CLARU
Spa-Benchapache-2.0

Spa-Bench Evaluation Rollouts — VLA-0 (Partial)

Physical Spa-Bench rollout trajectories with synchronized middle and wrist RGB video, robot state, action, timestamps, episode indices, and task indices for evaluation of VLA-0 model across multiple task families.

Downloads30
Episodes72

Why This Matters for Physical AI

This dataset provides benchmark evaluation rollouts for assessing vision-language-action models on diverse spatial reasoning and manipulation tasks in physical environments.

Technical Profile

Modalities
rgbproprioception
Environment
lab
Task Types
countingordinal_positionphysical_statereferential_descriptionrelational_placementrelative_size
Episodes
72
Data Format
LeRobot
Annotation Types
task_indicesepisode_indices
License
apache-2.0
Part of the Spa-Bench family

Access

View on HuggingFace

Need custom rgb data?

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

Request a Sample Pack

Related Datasets

Spa-Bench Evaluation Rollouts — pi0.5 (Full)

Physical Spa-Bench rollout trajectories with synchronized middle and wrist RGB video, robot state, action, timestamps, episode indices, and task indices for benchmark evaluation. Contains 920 recordings evaluated on task families including Counting, Ordinal Position, Physical State, Referential Description, Relational Placement, and Relative Size.

rgbproprioception
48 downloadsSep 2026apache-2.0

Spa-Bench Evaluation Rollouts — MolmoAct2 (Full)

Physical Spa-Bench rollout trajectories with synchronized middle and wrist RGB video, robot state, action, timestamps, episode indices, and task indices evaluated with the MolmoAct2 model. Coverage includes 900 scored recordings across task families including Counting, Ordinal Position, Physical State, Referential Description, Relational Placement, and Relative Size.

rgbproprioception
35 downloadsSep 2026apache-2.0

Spa-Bench Evaluation Rollouts — GR00T-N1.7 Full Fine-Tune (Partial)

Physical Spa-Bench rollout trajectories with synchronized middle and wrist RGB video, robot state, action, timestamps, episode indices, and task indices collected at 30 FPS. Covers 120 scored familiar-condition recordings across six task families for evaluating the GR00T-N1.7 model.

rgbproprioception
33 downloadsSep 2026apache-2.0

Spa-Bench Teleoperated Training Dataset — 1,200 Episodes

A teleoperated demonstration dataset of 1,200 episodes with five synchronized camera streams used for compositional generalization experiments across six task families involving manipulation and object interactions.

rgbproprioception
21 downloadsSep 2026apache-2.0

s1312003/addition1_20260721_145600

A robotics dataset created using LeRobot framework containing 10 episodes of robot manipulation trajectories with RGB video observations and joint position actions.

rgbproprioception
null downloadsJul 2026apache-2.0gated

s1312003/cam2_2_20260721_152437

A robotics dataset created using LeRobot containing 50 episodes of robot manipulation data with RGB video observations and joint position actions.

rgbproprioception
null downloadsJul 2026apache-2.0gated