CLARU
VincentNiapache-2.0

RoboTwin SFT 8-Tasks SAPIEN Eval

SAPIEN execution results for 1280 SFT-generated videos across 8 RoboTwin manipulation tasks, with actions extracted via inverse dynamics model and replayed in simulation.

Downloads0
Episodes1280

Why This Matters for Physical AI

This dataset provides end-to-end evaluation of video-to-action generation for robotic manipulation in simulation, enabling assessment of SFT model quality and inverse dynamics model fidelity for embodied AI systems.

Technical Profile

Modalities
rgbvideo
Robot Embodiments
manipulator
Action Space
joint_positions
Environment
simulation
Task Types
manipulationpick_and_placeobject_interaction
Episodes
1280
Data Format
mp4
Annotation Types
action_labelssuccess_labelstask_diagnostics
License
apache-2.0
Part of the RoboTwin family

Access

View on HuggingFace

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

RoboTwin blocks_ranking_rgb — Wan2.2 TI2V Rollouts

160 text+image-to-video rollouts for the blocks_ranking_rgb task from RoboTwin, generated with Wan2.2 TI2V (5B) diffusion model and scored with blocks_ranking_v2 reward using SAM3 object tracking and inverse dynamics.

rgblanguage
0 downloadsApr 2026apache-2.0

CoRL2026-CSI square_ring_on_peg

A robotics manipulation dataset created using LeRobot containing demonstrations of a square ring on peg task performed by a SO follower robot.

rgbproprioception
null downloadsApr 2026apache-2.0

Flip Object - Franka FR3

A robotics dataset containing 106 episodes of a Franka FR3 robot performing object flipping tasks, recorded with multi-view RGB video and proprioceptive state observations.

rgbproprioception
null downloadsApr 2026apache-2.0

EgoTraj-Bench

The first real-world benchmark for pedestrian trajectory prediction under ego-centric noisy observations, pairing noisy first-person-view derived trajectories with clean bird's-eye-view ground truth.

rgbtrajectory
null downloadsApr 2026cc-by-nc-4.0

Xperience-10M

A large-scale egocentric multimodal dataset of human experience containing 10 million interactions and 10,000 hours of synchronized first-person recordings with six video streams, audio, stereo depth, camera pose, hand mocap, full-body mocap, IMU, and hierarchical language annotations for embodied AI, robotics, and world modeling research.

rgbaudiodepthproprioception+3
2.3M downloadsApr 2026other

LingBot-Depth Dataset

Self-curated RGB-D dataset for training masked depth modeling approaches, containing real-world indoor scenes, VLA robot manipulation tasks, and simulated data across multiple camera types and robot platforms.

rgbdepth
261K downloadsApr 2026CC BY-NC-SA 4.0