Skip to main content
Search

# Rising Weight Precision over Flow Steps

1 related articles

Subscribe to updates
  1. arXiv

    Robot action-head weights rise in precision across flow steps: 2 to 4 bits keep closed-loop success, and inference runs 1.7 times faster than NVFP4 on Thor

    A robot vision-language-action model's diffusion action head re-reads all its weights at every step but applies each to only about fifty tokens, so inference is bound by memory bandwidth. Because later flow steps attenuate the quantization errors of earlier ones, weight precision can rise across the steps; with plain per-channel rounding and no calibration or training, a [2,3,3,4] schedule preserves the closed-loop success of three policies while transferring 25% fewer bits than 4-bit weights at every step. Current GPU kernels cannot exploit this because of fixed overheads, so a programmable accelerator scales latency and energy with the transferred weight bits and runs the GR00T N1.6 action head in 8.02 ms per inference under LPDDR5X timing simulation.

You’re all caught up