3 Papers Accepted at ACM MM 2025

The 33^rd ACM International Conference on Multimedia (ACM MM) will be held in Dublin, Ireland, from 27 – 31 October. As a premier international conference, ACM MM gathers leading minds from academia and industry to explore cutting-edge developments in multimedia. The conference will spotlight the latest innovations and applications across a wide range of fields, including video, haptics, virtual and augmented reality, audio, speech, music, sensor, and social data.

Congratulations to the following scientists from A*STAR Centre for Frontier AI Research (A*STAR CFAR) on having their papers accepted at ACM MM 2025:

Prof Ong Yew Soon, Chief Artificial Intelligence (AI) Scientist and Advisor
Dr Joey Zhou, Deputy Director, Principal Scientist
Dr Yin Haiyan, Senior Scientist
Dr Zhang Jie, Scientist

1.	Crowd Dynamics Demand Adaptivity: Self-Adaptive Physics-Informed Neural Network for Crowd Simulation Ziying Tan, Linbo Luo, Haiyan Yin, Yew-Soon Ong, Wentong Cai This paper presents a self-adaptive, physics-informed neural network for crowd simulation that dynamically adjusts to different scene demands. By integrating physical priors into learning and enabling adaptive fidelity control, the model achieves realistic, efficient, and scalable crowd behaviour generation. The proposed system paves the way for intelligent simulation in complex public environments such as transportation hubs and urban planning.
2.	Learning from Heterogeneity: Generalising Dynamic Facial Expression Recognition via Distributionally Robust Optimisation Feng-Qi Cui, Anyang Tong, Jinyang Huang, Jie Zhang, Dan Guo, Zhi Liu, Meng Wang This paper proposes a heterogeneity-aware framework (HDF) for dynamic facial expression recognition, combining a dual-branch time-frequency attention module and an adaptive optimisation module to improve robustness under real-world distribution shifts.
3.	Learning to Be A Doctor: Agent Architecture Search for A Good Medical Agent Yangyang Zhuang, Wenjia Jiang, Jia-Yu Zhang, Ze Yang, Joey Tianyi Zhou, Chi Zhang Existing medical agent systems often rely on static, manually crafted workflows that lack the flexibility to accommodate diverse diagnostic requirements and adapt to emerging clinical scenarios. Motivated by the success of automated machine learning (AutoML), this paper introduces a novel framework for the automated design of medical agent architectures

Learn more about ACM MM 2025.