The first unmistakable signal from another intelligence would rank alongside fire and writing in the story of our species, ...

Abstract: Speculative decoding has proven to be a powerful method for speeding up autoregressive inference by allowing tokens to be generated in parallel using a draft-then-verify approach. However, ...

Speculative decoding is a widely adopted technique for accelerating inference in large language models (LLMs), yet its application to vision-language models (VLMs) remains underexplored, with existing ...

Vision-Language-Action (VLA) models demonstrate remarkable potential for generalizable robotic manipulation. The performance of VLA models can be improved by integrating with action chunking, a ...

We propose FreeDave (Free Draft-and-Verification), a fast sampling algorithm for diffusion language models, which achieves lossless parallel decoding via a pipeline of parallel-decoded candidate ...

Imagine you’re a physician and you are called in to evaluate a patient who has had a sudden change in his neurological status, likely a stroke. You find him alert, mobile, and talking. But when you ...