Wall-OSS-0.5 Released: 4B VLA with Open Training Code and Zero-Shot Robot Eval

Core summary

One sentence to understand this update

X Square Robot released Wall-OSS-0.5, a new 4B Vision-Language-Action (VLA) model built on a 3B VLM backbone with a Mixture-of-Transformers layout, featuring open training code and zero-shot real-robot evaluation.

Impact & opportunity

What this could mean

Robotics AI developers and researchers can utilize Wall-OSS-0.5's open code and evaluation methods to accelerate their work in embodied AI and real-robot control.