Research
My research interests focus on LLM agents, deep reinforcement learning, and their applications in decision-making and robotics.
|
|
Verlog: A Multi-turn RL framework for LLM agentsL
Wen-Tse Chen,
Jiayu Chen,
Hao Zhu,
Jeff Schneider,
Proposed a multi-turn reinforcement learning framework built for long-horizon LLM-agentic tasks with highly variable episode lengths.
|
|
Decentralized Navigation of a Cable-Towed Load using Quadrupedal Robot Team via MARL
Wen-Tse Chen*,
Minh Nguyen*,
Zhongyu Li*,
Guo Ning Sue,
Koushil Sreenath
Proposed a scalable, decentralized MARL-based system for coordinating a team of quadrupedal robots to collaboratively tow a cable-connected load through cluttered environments, ensuring flexibility, scalability, and robustness across varying team sizes and environmental conditions.
|
|
Fine-tuning LLM Agents with Retrospective In-Context Online Learning
Wen-Tse Chen,
Jiayu Chen,
Fahim Tajwar,
Hao Zhu,
Xintong Duan,
Russ Salakhutdinov,
Jeff Schneider
NeurIPS Adaptive Foundation Models Workshop, 2024 (Oral presentation)
Presented a sample-efficient method for online fine-tuning LLM agents by using in-context learning to convert sparse feedback into dense signals, enabling LLMs to adapt to dynamic environments with minimal data.
|
|
DGPO: Discovering Multiple Strategies with Diversity-Guided Policy Optimization
Wen-Tse Chen,
Shiyu Huang,
Yuan Chiang,
Tim Pearce,
Wei-Wei Tu,
Chen Ting,
Zhu Jun,
The 38th Annual AAAI Conference on Artificial Intelligence (AAAI2024)
Proposed an on-policy framework for discovering multiple diverse optimal strategies for the same task in a single training process.
|
|