Search papers, labs, and topics across Lattice.
University of California Los Angeles
4
0
4
Predicting fine-grained traffic from coarse data could revolutionize traffic management systems by drastically improving prediction accuracy without the burden of extensive data collection.
A training-free framework can boost VLA model success rates by over 4% without retraining or additional demonstrations.
Attention heads in VLA models can be repurposed to achieve real-time collision avoidance, outperforming traditional safety filters by 43% in dynamic environments.
VLM agents exhibit vastly different skill evolution patterns, revealing that initial performance scores can be misleading without considering improvement dynamics.
