CLARU
VLA-Arena2025apache-2.0

VLA-Arena Dataset (L1 - Large Variant)

An open-source benchmark for systematic evaluation of Vision-Language-Action (VLA) models, featuring 55 tasks at difficulty level 1 with 2,750 human demonstrations across safety, distractor, extrapolation, and long horizon domains.

Downloads77
Episodes2750

Why This Matters for Physical AI

VLA-Arena provides a comprehensive benchmark for evaluating vision-language-action models on safety, generalization, and long-horizon reasoning critical for deploying robotic agents in real-world environments.

Technical Profile

Modalities
rgbproprioceptionlanguage
Action Space
end_effector_delta
Environment
simulation
Task Types
manipulationgraspingpick_and_place
Episodes
2750
Data Format
RLDS
Annotation Types
language_instructionsaction_labels
License
apache-2.0
Part of the VLA-Arena 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

VLA-Arena Dataset (L0 - Medium Variant)

Level 0 (L0) - Medium (M) variant of the VLA-Arena benchmark containing 1,800 human demonstrations across 60 distinct tasks for training Vision-Language-Action models. The dataset is formatted according to Lerobot specifications and optimized for training the SmolVLA model.

rgbproprioceptionlanguage
80 downloadsDec 2025apache-2.0

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
337K downloadsApr 2026CC BY-NC-SA 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
250K downloadsApr 2026other

Open-H-Embodiment

A multi-embodiment community-driven dataset of paired kinematics and video for training AI autonomy models in surgical robotics and ultrasound applications. The dataset includes tabletop exercises, clinical procedures, and simulations of healthcare robotics tasks.

rgbproprioception
111K downloadsMar 2026cc-by-4.0

OmniAction

A large-scale multimodal dataset for proactive robot manipulation comprising 141,162 episodes with cross-modal contextual instructions derived from spoken dialogue, environmental sounds, and visual cues rather than explicit commands.

rgbaudiolanguage
63K downloadsMar 2026cc-by-nc-4.0

Egocentric-100K

The largest dataset of manual labor with 100,405 hours of egocentric video from head-mounted fisheye cameras, featuring state-of-the-art hand visibility and active manipulation density.

rgbvideo
50K downloadsFeb 2026apache-2.0