Embodied AIVision-Language-Action ModelModel CompressionReal-Time Inference3D Spatial Representation
Knowledge Distillation
2 related stories
Samsung Researchers Show Compressed VLA Model Outperforms Larger Teacher on Real Robots
Researchers from Samsung R&D Institute demonstrate that compressing a visual-language-action model's Transformer layers from 18 to 6 via knowledge distillation reduces inference latency by more than half while maintaining success rates, enabling faster real-world robotic deployment.
Track4Action: Distilling 3D World Dynamics Into VLA Policies for Robust Robot Control
A new approach by Shanghai Jiao Tong University and Zhejiang University integrates 3D world tracking knowledge into Vision-Language-Action models to help robots understand how their actions change the physical environment.