Vrushabh272026MIT

Cosmic — MolmoSpaces Benchmark Evaluation Artifacts

Complete evaluation artifacts for Cosmic, a TAMP + foundation-model manipulation policy evaluated on the MolmoSpaces benchmark suite with a Franka arm using joint-position action space. Results from 11 manipulation tasks with success rates ranging from 40.7% to 87.9%.

Downloads35
Episodes10900

Why This Matters for Physical AI

This benchmark evaluation dataset demonstrates the performance of foundation-model-based manipulation policies on diverse simulated robotic tasks, providing critical benchmarks for evaluating progress in physical AI and TAMP-integrated learning approaches.

Technical Profile

Robot Embodiments
Franka Panda
Action Space
joint_positions
Environment
simulation
Task Types
manipulationpick_and_placegraspingopeningclosing
Episodes
10900
Data Format
HDF5
Annotation Types
reward_labelsaction_labels
License
MIT
Part of the MolmoSpaces 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