Reasoning on Efficient Knowledge Paths:Knowledge Graph Guides Large Language Model for Domain Question Answering

TL;DR

The paper addresses hallucinations and inefficient reasoning in domain-specific QA by grounding LLMs in a knowledge graph through Reasoning on Efficient Knowledge Paths (RoK). RoK expands questions with chain-of-thought, links to a KG to form a subgraph, and uses PageRank-based main path selection plus neighbor branches to ground prompts, enabling high-quality answers with few LLM calls. The approach yields improved or competitive results on GenMedGPT-5k, WebQuestions, and CMCQA compared with vanilla LLMs and retrieval-based baselines, while reducing computational cost. This KG-guided, multi-hop reasoning framework offers a practical path to reliable domain QA with reduced hallucination risk and better interpretability.

Abstract

Large language models (LLMs), such as GPT3.5, GPT4 and LLAMA2 perform surprisingly well and outperform human experts on many tasks. However, in many domain-specific evaluations, these LLMs often suffer from hallucination problems due to insufficient training of relevant corpus. Furthermore, fine-tuning large models may face problems such as the LLMs are not open source or the construction of high-quality domain instruction is difficult. Therefore, structured knowledge databases such as knowledge graph can better provide domain background knowledge for LLMs and make full use of the reasoning and analysis capabilities of LLMs. In some previous works, LLM was called multiple times to determine whether the current triplet was suitable for inclusion in the subgraph when retrieving subgraphs through a question. Especially for the question that require a multi-hop reasoning path, frequent calls to LLM will consume a lot of computing power. Moreover, when choosing the reasoning path, LLM will be called once for each step, and if one of the steps is selected incorrectly, it will lead to the accumulation of errors in the following steps. In this paper, we integrated and optimized a pipeline for selecting reasoning paths from KG based on LLM, which can reduce the dependency on LLM. In addition, we propose a simple and effective subgraph retrieval method based on chain of thought (CoT) and page rank which can returns the paths most likely to contain the answer. We conduct experiments on three datasets: GenMedGPT-5k [14], WebQuestions [2], and CMCQA [21]. Finally, RoK can demonstrate that using fewer LLM calls can achieve the same results as previous SOTAs models.

Figures (7)

• Figure 1: An overview of the architecture of RoK.
• Figure 2: An example workflow of generating reasoning process step by step related to the input problem. Then we leverage LLM to extract key entities from the reasoning process and link to the knowledge graph.
• Figure 3: The generation process of the main reasoning path.
• Figure 4: The process of selecting neighbor triple in main reasoning path. The pruning process is divided into two steps, which are removing repeat triples based on same relation and using LLM to select question-related triples.
• Figure 5: Comparison of RoK and MindMap's final answers.