Brief

An open-weights vision-language-action (VLA) model named Wall-OSS-0.5 has demonstrated significant progress on real-world robotic tasks without task-specific fine-tuning. The model achieved over 80% task completion on four out of seventeen tasks, including novel ones like tightening a deformable rope. Researchers highlight that the model appears to maintain general image and language understanding while improving embodied grounding, a balance often difficult to achieve in robotics. AI

X Square Robot has released Wall-OSS-0.5, a 4 billion parameter vision-language-action (VLA) model. The model is built upon a 3 billion parameter vision-language model backbone and incorporates action experts using a Mixture-of-Transformers architecture. Notably, the research evaluates the model's performance on real robots before fine-tuning, demonstrating strong zero-shot capabilities and significant improvements after task-specific adaptation. AI