Search papers, labs, and topics across Lattice.
University of New South Wales
3
3
6
6
By disentangling structure and motion in the latent space, CoWVLA achieves superior visuomotor learning compared to standard world-model and latent-action approaches.
Standard panoramic segmentation models crumble when faced with real-world camera rotations, but SO3UFormer maintains high accuracy even under arbitrary 3D reorientations by learning rotation-invariant spherical features.
VLMs can now excel at industrial anomaly detection by injecting domain-specific facts and aligning with expert preferences, achieving state-of-the-art zero-shot and one-shot performance.
