Papers
14
Total Citations
316
H-Index
6
About
Paul Pu Liang is a pioneering researcher at the forefront of multimodal machine learning, a field dedicated to building AI systems that can understand and reason across diverse communicative modalities — including language, vision, audio, and touch. His most influential contribution, "Foundations & Trends in Multimodal Machine Learning," has rapidly accumulated over 150 citations since 2024, establishing itself as a foundational reference for researchers and practitioners alike. Liang has been instrumental in shaping the theoretical and practical landscape of the field, authoring comprehensive tutorials and benchmark frameworks such as MultiBench, which systematically evaluates multimodal representation learning across real-world domains like healthcare, robotics, and affective computing. His work on DIME advances interpretability in multimodal AI, addressing the critical need for transparent and trustworthy decision-making systems. More recently, Liang has extended his research into social AI agents, robotic learning, and high-modality transformer architectures, as well as contributing to OpenFace 3.0 for facial behavior analysis. Through rigorous benchmarking, principled theoretical frameworks, and interdisciplinary collaboration, Liang's cumulative body of work is actively defining the standards and open questions that will guide multimodal AI research for years to come.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5MultiBench: Multiscale Benchmarks for Multimodal Representation Learning22 citations · 2021
- 6
- 7
- 8
- 9
- 10