Never Give Up ICLR 2020

Random Disco Maze

Pitfall (exploratory policy)

Pitfall (exploitative policy)

MR example seed that passes the level

MR (exploitative policy)

MR (exploratory policy)

MR first episode (hand-crafted controllable states)

MR second episode (hand-crafted controllable states)

MR third episode (hand-crafted controllable states)