Search papers, labs, and topics across Lattice.
1
0
3
13
By decoupling visual and motor information during pretraining, FutureVLA unlocks more effective visuomotor prediction for vision-language-action models, boosting performance without modifying downstream architectures.
