Preprint
Aug 2026
GORDON: Graph-based Object-centric Rewards for Decomposition of Long-Horizon Manipulation
This work proposes GORDON, a graph-based object-centric reward learning framework that learns dense rewards from action-free video demonstrations that improves reinforcement learning in short-horizon settings and enables successful policy learning in complex long-horizon tasks through automatic decomposition.
Andrea Protopapa, Davide Buoso, Francesca Pistilli et al.
· 0 citations