MMFlash: Faster Speculative Decoding for Vision-Language Models
Sicheng Feng, Zigeng Chen, Song Wang, Gongfan Fang, Xinchao Wang
MMFlash: Faster Speculative Decoding for Vision-Language Models: 40 upvotes on Hugging Face Daily Papers, #5 of 40 papers on 2026-11-16. Day-by-day upvote history.
Upvotes over time
It had 35 upvotes by its first day, 38 upvotes by the next day, 40 upvotes by day 3, 40 upvotes by day 7, and has 40 now. It ranked #15 of 146 papers in its week and #63 of 520 in its month. Its code repository has 42 GitHub stars.
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 MMFlash, a simple and effective framework that adaptively upweights visually dependent tokens that the drafter predicts poorly. For long video inputs, MMFlash further employs sparse attention to reduce redundant attention computation and improve drafting quality. Across various multimodal benchmarks, MMFlash achieves lossless decoding speedups of up to 3.58times, outperforming EAGLE-3 and DFlash by up to 195.4% and 17.1%, respectively.
Paper page on Hugging Face · arXiv
Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.