@animesh_garg @_k_sridhar https://veo-robotics.github.io/ Enjoy, Animesh ;)
@_k_sridhar But you didnt release any of it. Did you?
A Google DeepMind research team shared a project page describing how a fine-tuned Veo video model can generate varied robot scenes for testing Gemini robotics policies, including in-distribution checks, out-of-distribution shifts, and safety probes, with claims validated against real-world trials on bimanual tasks.
@animesh_garg @_k_sridhar https://veo-robotics.github.io/ Enjoy, Animesh ;)
@_k_sridhar But you didnt release any of it. Did you?
The public page offers videos, plots, and method details but no code, weights, or runnable simulator, leaving external researchers to read the arXiv paper without hands-on tools.
Rankings from the simulator matched outcomes across 1600 real evaluations, yet gaps in physics realism and multi-view consistency remain explicit in the write-up.
No Digg Deeper questions have been answered for this story yet.
@keerthanpg @_k_sridhar ah of course!
@animesh_garg @_k_sridhar https://veo-robotics.github.io/ Enjoy, Animesh ;)