VLA-IAP: Training-Free Visual Token Pruning via Interaction Alignment for Vision-Language-Action Models

2026-03-24

AI Summary

This work addresses the high inference cost of existing vision-language-action (VLA) models on resource-constrained platforms, which stems from their large number of visual tokens. Conventional pruning methods often neglect the physical interaction characteristics inherent in robotic tasks, inadvertently removing structurally critical regions and causing early-stage behavioral instability. To overcome this, the authors propose a training-free, dynamic visual token pruning approach that introduces an “interaction-first” paradigm by explicitly modeling physical interactions as the pruning criterion. The method leverages geometric priors to preserve structural anchor points and adaptively modulates pruning intensity based on a semantic-motion alignment metric, enabling a scheduling strategy that transitions from conservative to aggressive pruning. This plug-and-play technique requires no fine-tuning, achieves a 97.8% success rate on the LIBERO benchmark, accelerates inference by up to 1.54× (1.25× on average), and demonstrates consistent performance across multiple models and both simulated and real-world robotic platforms.

Technology Category

• Computer Vision: Large Vision Models
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Vision-Language-Action (VLA) models have rapidly advanced embodied intelligence, enabling robots to execute complex, instruction-driven tasks. However, as model capacity and visual context length grow, the inference cost of VLA systems becomes a major bottleneck for real-world deployment on resource-constrained platforms. Existing visual token pruning methods mainly rely on semantic saliency or simple temporal cues, overlooking the continuous physical interaction, a fundamental property of VLA tasks. Consequently, current approaches often prune visually sparse yet structurally critical regions that support manipulation, leading to unstable behavior during early task phases. To overcome this, we propose a shift toward an explicit Interaction-First paradigm. Our proposed \textbf{training-free} method, VLA-IAP (Interaction-Aligned Pruning), introduces a geometric prior mechanism to preserve structural anchors and a dynamic scheduling strategy that adapts pruning intensity based on semantic-motion alignment. This enables a conservative-to-aggressive transition, ensuring robustness during early uncertainty and efficiency once interaction is locked. Extensive experiments show that VLA-IAP achieves a \textbf{97.8\% success rate} with a \textbf{$1.25\times$ speedup} on the LIBERO benchmark, and up to \textbf{$1.54\times$ speedup} while maintaining performance \textbf{comparable to the unpruned backbone}. Moreover, the method demonstrates superior and consistent performance across multiple model architectures and three different simulation environments, as well as a real robot platform, validating its strong generalization capability and practical applicability. Our project website is: \href{https://chengjt1999.github.io/VLA-IAP.github.io/}{VLA-IAP.com}.

Problem

• Vision-Language-Action models
• visual token pruning
• embodied intelligence
• inference efficiency
• physical interaction

Innovation

• training-free pruning
• interaction alignment
• vision-language-action models
• geometric prior
• dynamic scheduling