One of the coolest reward hacks I have seen! Puffer RL agent playing Nethack with a high gold reward learns to kick stacks of gold to scatter the coins. Since single-instance rewards get clamped, it split the pile into several smaller rewards!
Our intern @finlay_sanders is working on this. Props to him. I was just checking on progress and trying to speed up training a bit on stream today
Users praised their intern for work on the RL agent exploiting reward hacks by kicking gold in NetHack, giving props for efforts to speed up training.
Based on 1 visible X reactions from 1 accounts; directional sample.
Ask a question below.
Published answers will appear here.
Our intern @finlay_sanders is working on this. Props to him. I was just checking on progress and trying to speed up training a bit on stream today
One of the coolest reward hacks I have seen! Puffer RL agent playing Nethack with a high gold reward learns to kick stacks of gold to scatter the coins. Since single-instance rewards get clamped, it split the pile into several smaller rewards!
Users praised their intern for work on the RL agent exploiting reward hacks by kicking gold in NetHack, giving props for efforts to speed up training.
Based on 1 visible X reactions from 1 accounts; directional sample.
Ask a question below.
Published answers will appear here.
