OpenVLA-OFT
Emerging4papers using it
2025first seen
The 'OpenVLA-OFT' dataset/benchmark is used to evaluate Vision-Language-Action models in the context of robotic manipulation tasks, focusing on their ability to integrate visual inputs and natural language instructions while adhering to physical principles.
Papers using OpenVLA-OFT (4)
- PhysVLA: Towards Physically-Grounded VLA for Embodied Robotic ManipulationVISTA: Enhancing Visual Conditioning via Track-Following Preference Optimization in Vision-Language-Action ModelsMAPS: Preserving Vision-Language Representations via Module-Wise Proximity Scheduling for Better Vision-Language-Action GeneralizationSimpleVLA-RL: Scaling VLA Training via Reinforcement Learning