The field of artificial intelligence is witnessing significant advancements in the development of multi-agent systems, particularly in the context of complex task solving. These systems, which involve the interaction of multiple agents, are being designed to tackle challenging tasks such as visual question answering, data analysis, and scientific discovery. A key trend in this area is the use of large language models (LLMs) as a core component of these systems, enabling them to reason, learn, and adapt in complex environments. Another notable direction is the integration of multimodal interaction, where agents can process and generate multiple forms of data, such as text, images, and code. This allows for more effective and efficient problem-solving, as well as improved transparency and interpretability. Noteworthy papers in this area include the introduction of iMAD, a token-efficient framework for multi-agent debate, and DataSage, a novel multi-agent framework for insight discovery with external knowledge retrieval and multi-path reasoning. Additionally, the development of OpenBioLLM, a modular multi-agent framework for genomic question answering, and TIM, a framework for DeFi user transaction intent mining, demonstrate the potential of multi-agent systems in diverse applications.