Abstract: With the rapid development of artificial intelligence technology, multi-robot systems have shown significant advantages in task execution efficiency, environmental adaptability and robustness. However, traditional multi-robot systems face challenges such as communication coordination, real-time decision-making and complex task decomposition. In recent years, the explosive progress of large language model (LLM) provides a new paradigm for the intelligentization of multi-robot systems, enabling efficient task planning, human-robot interaction, and inter-robot collaboration through its powerful natural language understanding, logical reasoning and generalization capabilities. We systematically review the research progress of LLM-based multi-robot systems, sort out the current mainstream benchmarks, and analyze the core challenges encountered by LLM-based multi-robot systems.