analiza aiqbench
SAPrune introduces a stage-aware pruning strategy for Vision-Language-Action models, dynamically selecting layers based on evolving attention patterns rather than fixed schedules. By protecting critical visual tokens and context until the model's action-relevant focus stabilizes, this method achieves significant inference speedups without sacrificing task success rates in robotic applications.
When and What to Prune? Stage-Aware Visual Token Pruning for Efficient VLA
arXiv:2610.05273v1 Announce Type: cross Abstract: Visual token pruning is an effective way to accelerate vision-language models and is especially useful for vision-language-action (VLA) inference, where many visual tokens must be processed before predicting robot actions. Existing pruning methods usually estimate which tokens can be pruned based on attention scores or feature diversity, retaining tokens that are either highly attended or visually different from others. However, most of them use fixed pruning schedules, such as pruning once at a preset layer or pruning at uniformly spaced layers. Such schedules can be risky for VLA models, because the model may not know which visual regions matter for the action in early layers. Tokens that look unimportant at first may become useful after the model combines visual observations with the language instruction. In this work, we propose SAPrune, a training-free visual token pruning framework for efficient VLA inference. Instead of pruning at fixed layers, SAPrune uses a small calibration set to observe how action-to-visual attention changes across layers, and chooses pruning layers only after the attention pattern becomes more reliable. At each selected layer, SAPrune applies a dual-path pruning rule: one path protects strongly attended visual tokens from pruning, while the other prevents useful surrounding context from being discarded. Experiments on LIBERO, SIMPLER, and real-world robotic tasks show that SAPrune prunes 87.5% of visual tokens and achieves up to 1.718x inference speedup while maintaining competitive task success rates.
Title:When and What to Prune? Stage-Aware Visual Token Pruning for Efficient VLA
Abstract:Visual token pruning is an effective way to accelerate vision-language models and is especially useful for vision-language-action (VLA) inference, where many visual tokens must be processed before predicting robot actions. Existing pruning methods usually estimate which tokens can be pruned based on attention scores or feature diversity, retaining tokens that are either highly attended or visually different from others. However, most of them use fixed pruning schedules, such as pruning once at a preset layer or pruning at uniformly spaced layers. Such schedules can be risky for VLA models, because the model may not know which visual regions matter for the action in early layers. Tokens that look unimportant at first may become useful after the model combines visual observations with the language instruction. In this work, we propose SAPrune, a training-free visual token pruning framework for efficient VLA inference. Instead of pruning at fixed layers, SAPrune uses a small calibration set to observe how action-to-visual attention changes across layers, and chooses pruning layers only after the attention pattern becomes more reliable. At each selected layer, SAPrune applies a dual-path pruning rule: one path protects strongly attended visual tokens from pruning, while the other prevents useful surrounding context from being discarded. Experiments on LIBERO, SIMPLER, and real-world robotic tasks show that SAPrune prunes 87.5% of visual tokens and achieves up to 1.718x inference speedup while maintaining competitive task success rates.
Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
arXiv:2610.05273 [cs.CV]
(or arXiv:2610.05273v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2610.05273 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 341 cuvinte
Subiecte conexe
When Attention Collapses: Stage-Aware Visual Token Pruning from Structure to Semantics
arXiv · 3 iun. 2026
[AINews] not much happened today
Latent Space · 10 sept. 2026
DCP-Prune: Ultra-Low Token Pruning with Distribution Consistency Preservation
arXiv · 16 iun. 2026
SLICEChat: Progressive In-Encoder Token Pruning for Whole-Slide Pathology Language Models
arXiv · 22 sept. 2026