Chaoyi Deng

About

Chaoyi Deng is a researcher advancing the frontiers of reinforcement learning through unsupervised pre-training and world models. Their work centers on developing methods that enable AI agents to learn from vast, unstructured video data—a paradigm shift from traditional domain-specific or simulated training. Deng’s most notable contribution, "Pre-training Contextualized World Models with In-the-wild Videos for Reinforcement Learning" (2023, 3 citations), introduces a novel approach to leveraging large-scale, real-world video datasets for model-based RL. This work demonstrates how agents can acquire rich, contextualized representations of environments without explicit task supervision, significantly reducing the need for costly, hand-crafted simulations. By bridging unsupervised pre-training with model-based RL, Deng addresses a critical bottleneck in scaling intelligent systems to complex, open-world tasks. Though early in its citation impact, this research has already sparked interest for its potential to democratize RL training—making it more data-efficient and generalizable. Deng’s work sits at the intersection of computer vision, representation learning, and decision-making, offering a promising path toward more adaptable and autonomous AI systems.

Research Focus

Key Achievements

Top Papers

1. 1
   Pre-training Contextualized World Models with In-the-wild Videos for Reinforcement Learning

   3 citations · 2023

Key Collaborators

Contact & Links