Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Zhang, Dingkun, Qi, Shuhan, Wu, Yulin, Xiao, Xinyu, Wang, Xuan, Chen, Long
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2602.03815
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912873669722112
author Zhang, Dingkun
Qi, Shuhan
Wu, Yulin
Xiao, Xinyu
Wang, Xuan
Chen, Long
author_facet Zhang, Dingkun
Qi, Shuhan
Wu, Yulin
Xiao, Xinyu
Wang, Xuan
Chen, Long
contents Multimodal Large Language Models (MLLMs) suffer from severe training inefficiency issue, which is associated with their massive model sizes and visual token numbers. Existing efforts in efficient training focus on reducing model sizes or trainable parameters. Inspired by the success of Visual Token Pruning (VTP) in improving inference efficiency, we are exploring another substantial research direction for efficient training by reducing visual tokens. However, applying VTP at the training stage results in a training-inference mismatch: pruning-trained models perform poorly when inferring on non-pruned full visual token sequences. To close this gap, we propose DualSpeed, a fast-slow framework for efficient training of MLLMs. The fast-mode is the primary mode, which incorporates existing VTP methods as plugins to reduce visual tokens, along with a mode isolator to isolate the model's behaviors. The slow-mode is the auxiliary mode, where the model is trained on full visual sequences to retain training-inference consistency. To boost its training, it further leverages self-distillation to learn from the sufficiently trained fast-mode. Together, DualSpeed can achieve both training efficiency and non-degraded performance. Experiments show DualSpeed accelerates the training of LLaVA-1.5 by 2.1$\times$ and LLaVA-NeXT by 4.0$\times$, retaining over 99% performance. Code: https://github.com/dingkun-zhang/DualSpeed
format Preprint
id arxiv_https___arxiv_org_abs_2602_03815
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Fast-Slow Efficient Training for Multimodal Large Language Models via Visual Token Pruning
Zhang, Dingkun
Qi, Shuhan
Wu, Yulin
Xiao, Xinyu
Wang, Xuan
Chen, Long
Computer Vision and Pattern Recognition
Machine Learning
Multimodal Large Language Models (MLLMs) suffer from severe training inefficiency issue, which is associated with their massive model sizes and visual token numbers. Existing efforts in efficient training focus on reducing model sizes or trainable parameters. Inspired by the success of Visual Token Pruning (VTP) in improving inference efficiency, we are exploring another substantial research direction for efficient training by reducing visual tokens. However, applying VTP at the training stage results in a training-inference mismatch: pruning-trained models perform poorly when inferring on non-pruned full visual token sequences. To close this gap, we propose DualSpeed, a fast-slow framework for efficient training of MLLMs. The fast-mode is the primary mode, which incorporates existing VTP methods as plugins to reduce visual tokens, along with a mode isolator to isolate the model's behaviors. The slow-mode is the auxiliary mode, where the model is trained on full visual sequences to retain training-inference consistency. To boost its training, it further leverages self-distillation to learn from the sufficiently trained fast-mode. Together, DualSpeed can achieve both training efficiency and non-degraded performance. Experiments show DualSpeed accelerates the training of LLaVA-1.5 by 2.1$\times$ and LLaVA-NeXT by 4.0$\times$, retaining over 99% performance. Code: https://github.com/dingkun-zhang/DualSpeed
title Fast-Slow Efficient Training for Multimodal Large Language Models via Visual Token Pruning
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2602.03815