About this paper
MMFlash improves multimodal speculative decoding by adaptively upweighting visually dependent tokens and using sparse attention for long video, yielding up to 3.58× lossless speedup.
Speculative decoding accelerates generation by letting a lightweight drafter propose tokens for parallel verification by a target model. Diffusion drafters such as DFlash propose an entire token block in a single forward pass, but mainly focus on text-only tasks, leaving multimodal drafting underexplored. To study this setting, we first introduce a metric of token-level visual dependence based on the KL divergence between next-token distributions with and without visual input. We find that acceptance rates decrease from 46% in the lowest visual-dependence quartile to 22% in the highest, with this disparity persisting after multimodal training. Motivated by this observation, we propose…
speculative decodinglightweight draftertarget modelparallel verificationdiffusion draftersDFlashtoken blocksingle forward passtext-only tasksmultimodal drafting
- Upvotes
- 21
- Authors
- 5
- Published
- Nov 13, 2026
Summary and abstract excerpt via Hugging Face Papers ↗ · arXiv.
Trending history
MMFlash: Faster Speculative Decoding for Vision-Language Models was first seen on Hugging Face · Daily papers on Nov 16, 2026. Since then it has appeared in 15 checks on 2 different days. Its best position was #5 of 30, reached on Nov 16, 15:28 UTC. It dropped off the list after Nov 17, 11:28 UTC.