Berkeley-ICON-Lab

pi0.5 TACO LIBERO-32 Evaluations

Evaluation artifacts for 36 checkpoints of the pi0.5 TACO model on LIBERO-32 tasks, including results across five random seeds with statistical summaries and rollout videos.

Downloads0
Episodes10

Why This Matters for Physical AI

This evaluation dataset documents systematic benchmarking of continual learning in robotic manipulation, providing comprehensive performance metrics across multiple seeds and tasks that are essential for assessing robot learning generalization.

Technical Profile

Environment
simulation
Task Types
manipulation
Episodes
10
Data Format
JSON
Annotation Types
reward_labels
Part of the LIBERO 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