跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.21839 · v1

VisHead-Spec: Head-Routed Visual Attention for Multimodal Speculative Decoding

首次提交 2026-11-17 03:57
版本更新 2026-11-17 03:57
雷达观测 2026-11-18 19:44
arXiv ↗PDF ↗

摘要

Speculative decoding accelerates vision-language models(VLMs) while preserving the target model's outputs, but its speedup diminishes under high concurrency serving as the system becomes more compute bounded as the batch size increases. This problem is pronounced for multimodal speculative decoding as the long range of the image tokens in the prompt can significantly increase the computational cost during draft token prediction. The block-diffusion drafters (e.g. DFlash, DSpark) can effectively speed up the draft model efficiency by predicting multiple draft tokens in one forward pass, but their speedup for VLMs under high concurrency is still modest (<9%). In order to understand this gap, we examine the information flow between different modalities from target model to draft model and identify the sparsity pattern of visual attention between the target and draft models: only small set of key-value heads in the draft model need direct visual feature attention in the context and only visual features from top level layers in the target model need to be attended in the context. We therefore propose VisHead-Spec, which selects a small, fixed set of key-value (KV) heads to attend to all image features from top level layers in the target model and restricts the remaining heads to text features. On Qwen3-VL-8B and MMSpec, VisHead-Spec-8 lets 8 of the drafter's 40 KV heads read the image while keeping accepted length within 0.3% of the block-diffusion draft model baseline (i.e. DSpark). At batch size 64, its throughput gain over autoregressive decoding increases to 16.0%, compared with 9.5% for DSpark; at batch size 1, its speedup remains comparable to DSpark.

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-17 03:57 · 最新观测版本

VisHead-Spec: Head-Routed Visual Attention for Multimodal Speculative Decoding

Speculative decoding accelerates vision-language models(VLMs) while preserving the target model's outputs, but its speedup diminishes under high concurrency serving as the system becomes more compute bounded as the batch size increases. This problem is pronounced for multimodal speculative decoding as the long range of the image tokens in the prompt can significantly increase the computational cost during draft token prediction. The block-diffusion drafters (e.g. DFlash, DSpark) can effectively speed up the draft model efficiency by predicting multiple draft tokens in one forward pass, but their speedup for VLMs under high concurrency is still modest (<9%). In order to understand this gap, we examine the information flow between different modalities from target model to draft model and identify the sparsity pattern of visual attention between the target and draft models: only small set of key-value heads in the draft model need direct visual feature attention in the context and only visual features from top level layers in the target model need to be attended in the context. We therefore propose VisHead-Spec, which selects a small, fixed set of key-value (KV) heads to attend to all image features from top level layers in the target model and restricts the remaining heads to text features. On Qwen3-VL-8B and MMSpec, VisHead-Spec-8 lets 8 of the drafter's 40 KV heads read the image while keeping accepted length within 0.3% of the block-diffusion draft model baseline (i.e. DSpark). At batch size 64, its throughput gain over autoregressive decoding increases to 16.0%, compared with 9.5% for DSpark; at batch size 1, its speedup remains comparable to DSpark.

查看此版本解读 →查看此版本原文 ↗