Search papers, labs, and topics across Lattice.
2
0
5
0
Vision-language-action models let robots follow language instructions, but their language backbones of several billion parameters are the main obstacle to running them on robot hardware, and removing 63% of it from OpenVLA-OFT drops LIBERO-Long success from 93.2% to 0.8%.
CoTinyVLA outperforms larger models by 15.9 points in the most challenging tasks, proving that smarter supervision beats sheer size in VLA systems.