Efficient Knowledge Infusion via KG-LLM Alignment
This work addresses domain-specific knowledge infusion for LLMs, particularly in biomedical QA, representing an incremental improvement over existing knowledge graph-retrieval augmented methods.
The paper tackles the problem of domain-specific knowledge scarcity in large language models by addressing knowledge mismatch and poor information compliance between knowledge graphs and LLMs. The proposed approach constructs domain-specific knowledge graphs using an LLM and implements a three-stage alignment strategy, achieving superior performance on two biomedical QA datasets compared to existing baselines.
To tackle the problem of domain-specific knowledge scarcity within large language models (LLMs), knowledge graph-retrievalaugmented method has been proven to be an effective and efficient technique for knowledge infusion. However, existing approaches face two primary challenges: knowledge mismatch between public available knowledge graphs and the specific domain of the task at hand, and poor information compliance of LLMs with knowledge graphs. In this paper, we leverage a small set of labeled samples and a large-scale corpus to efficiently construct domain-specific knowledge graphs by an LLM, addressing the issue of knowledge mismatch. Additionally, we propose a three-stage KG-LLM alignment strategyto enhance the LLM's capability to utilize information from knowledge graphs. We conduct experiments with a limited-sample setting on two biomedical question-answering datasets, and the results demonstrate that our approach outperforms existing baselines.