Lihan Zha
Papers
3
Total Citations
46
H-Index
3
About
Lihan Zha is a rising researcher at the intersection of robotics, natural language processing, and human-robot interaction. Their work focuses on enabling robots to generalize to novel environments by leveraging human feedback and large language models (LLMs). Zha’s major contributions include developing methods for robots to learn from online human corrections, distilling generalizable knowledge for manipulation tasks, and grounding LLMs in physical execution. Their most-cited paper, "Distilling and Retrieving Generalizable Knowledge for Robot Manipulation via Language Corrections" (2024, 25 citations), addresses the critical challenge of robot policy generalization by using human corrective feedback to improve performance in unfamiliar settings. Another key work, "DoReMi: Grounding Language Model by Detecting and Recovering from Plan-Execution Misalignment" (2024, 17 citations), introduces a framework that detects and recovers from mismatches between high-level plans and low-level execution, bridging the gap between LLM-generated plans and real-world action. This research has significant implications for deploying robots in dynamic, unstructured environments, such as homes or factories. Zha’s innovative approach to combining language models with robotic control is paving the way for more adaptable and intelligent autonomous systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3