Yaofeng Su

About

I am an undergraduate at Fudan University (GPA 3.84/4.00). Currently, I am an intern at Joy Future Academy, JD, working on streaming audio-visual generation, DMD distillation, and memory for world models. Our team’s JoyAI-Echo achieves minute-long coherent audio-video story generation. My work OmniForcing was selected for a Long Oral at ECCV 2026 (0.27% of submissions), and I presented it in an invited talk to the Lightricks LTX-Video team, including their CTO and Director of GenAI Research. In 2025, I was glad to work with Prof. Huaxiu Yao as an exchange student at UNC Chapel Hill, where I contributed to SimpleMem (ICML 2026) based on deep analysis of long-term conversation characteristics and memory structures.

I’m excited to begin my PhD with Prof. Yinghao Xu in Fall 2027. If you find my research interesting or would like to connect, feel free to reach out at yaofengsu@outlook.com.

Research Interests

My research interest lies in multimodal generative models as a means to build faithful representations of the world. Through reinforcement learning, distillation, and agent frameworks, I have contributed to applications spanning audio-visual generation, autonomous driving, and video understanding. I am particularly drawn to developing principled approaches that bridge generative modeling with physically grounded world understanding.

Education

Invited Talks

Publications

CV