Mingju Gao
Papers
1
Total Citations
2
H-Index
1
About
Mingju Gao is a rising researcher in computer vision and generative modeling, with a focus on part-level dynamics and world models. Their most notable contribution, "PartRM: Modeling Part-Level Dynamics with Large Cross-State Reconstruction Model" (2025), addresses a critical gap in predicting how individual components of objects evolve over time from observations and actions. Unlike prior methods that require fine-tuning large pre-trained video diffusion models, Gao’s work introduces a scalable reconstruction-based framework that captures fine-grained, part-level motion without extensive retraining. This innovation has immediate relevance for robotics, animation, and simulation, where understanding object articulation is key. While early in its trajectory, the paper has already garnered 2 citations, signaling growing interest from the community. Gao’s research sits at the intersection of generative AI and physical reasoning, pushing toward more interpretable and controllable world models. Their work is particularly notable for challenging the dominant paradigm of full-scene video prediction, instead advocating for modular, part-centric approaches that could enable more efficient and accurate long-horizon planning. As the field moves toward embodied AI and interactive environments, Gao’s contributions offer a promising direction for building models that truly understand object-level dynamics.
Research Focus
Key Achievements
Top Papers
- 1