Biography
Run Peng is a Ph.D. student at University of Michigan, advised by Professor Joyce Chai in the SLED Lab. Before, he actively worked in Lee Lab, advised by Professor Honglak Lee, and worked as a reserach intern at LG AI Research, mentored by Lajanugen Logeswaran and Sungryull Sohn. His research interests include embodied agent modeling and reinforcement learning.
Interests
- Embodied Collaboration
- Computational Linguistics
- Reinforcement Learning
Education
MSE in Computer Science & Engineering, 2024
University of Michigan
BSE in Computer Science & Engineering, 2022
University of Michigan
BSE in Electrical & Computer Engineering, 2022
Shanghai Jiao Tong University
Experience
Research Intern
LG AI Research
Responsibilities include:
- Design and implement intuitive GPT-based evaluation on language generation tasks, which enables automatic, human-like evaluation on open-end dialogues without human evaluators.
- Equip LLM with specific personas to act as humans, having diverse task-oriented conversations with given baselines, to augment the existing datasets with more randomness and diversity.
Graduate student research asssistant
Responsibilities include:
- Leading or playing a major role in multiple projects related to theory-of-mind modeling, human-robot collaboration, language grounding, and embodied environment understanding.
- Detailed descriptions can be found in the project list.
Graduate & Undergraduate student research assistant
Lee Lab @Umich
Responsibilities include:
- Intrinsic motivation design in hard exploration with reinforcement learning
- Task representation refining for meta-learning via contrastive learning
Recent Publications
Quickly discover relevant content by filtering publications.
(2024).
Shall We Team Up: Exploring Spontaneous Cooperation of Competing LLM Agents.
EMNLP 2024 (findings).
(2023).
Towards A Holistic Landscape of Situated Theory of Mind in Large Language Models.
In EMNLP 2023 Findings.
(2023).
Think, Act, and Ask: Open-World Interactive Personalized Robot Navigation.
ICRA 2024.
(2023).
Go Beyond Imagination: Maximizing Episodic Reachability with World Models.
In ICML 2023, Poster.