Search papers, labs, and topics across Lattice.
This paper introduces GaussianDream++, an efficient approach to 3D Gaussian world modeling that enhances robotic manipulation through the integration of World State Tokens and World Prediction Tokens into Vision-Language-Action (VLA) policies. By employing a training-only World Representation Head that decodes these tokens into a Current World and Future Prediction, the method achieves high performance on benchmark tasks while significantly reducing deployment costs. The results show a remarkable improvement in real-robot success rates, increasing from 29.2% to 52.5%, demonstrating the effectiveness of the proposed architecture in dynamic environments.
Achieving a 52.5% success rate in real-robot manipulation tasks, GaussianDream++ redefines efficiency in 3D world modeling without the need for online Gaussian decoding.
Vision-Language-Action (VLA) policies have advanced language-conditioned robotic manipulation, yet action-imitation objectives provide only weak supervision for metric 3D structure and short-horizon physical evolution. Geometry-enhanced policies mainly improve current-scene grounding, whereas predictive policies often model future dynamics in RGB or latent spaces and may incur substantial deployment cost. GaussianDream demonstrates that training-time current Gaussian reconstruction and future Gaussian prediction provide effective 3D supervision, but its dense VGGT/TGE-based prefix jointly carries state, dynamics, and action-conditioning information. We present \textbf{\methodname}, a compact, policy-native extension that inserts \textbf{World State Tokens} and \textbf{World Prediction Tokens} directly into the VLA backbone. A training-only \textbf{World Representation Head} decodes these tokens into a Current World and coupled Future Prediction over shared Gaussian primitives, while static--dynamic factorization preserves persistent structure and focuses residual motion on interaction-relevant regions. At inference, the head, renderer, auxiliary objectives, and VGGT/TGE pathway are removed, leaving only 20 world tokens without online Gaussian decoding or rollout. \method achieves \textbf{98.6\%} on LIBERO and \textbf{87.8\%} on LIBERO-Plus, with clear gains under Camera and Layout shifts. Real-robot experiments further improve average success from 29.2\% to 52.5\% over reproduced $\pi_{0.5}$ while maintaining efficient closed-loop control.
