation
• Natural Language Processing: Discourse, Pragmatics & Argument Mining
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Retrieval-augmented generation (RAG) is key to enhancing large language models (LLMs) to systematically access richer factual knowledge. Yet, using RAG brings intrinsic challenges, as LLMs must deal with potentially conflicting knowledge, especially in multilingual retrieval, where the heterogeneity of knowledge retrieved may deliver different outlooks. To make RAG more analytical, critical and grounded, we introduce Dialectic-RAG (DRAG), a modular approach guided by Argumentative Explanations, i.e., structured reasoning process that systematically evaluates retrieved information by comparing, contrasting, and resolving conflicting perspectives. Given a query and a set of multilingual related documents, DRAG selects and exemplifies relevant knowledge for delivering dialectic explanations that, by critically weighing opposing arguments and filtering extraneous content, clearly determine the final response. Through a series of in-depth experiments, we show the impact of our framework both as an in-context learning strategy and for constructing demonstrations to instruct smaller models. The final results demonstrate that DRAG significantly improves RAG approaches, requiring low-impact computational effort and providing robustness to knowledge perturbations.

Problem

• Enhancing multilingual retrieval-augmented models with dialectic reasoning
• Resolving conflicting knowledge in multilingual document retrieval
• Improving robustness and efficiency in retrieval-augmented generation

Innovation

• Modular DRAG approach with Argumentative Explanations
• Compares and contrasts multilingual retrieved information
• Filters extraneous content for robust responses