Grace Zhang

Generalize and Guide: Decomposing Rewards for Few-Shot Inverse Reinforcement Learning

Ziyi Liu, Grace Zhang

Inverse reinforcement learning (IRL) provides a powerful framework for learning from demonstrations. However, real-world tasks often exhibit substantial natural variations (e.g., picking up mugs with varying shapes), making it impractical to collect demonstrations that fully spec…Read more