Olivier Sigaud

Moving quality-diversity optimization from raw policy parameters into the latent space of a behavioral foundation model rescues evolutionary search from total collapse in sparse, deceptive environments while eliminating critic training entirely.

LG-AC outperforms traditional goal-conditioned methods by explicitly leveraging intermediate goals, tackling the notorious challenges of goal chaining in reinforcement learning.

MODIP outperforms traditional behavioral cloning and rivals advanced reinforcement learning methods, showcasing a new frontier in efficient policy fine-tuning for robot learning.