Announcement
OCBench offers ‘human-like’ scripted policies for robotics research
A post introducing the benchmark says it enables studies of behavior cloning and reinforcement learning with virtually unlimited data.
TLDR
A post introducing OCBench says human demonstrations are expensive and less controllable, making robotics research difficult. It describes the new benchmark as providing ‘human-like’ scripted policies that enable scientific studies of behavior cloning (BC) and reinforcement learning (RL) with virtually unlimited data.
Combined views
3.8K
2 Sources, first seen ago
