Artificial intelligence is rapidly advancing from single-modality perception toward cross-modal representation learning, generative content creation, and decision intelligence for real-world scenarios. Breakthroughs in multimodal large language models, cross-modal alignment, and generative modeling are providing key momentum for this transformation. As multimodal systems enter complex and dynamic real-world environments, challenges related to robustness, trustworthiness, efficient deployment, and generalizable evaluation are becoming increasingly prominent.
This forum aims to bring together researchers and industry experts from multimodal learning, generative artificial intelligence, and application domains such as healthcare, transportation, industry, and public safety. The discussions will cover multimodal foundation models, cross-modal perception and representation learning, multimodal data fusion and spatio-temporal modeling, vision-language understanding, data-efficient learning, multimodal generation, human-object interaction understanding and motion generation, as well as domain-specific multimodal reasoning, decision-making, and applications.
We warmly welcome contributions on the theoretical foundations and engineering practices of trustworthy, robust, and efficient multimodal intelligent systems.
人工智能正快速从单一模态感知迈向跨模态表征学习、生成式内容创作与面向真实场景的决策智能，多模态大模型、跨模态对齐和生成式建模的突破为这一进程提供了核心驱动力。随着多模态系统进入复杂动态的真实环境，鲁棒性、可信性、高效部署与可泛化评测等问题日益突出。
本分论坛旨在汇聚多模态学习、生成式人工智能及医疗、交通、工业、公共安全等领域的学者与业界专家，研讨多模态基础模型、跨模态感知与表征学习、数据融合与时空建模、视觉—语言理解、数据高效学习、多模态生成、人—物交互理解与动作生成，以及垂直领域的多模态推理、决策与应用。
诚挚欢迎围绕可信、鲁棒、高效的多模态智能系统开展理论研究与工程实践交流。
The following topics are within the scope of this special session, but are not limited to: 以下为本分论坛征稿范围，但不限于：