Search papers, labs, and topics across Lattice.
4
0
5
4
Robots can now plan 9x faster and achieve significantly higher success rates by decoupling action prediction from video generation in World-Action Models.
Achieve 100% success rates in visually ambiguous manipulation tasks by fusing high-frequency tactile data with low-frequency visual planning, outperforming visual-only baselines and satisfying hard real-time constraints.
Flow-based VLAs can now learn online without likelihoods or value networks, unlocking better generalization in complex embodied control tasks.
By compressing future observations into a learned "condition space," WoG enables VLA models to generate more precise actions and generalize better than models relying on direct future prediction.
