CALVIN-3D_PCD-ABCD_D: Point Cloud Data for Vision-Language-Action Models

FALCON-VLA's dataset supports grounding Vision-Language-Action models in spatial foundation priors, as described in an ICLR 2026 paper. The dataset page was last updated on May 25, 2026. Authors include Zhengshen Zhang, Hao Li, Yalun Dai, and others from the FALCON-VLA organization.

Use Cases

• Training spatial reasoning models based on 3D point cloud data.
• Benchmarking Vision-Language-Action models on tasks requiring spatial grounding.
• Developing foundation models for robotics based on the described spatial priors.

Strengths

• Associated with a peer-reviewed ICLR 2026 conference paper.
• Last updated on 2026-05-25, suggesting recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

FALCON-VLA

Freshness

Last updated 2026-05-25 09:50:56; freshness should be verified.