Resource-Friendly Dynamic Enhancement Chain for Multi-Hop Question Answering

2025-06-21

AI Summary

To address hallucination, semantic drift, and high computational overhead in knowledge-intensive multi-hop question answering—particularly with lightweight large language models—this paper proposes the Dynamic Augmentation Chain (DAC) framework. DAC introduces three key innovations: (1) a logically coherent subquestion decomposition mechanism that explicitly models multi-hop reasoning paths; (2) a context-aware query rewriting strategy to mitigate semantic drift; and (3) a low-overhead discriminative keyword extraction module that replaces verbose prompts for precise, targeted retrieval. Evaluated on HotpotQA, 2WikiMQA, and MuSiQue, DAC achieves state-of-the-art or competitive performance using only an 8B-parameter model, while reducing average token consumption by 37%. This significantly enhances both inference efficiency and reliability in resource-constrained settings.

Technology Category

• Natural Language Processing: Question Answering
• Data Mining & Knowledge Management: Intelligent Query Processing
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Web Mining and Content Analysis: Community question answering

Abstract

Knowledge-intensive multi-hop question answering (QA) tasks, which require integrating evidence from multiple sources to address complex queries, often necessitate multiple rounds of retrieval and iterative generation by large language models (LLMs). However, incorporating many documents and extended contexts poses challenges -such as hallucinations and semantic drift-for lightweight LLMs with fewer parameters. This work proposes a novel framework called DEC (Dynamic Enhancement Chain). DEC first decomposes complex questions into logically coherent subquestions to form a hallucination-free reasoning chain. It then iteratively refines these subquestions through context-aware rewriting to generate effective query formulations. For retrieval, we introduce a lightweight discriminative keyword extraction module that leverages extracted keywords to achieve targeted, precise document recall with relatively low computational overhead. Extensive experiments on three multi-hop QA datasets demonstrate that DEC performs on par with or surpasses state-of-the-art benchmarks while significantly reducing token consumption. Notably, our approach attains state-of-the-art results on models with 8B parameters, showcasing its effectiveness in various scenarios, particularly in resource-constrained environments.

Problem

• Reduces hallucinations in lightweight LLMs for multi-hop QA
• Improves retrieval precision with low computational overhead
• Optimizes token usage while maintaining high QA performance

Innovation

• Decomposes questions into coherent subquestions
• Refines subquestions via context-aware rewriting
• Uses lightweight keyword extraction for retrieval