Bridging the Semantic-Action Gap in Visual Token Pruning for Efficient VLA Inference

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liu, Ziyan, Chen, Yeqiu, Cai, Hongyi, Lin, Tao, Yang, Shuo, Liu, Zheng, Zhao, Bo
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916048313253888
author Liu, Ziyan
Chen, Yeqiu
Cai, Hongyi
Lin, Tao
Yang, Shuo
Liu, Zheng
Zhao, Bo
author_facet Liu, Ziyan
Chen, Yeqiu
Cai, Hongyi
Lin, Tao
Yang, Shuo
Liu, Zheng
Zhao, Bo
contents Vision-Language-Action (VLA) models have shown great potential for embodied AI by integrating visual perception, language understanding, and action execution. In real-time deployment, these models must process continuous visual streams, incurring substantial computational overhead. Visual token pruning -- a mainstream technique for accelerating Vision-Language Models (VLMs) by retaining salient tokens while discarding redundant ones -- offers a natural candidate solution to this challenge. However, directly applying VLM-oriented pruning methods to VLA inference can cause severe degradation in manipulation performance. Our analysis attributes this degradation to a key mismatch: VLA inference exhibits distinct attention patterns between the vision-language prefill stage and the action-decode stage, so pruning based only on context-prefill semantic salience is biased toward semantic cues and may remove action-critical visual tokens. Motivated by this observation, we propose VLA-Pruner, an effective plug-and-play token pruning method grounded in the visual requirements of VLA inference, further exploiting the temporal continuity of robot manipulation. Specifically, VLA-Pruner estimates visual-token importance from both semantic prefilling and temporally smoothed action relevance, and then applies a Combine-then-Filter strategy to retain compact, non-redundant tokens under the compute budget. Experiments show that VLA-Pruner outperforms state-of-the-art approaches across multiple VLA architectures, achieving up to 1.99x speedup with comparable manipulation quality.
format Preprint
id arxiv_https___arxiv_org_abs_2511_16449
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Bridging the Semantic-Action Gap in Visual Token Pruning for Efficient VLA Inference
Liu, Ziyan
Chen, Yeqiu
Cai, Hongyi
Lin, Tao
Yang, Shuo
Liu, Zheng
Zhao, Bo
Computer Vision and Pattern Recognition
Artificial Intelligence
Vision-Language-Action (VLA) models have shown great potential for embodied AI by integrating visual perception, language understanding, and action execution. In real-time deployment, these models must process continuous visual streams, incurring substantial computational overhead. Visual token pruning -- a mainstream technique for accelerating Vision-Language Models (VLMs) by retaining salient tokens while discarding redundant ones -- offers a natural candidate solution to this challenge. However, directly applying VLM-oriented pruning methods to VLA inference can cause severe degradation in manipulation performance. Our analysis attributes this degradation to a key mismatch: VLA inference exhibits distinct attention patterns between the vision-language prefill stage and the action-decode stage, so pruning based only on context-prefill semantic salience is biased toward semantic cues and may remove action-critical visual tokens. Motivated by this observation, we propose VLA-Pruner, an effective plug-and-play token pruning method grounded in the visual requirements of VLA inference, further exploiting the temporal continuity of robot manipulation. Specifically, VLA-Pruner estimates visual-token importance from both semantic prefilling and temporally smoothed action relevance, and then applies a Combine-then-Filter strategy to retain compact, non-redundant tokens under the compute budget. Experiments show that VLA-Pruner outperforms state-of-the-art approaches across multiple VLA architectures, achieving up to 1.99x speedup with comparable manipulation quality.
title Bridging the Semantic-Action Gap in Visual Token Pruning for Efficient VLA Inference
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2511.16449