bitvla-robot-control

Build fully ternary quantized vision-language-action models for robotic manipulation, achieving 11x memory reduction and 4.4x speedup while maintaining task performance on edge devices.

Independent GitHub snapshot · Automated classification · Original documentation governs