Wonje Jeung
Papers
1
Total Citations
3
H-Index
1
About
Wonje Jeung is a researcher at the forefront of embodied AI and instruction following, with a focus on bridging the gap between simulated agents and real-world robotics. His most notable contribution is the development of **ReALFRED**, a benchmark designed to evaluate embodied agents in photo-realistic environments, which challenges models to follow natural language instructions while navigating and manipulating objects in visually complex, dynamic settings. This work, published in 2024 and already garnering early citations, addresses critical limitations in prior benchmarks by emphasizing perceptual realism and task diversity. Jeung’s research pushes the boundaries of how AI systems understand and execute human commands in physical spaces, with implications for household robotics and assistive technologies. By creating standardized, rigorous evaluation frameworks, he enables the community to measure progress in grounding language to action. His work is essential reading for students and researchers interested in embodied AI, vision-language navigation, and the intersection of simulation and reality.
Research Focus
Key Achievements
Top Papers
- 1