OpenVLA

Splat2Act is a new automated pipeline that bridges the gap between generalist robot policies and specific user environments. By leveraging 3D Gaussian Splatting, it creates high-fidelity simulations from simple mobile scans to train robots for personalized tasks.

TurboVLA is a breakthrough Vision-Language-Action model designed for real-time robotics with unprecedented efficiency. It delivers high success rates on the LIBERO benchmark while maintaining a tiny memory footprint and low latency.

Researchers have uncovered a critical vulnerability in World-Action Models (WAMs) where adversarial visual perturbations can decouple a robot's predicted future from its actual movements. This 'World-Action Drift Attack' allows an attacker to hijack robot actions while leaving the model's imagined future appearing visually plausible and safe.