The article presents the latest advancements in the field of knowledge-based question answering, including the application and performance of the S3-DST method in open-domain dialogue systems. Additionally, the article introduces the PDFTriage tool, which enhances the question-answering capabilities of large language models by incorporating structural metadata, particularly in handling structured documents. The author also discusses the divide-and-conquer strategy for tool interaction and novel evaluation methods, which can improve the performance of goal-oriented dialogue systems. The article points out the challenges faced by LLM-based ToD systems in responding to user queries, and the advantages of introducing future dialogue behavior prompts to increase the success rate of outputs. Finally, it emphasizes the need for new evaluation methods to comprehensively assess the performance of dialogue systems, including goal achievement rate, efficiency, and user satisfaction.