RoboCerebra

Overview RoboCerebra is a large-scale benchmark designed to evaluate robotic manipulation in long-horizon tasks, shifting the focus of robotic imitation learning from fast, reactive System 1 behavior to slow, deliberate System 2 reasoning. Resources Code (conversion scripts and evaluation): Benchmark (pre-converted RLDS): Key Features Long-Horizon… See the full description on the dataset page:

Source: Hugging Face Hub (qiukingballball/RoboCerebra). Metadata imported from the dataset’s Hub tags.