The paper introduces the KG-RAG (Knowledge Graph-Retrieval Augmented Generation) pipeline, a novel framework designed to enhance the knowledge capabilities of Large Language Models (LLMs) by integrating structured Knowledge Graphs (KGs) with LLM functionalities. This approach aims to reduce the reliance on LLMs' latent knowledge and address issues such as information hallucinations, catastrophic forgetting, and limitations in processing long contexts. The KG-RAG pipeline constructs a KG from unstructured text, performs information retrieval over the graph for Knowledge Graph Question Answering (KGQA), and leverages a Chain of Explorations (CoE) algorithm to sequentially explore nodes and relationships within the KG. Preliminary experiments on the ComplexWebQuestions dataset show promising results, reducing hallucinated content and suggesting a promising path for developing intelligent systems adept at handling knowledge-intensive tasks. The paper also discusses related work, methodology, experimental setup, and limitations, highlighting the potential for future research in KG construction and knowledge representation.The paper introduces the KG-RAG (Knowledge Graph-Retrieval Augmented Generation) pipeline, a novel framework designed to enhance the knowledge capabilities of Large Language Models (LLMs) by integrating structured Knowledge Graphs (KGs) with LLM functionalities. This approach aims to reduce the reliance on LLMs' latent knowledge and address issues such as information hallucinations, catastrophic forgetting, and limitations in processing long contexts. The KG-RAG pipeline constructs a KG from unstructured text, performs information retrieval over the graph for Knowledge Graph Question Answering (KGQA), and leverages a Chain of Explorations (CoE) algorithm to sequentially explore nodes and relationships within the KG. Preliminary experiments on the ComplexWebQuestions dataset show promising results, reducing hallucinated content and suggesting a promising path for developing intelligent systems adept at handling knowledge-intensive tasks. The paper also discusses related work, methodology, experimental setup, and limitations, highlighting the potential for future research in KG construction and knowledge representation.