a3124371940cc-by-4.0

RadeonVLA-Reflex Physical-2K

2,000 strictly validated successful Genesis simulation episodes for language-conditioned Franka fruit sorting, covering five fruits and four bowl positions with 100 episodes per task variation.

Downloads0
Episodes2,000

Why This Matters for Physical AI

This dataset provides a large-scale, rigorously validated benchmark for training vision-language-action models on robotic manipulation with strict success metrics and diverse task variations, though it is simulation-only and does not establish real-robot performance.

Technical Profile

Modalities
rgbproprioceptionlanguage
Robot Embodiments
Franka Panda
Action Space
joint_positions
Environment
simulation
Task Types
manipulationpick_and_placesorting
Episodes
2,000
Data Format
LeRobot
Annotation Types
language_instructionsaction_labels
License
cc-by-4.0
Part of the RadeonVLA-Reflex family

Access

Need custom rgb data?

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

Request a Sample Pack

Related Datasets