Dual Advantage Fields paper for offline goal-conditioned RL

Arxiv CS.LG (machine learning)

Proposes dual goal representations that capture both global goal reachability and local action comparisons. The method combines value fields for long-horizon reachability with local action selection.

reinforcement-learning

Arxiv CS.LG (machine learning)·Alexey ZemtsovMaxim BobrinAlexander NikulinDmitry V. DylovFakhri KarrayVladislav KurenkovMartin Tak\'a\v{c}Arip Asadulaev

7 days ago