Papers
1
Total Citations
4
H-Index
1
About
Haozhe Liu is a rising researcher in artificial intelligence, with a primary focus on deep reinforcement learning (DRL) and representation learning from visual data. His most notable contribution is the development of methods that enable agents to learn effective policies from passive video observations—without requiring explicit action labels. In his highly cited 2023 work, "Learning to Identify Critical States for Reinforcement Learning from Videos," Liu introduced a novel framework that extracts algorithmic information about good policies from offline video data, addressing a fundamental challenge in DRL: learning from implicit, unlabeled demonstrations. This approach has significant implications for robotics and autonomous systems, where collecting labeled action data is often impractical. With 4 citations in its first year, this paper is gaining traction for its potential to bridge the gap between observational learning and policy optimization. Liu’s work stands out for its innovative use of critical state identification, allowing agents to focus on decision-relevant moments in video streams. As an early-career researcher, his contributions are shaping how machines learn from passive visual inputs, promising more sample-efficient and generalizable reinforcement learning systems.
Research Focus
Key Achievements
Top Papers
- 1