Motivation. Artificial Intelligence (AI) agents powered by Large Language Models (LLMs) have demonstrated advanced capabilities in automating complex tasks. With the rapid evolution of LLMs’ reasoning abilities, such agents have achieved notable success in a variety of domains, including software development and robotics. More recently, the emergence of Multimodal Foundation Models (MFMs) marks a significant advancement in artificial intelligence. These models integrate multiple sensory modalities—such as vision, language, and audio—to substantially enhance the cognitive and perceptual capabilities of AI agents. Recent breakthroughs reveal that MFMs enable fine-grained cross-modal reasoning by jointly processing visual, textual, and auditory data. These systems excel at parsing complex scenarios, generating context-aware descriptions, and tackling tasks requiring synergistic perception and language understanding—capabilities critical to robotics, human-computer interaction, and beyond. Thus, this workshop aims to explore how AI agents can be further empowered through MFMs to operate effectively in multimodal environments.

Background & Application. Recent advances have demonstrated the potential of AI agents to operate in increasingly complex, multimodal settings. For instance, OS-Copilot exhibits human-like interaction capabilities with operating systems, encompassing activities such as web browsing, programming, and engaging with third-party applications. Another prominent application area involves AI Scientist Agents, designed to automate research workflows—including experiment design, execution, and analysis. However, existing implementations of such agents are often limited to processing textual inputs, thereby overlooking the rich, complementary information available from visual and auditory modalities. Lastly, Embodied Agents represent another rapidly evolving area, where AI agents interact with the physical (or simulated) world using sensors and robotic effectors. These applications exemplify the growing need for agents that can reason across modalities to achieve human-level situational awareness and actionability.

Challenges. Multimodal reasoning introduces a distinct set of theoretical and technical challenges that extend beyond those faced in unimodal contexts. First, integrating heterogeneous data (text, images, video) demands innovations in model architectures, training paradigms, and evaluation frameworks. The increased computational demands associated with processing multimodal inputs also raise concerns regarding scalability, efficiency, and real-time performance.

Second, applying these models to frontier domains—such as multi-agent collaboration, scientific discovery, and embodied intelligence systems—requires overcoming bottlenecks in cross-modal semantic understanding and knowledge transfer. Crucially, the interpretability and robustness of multimodal reasoning systems remain unresolved foundational issues, directly impacting the deployment of reliable real-world applications (e.g., scientific AI agents, bio-inspired robotics). Addressing these challenges necessitates breakthroughs in cross-modal representation alignment, dynamic attention mechanisms, and uncertainty modeling of multi-source information.

About this Workshop. All discussions under the scope of multimodal reasoning are welcome. This workshop aims to bring. together researchers with various backgrounds to study the next generation of multimodal reasoning systems. To foster an inclusive dialogue and debate space, we invite speakers and panelists from diverse backgrounds and areas of expertise. Our roster includes both renowned researchers and emerging investigators who have driven promising advances in the field.