Search papers, labs, and topics across Lattice.
6
0
7
13
World Pilot achieves an unprecedented 84.7% success rate in zero-shot manipulation tasks by integrating anticipatory scene and motion priors into VLA models.
TVIR-Agent reveals that integrating visual elements into report generation can dramatically improve the quality and reliability of analytical outputs.
Training agents in MobileGym transfers surprisingly well to real-world mobile devices, retaining over 95% of the simulation-side performance gains.
Existing GUI agents can parrot actions, but AutoGUI-v2 reveals they still lack a deep understanding of GUI functionality and struggle to predict the outcomes of even simple interactions.
You don't need billions of parameters to accurately ground GUI elements: GoClick, a 230M parameter model, matches the performance of much larger models, opening the door for on-device GUI agents.
By forecasting compact world dynamics before taking action, DynVLA leapfrogs traditional CoT methods to achieve more informed and physically grounded autonomous driving decisions.
