model · model--graspvla

GraspVLA

Grasping foundation model pretrained on one billion synthetic VLA frames.

Identity and state

PublicationReleased
DeploymentExternal Evaluation
Canonical claims5
Last reviewed2026-08-27

Reviewed relationships

StateRelated entityRelationScopeBoundary
VerifiedPKU EPIC Labcoauthorspaper affiliationThe research system is related to the product stack but remains a versioned evidence node.
VerifiedSynGrasp-1Btrained onsynthetic pretrainingThe source states one billion synthetic VLA frames; frames are not trajectories or hours.

Canonical evidence claims

Claims are projected from authored tables. A dash remains an explicit missing value; it is never filled from a neighboring product or name match.

MetricValueClass
GraspVLAarXiv:2505.03233 [research-systems-r17]paper-body-galbot-g1
GraspVLAhomepage product — homepage and founder profile call it the generalist grasp stack on Galbot G1 [research-systems-r21] [research-systems-p3]paper-body-galbot-g1
GraspVLApaper-body-galbot-g1
GraspVLAreal-robot evals; robot not named in the paper [research-systems-r17]paper-body-galbot-g1
GraspVLAHF shengliangd/GraspVLA [research-systems-r17]paper-body-galbot-g1

Source ledger

What is not inferred

No fuzzy entity matching, adjacent-column borrowing, universal compatibility, commercial maturity, or checkpoint-to-product identity. Absence of a relation means unreviewed or unsupported here—not proven incompatibility.