Search papers, labs, and topics across Lattice.
The paper introduces GaussVLA, a Vision-Language-Action model that enhances spatial reasoning by integrating a Gaussian Spatial Tokenizer and a Depth-Aware Chain-of-Thought module. This approach allows the model to leverage 3D geometric structures and perform structured reasoning, addressing the limitations of traditional flat 2D patch tokens. GaussVLA achieves a remarkable 93.5% average success rate on the LIBERO benchmark, significantly outperforming its predecessor, SpatialVLA, while maintaining a compact parameter footprint of only 200M parameters.
GaussVLA achieves a 19.7% improvement in spatial manipulation success rates while being more parameter-efficient than previous models.
Vision-Language-Action (VLA) models encode visual observations as flat 2D patch tokens that carry no intrinsic geometric structure, and augmenting them with dense monocular depth injects per-pixel scalar values that encode neither surface orientation nor geometric confidence. This leaves the policy with limited structured spatial reasoning for action prediction. We propose GaussVLA, a Mamba-based VLA that incorporates two custom modules: Gaussian Spatial Tokenizer (GST) to lift frozen semantic and depth features into compact 3D Gaussian tokens, pools geometrically salient regions with learned queries, and \emph{Depth-Aware Chain-of-Thought (DA-CoT)} that performs structured, non-autoregressive geometric reasoning under language and flow-time conditioning. Across both simulation and real-world evaluations, GaussVLA demonstrates strong spatial-manipulation performance while remaining parameter-efficient. On LIBERO, it achieves 93.5% average success and 100.0% success on the Spatial suite with only 200M parameters, improving over SpatialVLA by 19.7% relative average success while remaining significantly more parameter-efficient.
