1 post tagged with "training".
Agent RL's bottleneck moved from data to reward to the environment itself. The newest research tries to take humans out of environment-building entirely.