Gwanwoo2026MIT

OGBench puzzle-3x3-play-v0

Unofficial mirror of the puzzle-3x3-play-v0 dataset from OGBench, a benchmark for offline goal-conditioned reinforcement learning. Contains data for training and evaluating goal-conditioned RL agents on a 3x3 puzzle task in MuJoCo simulation.

Downloads68

Why This Matters for Physical AI

This offline goal-conditioned RL benchmark enables evaluation of agents that learn to reach arbitrary goals from fixed datasets, which is crucial for developing generalizable robotic control policies without online interaction.

Technical Profile

Environment
simulation
Task Types
manipulation
Data Format
npz
Annotation Types
goal_labels
License
MIT
Part of the OGBench family

Access

Need custom physical AI data?

Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.

Request a Sample Pack

Related Datasets