GEAR-VLA learns geometry-aware action representations via coarse-to-fine pretraining, gradient-decoupled DiT action expert, semantic-aligned 3D integration, and embodiment canonicalization, reporting SOTA results on LIBERO benchmarks and over 80% success on unseen embodiments and 212 unseen objects.
Dexgraspvla: A vision-language-action framework towards general dexterous grasping
2 Pith papers cite this work, alongside 3 external citations. Polarity classification is still indexing.
2
Pith papers citing it
3
external citations · Crossref
fields
cs.RO 2years
2026 2verdicts
UNVERDICTED 2representative citing papers
HandITL enables seamless human intervention in VLA policies for bimanual dexterous manipulation, cutting jitter by 99.8% and improving refined policies by 19% over standard teleoperation.
citing papers explorer
-
GEAR-VLA: Learning Geometry-Aware Action Representations for Generalizable Robotic Manipulation
GEAR-VLA learns geometry-aware action representations via coarse-to-fine pretraining, gradient-decoupled DiT action expert, semantic-aligned 3D integration, and embodiment canonicalization, reporting SOTA results on LIBERO benchmarks and over 80% success on unseen embodiments and 212 unseen objects.
-
Hand-in-the-Loop: Improving VLA Policies for Dexterous Manipulation via Seamless Hand-Arm Intervention
HandITL enables seamless human intervention in VLA policies for bimanual dexterous manipulation, cutting jitter by 99.8% and improving refined policies by 19% over standard teleoperation.