Anjali Narayan-Chen

Amazon (United States), University of Illinois Urbana-Champaign

Papers

Total Citations

156

H-Index

About

Anjali Narayan-Chen is a leading researcher at the intersection of natural language processing, embodied AI, and human-robot interaction. Her work focuses on enabling robots and virtual agents to communicate with humans through natural language to collaboratively solve grounded, real-world tasks. Her most impactful contribution is the TEACh dataset (Task-Driven Embodied Agents That Chat), which has garnered 89 citations. TEACh provides over 3,000 human-human interactive dialogues where agents must understand and execute instructions while using conversation to resolve ambiguity and recover from mistakes—a critical step toward deploying robots in human spaces. Earlier, she pioneered collaborative dialogue in Minecraft (56 citations), defining a grounded building task that simulates physical collaboration without requiring robots. This work demonstrated how game environments can accelerate research in interactive agents. Narayan-Chen also co-authored "Towards Problem Solving Agents that Communicate and Learn" (2017), a foundational paper linking language grounding with machine learning. Her research is essential reading for anyone interested in building agents that don’t just follow commands, but actively converse to achieve shared goals.