MASSV: Multimodal Adaptation and Self-Data Distillation for Speculative Decoding of Vision-Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ganesan, Mugilan, Segal, Shane, Aggarwal, Ankur, Sinnadurai, Nish, Lie, Sean, Thangarasa, Vithursan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908368441966592
author Ganesan, Mugilan
Segal, Shane
Aggarwal, Ankur
Sinnadurai, Nish
Lie, Sean
Thangarasa, Vithursan
author_facet Ganesan, Mugilan
Segal, Shane
Aggarwal, Ankur
Sinnadurai, Nish
Lie, Sean
Thangarasa, Vithursan
contents Speculative decoding significantly accelerates language model inference by enabling a lightweight draft model to propose multiple tokens that a larger target model verifies simultaneously. However, applying this technique to vision-language models (VLMs) presents two fundamental challenges: small language models that could serve as efficient drafters lack the architectural components to process visual inputs, and their token predictions fail to match those of VLM target models that consider visual context. We introduce Multimodal Adaptation and Self-Data Distillation for Speculative Decoding of Vision-Language Models (MASSV), which transforms existing small language models into effective multimodal drafters through a two-phase approach. MASSV first connects the target VLM's vision encoder to the draft model via a lightweight trainable projector, then applies self-distilled visual instruction tuning using responses generated by the target VLM to align token predictions. Comprehensive experiments across the Qwen2.5-VL and Gemma3 model families demonstrate that MASSV increases accepted length by up to 30% and delivers end-to-end inference speedups of up to 1.46x on visually-grounded tasks. MASSV provides a scalable, architecture-compatible method for accelerating both current and future VLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2505_10526
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MASSV: Multimodal Adaptation and Self-Data Distillation for Speculative Decoding of Vision-Language Models
Ganesan, Mugilan
Segal, Shane
Aggarwal, Ankur
Sinnadurai, Nish
Lie, Sean
Thangarasa, Vithursan
Machine Learning
Computation and Language
Computer Vision and Pattern Recognition
Speculative decoding significantly accelerates language model inference by enabling a lightweight draft model to propose multiple tokens that a larger target model verifies simultaneously. However, applying this technique to vision-language models (VLMs) presents two fundamental challenges: small language models that could serve as efficient drafters lack the architectural components to process visual inputs, and their token predictions fail to match those of VLM target models that consider visual context. We introduce Multimodal Adaptation and Self-Data Distillation for Speculative Decoding of Vision-Language Models (MASSV), which transforms existing small language models into effective multimodal drafters through a two-phase approach. MASSV first connects the target VLM's vision encoder to the draft model via a lightweight trainable projector, then applies self-distilled visual instruction tuning using responses generated by the target VLM to align token predictions. Comprehensive experiments across the Qwen2.5-VL and Gemma3 model families demonstrate that MASSV increases accepted length by up to 30% and delivers end-to-end inference speedups of up to 1.46x on visually-grounded tasks. MASSV provides a scalable, architecture-compatible method for accelerating both current and future VLMs.
title MASSV: Multimodal Adaptation and Self-Data Distillation for Speculative Decoding of Vision-Language Models
topic Machine Learning
Computation and Language
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2505.10526