Hugging Face Papers ↗ · 2026-11-17
VSpark: Speculative Decoding for Autoregressive Vision-Language-Action Models
Media reports Hugging Face Papers Source language: Chinese
VSpark speeds up autoregressive vision-language-action models by using calibrated action-noise rollouts, hidden-feature alignment, and reuse of rejected drafts, achieving lossless 1.28x–2.10x speedups across various robot manipulation tasks.
Full text
VSpark accelerates autoregressive vision-language-action models by improving speculative decoding with calibrated action-noise rollouts, hidden-feature alignment, and reuse of rejected-draft continuations, achieving lossless 1.28x–2.10x speedups across multiple robot manipulation tasks.