Mingsheng Long

About

Mingsheng Long is a leading researcher in machine learning, with a primary focus on predictive learning, world models, and reinforcement learning. His work addresses the critical challenge of building AI systems that can learn and adapt to changing environments over time. In his highly cited 2022 paper, "Continual Predictive Learning from Videos," Long introduced a novel framework for world modeling that handles sequential prediction tasks from different environments, overcoming the traditional assumption that all data is available at once. This work has garnered 5 citations and laid the groundwork for more adaptive AI. Expanding on this, his 2023 study, "Pre-training Contextualized World Models with In-the-wild Videos for Reinforcement Learning," pioneered unsupervised pre-training for model-based reinforcement learning using diverse, real-world video data—a significant departure from domain-specific or simulated datasets. With 3 citations, this contribution pushes the boundaries of scalable, data-efficient RL. Long’s research is pivotal for developing AI that can continuously learn and generalize across tasks, making him a key figure in advancing robust, real-world machine learning systems.

Research Focus

Key Achievements

Top Papers

1. 1
   Continual Predictive Learning from Videos

   5 citations · 2022
2. 2
   Pre-training Contextualized World Models with In-the-wild Videos for Reinforcement Learning

   3 citations · 2023

Key Collaborators

Contact & Links