Beyond VLAs: How World Action Models Reshape Robot Manipulation

<img width="600" height="338" src="https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models.gif" class="webfeedsFeaturedVisual wp-post-image" alt="A GIF of a robot following directions” with a description of the actual robot, task, objects, and motion shown." style="display: block; margin-bottom: 5px; clear:both;max-width: 100%;" link_thumbnail="" decoding="async" srcset="https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models.gif 600w, https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models-179x101.gif 179w, https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models-300x169.gif 300w, https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models-500x282.gif 500w, https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models-160x90.gif 160w, https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models-362x204.gif 362w, https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models-195x110.gif 195w" sizes="(max-width: 600px) 100vw, 600px" title="Cosmos-World-Models" />A central challenge in robotics is building policies that generalize beyond the demonstrations they’re trained on. A policy that succeeds in a training scene...<img width="600" height="338" src="https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models.gif" class="webfeedsFeaturedVisual wp-post-image" alt="A GIF of a robot following directions” with a description of the actual robot, task, objects, and motion shown." style="display: block; margin-bottom: 5px; clear:both;max-width: 100%;" link_thumbnail="" decoding="async" loading="lazy" srcset="https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models.gif 600w, https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models-179x101.gif 179w, https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models-300x169.gif 300w, https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models-500x282.gif 500w, https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models-160x90.gif 160w, https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models-362x204.gif 362w, https://developer-blogs.nvidia.com/wp-content/uploads/2026/07/Cosmos-World-Models-195x110.gif 195w" sizes="auto, (max-width: 600px) 100vw, 600px" title="Cosmos-World-Models" /><p>A central challenge in robotics is building policies that generalize beyond the demonstrations they’re trained on. A policy that succeeds in a training scene often fails when object shapes, positions, or lighting change. Generalizing to these new conditions requires the policy to understand the tasks underlying physics, not just mimic the demonstrations. This ability comes from the backbone it’s…</p> <p><a href="https://developer.nvidia.com/blog/beyond-vlas-how-world-action-models-reshape-robot-manipulation/" rel="nofollow" data-wpel-link="internal" target="_self">Source</a></p>