Search papers, labs, and topics across Lattice.
University of Virginia 鈭桬qual contribution.
5
0
9
Action-conditioned world model embeddings can revolutionize failure detection in long-horizon robotic tasks, achieving reliable monitoring without dense annotations.
Loop-aware planning combined with descriptor-aided localization cuts exploration time by 15% and travel distance by 14% in challenging environments.
MLLMs exhibit alarming Stochastic Collapse, failing to maintain randomness even under explicit random instructions, which could undermine their utility in diverse applications.
State-of-the-art shot boundary detection gets a major upgrade with a Transformer-based approach that not only improves accuracy but also offers more interpretable boundaries, thanks to a novel relational prediction framework and synthetic training data.
Achieve 2.6x faster autoregressive world model inference without retraining by caching and selectively reusing block-level residuals across generation chunks.
