DTKG: Dual-Track Knowledge Graph-Verified Reasoning Framework for Multi-Hop QA

TL;DR

The paper tackles the challenge of multi-hop QA in retrieval-augmented settings by addressing the heterogeneous needs of parallel fact-verification and chained reasoning. It introduces DTKG, a dual-track framework with a few-shot prompting-based task classifier that dynamically routes questions to either a parallel fact-verification branch or a chained reasoning branch, each equipped with KG-grounded, task-specific processing. A task-aware denoising component further cleans KG evidence by filtering irrelevant relations through a two-stage scoring mechanism and a keyword-based invalid-relation library. Empirical results on HotpotQA, Mintaka, CWQ, and QALD10 demonstrate superior accuracy and semantic alignment, validating the approach’s adaptability and efficiency across diverse multi-hop tasks. This work significantly reduces the strategy-task mismatch in LLM-KG systems and enhances practical QA performance in retrieval-augmented environments.

Abstract

Multi-hop reasoning for question answering (QA) plays a critical role in retrieval-augmented generation (RAG) for modern large language models (LLMs). The accurate answer can be obtained through retrieving relational structure of entities from knowledge graph (KG). Regarding the inherent relation-dependency and reasoning pattern, multi-hop reasoning can be in general classified into two categories: i) parallel fact-verification multi-hop reasoning question, i.e., requiring simultaneous verifications of multiple independent sub-questions; and ii) chained multi-hop reasoning questions, i.e., demanding sequential multi-step inference with intermediate conclusions serving as essential premises for subsequent reasoning. Currently, the multi-hop reasoning approaches singly employ one of two techniques: LLM response-based fact verification and KG path-based chain construction. Nevertheless, the former excels at parallel fact-verification but underperforms on chained reasoning tasks, while the latter demonstrates proficiency in chained multi-hop reasoning but suffers from redundant path retrieval when handling parallel fact-verification reasoning. These limitations deteriorate the efficiency and accuracy for multi-hop QA tasks. To address this challenge, we propose a novel dual-track KG verification and reasoning framework DTKG, which is inspired by the Dual Process Theory in cognitive science. Specifically, DTKG comprises two main stages: the Classification Stage and the Branch Processing Stage.

Figures (5)

• Figure 1: Problem statement and our solution: (a) Fact verification based on post-processing (e.g., KGR); (b) Knowledge graph verification based on multi-hop reasoning chains (e.g., TOG); (c) Dual-Track Knowledge Graph-Verified Reasoning Framework for Multi-Hop QA (DTKG).
• Figure 2: Overview of the DTKG framework. An LLM-powered classifier first categorizes the input query into either a "multi-hop reasoning chain" or "parallel fact-checking" type, then this query is routed to the corresponding specialized processing branch, wherein a knowledge graph is leveraged for tailored reasoning and verification to produce proper answer.
• Figure 3: Error analysis on 100 cases from the four datasets.
• Figure 4: Distribution of question type across datasets.
• Figure 5: Hyperparameter N for Top-N Candidate Selection