Announcement
ECHO reportedly beats pure RL on score and efficiency in small robotics-like experiments
A user says the setup had a turn cap and no thinking to simulate time-constrained environments.
TLDR
A user describes small ECHO experiments in a robotics-like environment, using a turn cap and no thinking to simulate time constraints. They say ECHO performed better than pure reinforcement learning (RL) on score and efficiency, and that a short ECHO phase followed by RL worked well.
Combined views
2.1K
4 Sources, first seen ago
