Search papers, labs, and topics across Lattice.
4
0
9
Forget training separate models for each video generation condition – Video2LoRA lets you control video semantics with a single, lightweight (150MB) model.
By fusing physics-based modeling with learned residuals, aerial manipulators can now achieve robust wall-contact execution even in the face of unpredictable wind and near-wall aerodynamic effects.
Forget manually curating training data: RPDR automatically synthesizes and selects "easy-to-learn" examples, boosting dense retriever performance on long-tail QA tasks.
Expert-annotated geographic reasoning data and specialized rewards unlock superior geolocation performance in RL agents, outperforming even large VLLMs.
