SpecPrune-VLA: Accelerating Vision-Language-Action Models via Action-Aware Self-Speculative Pruning

Pruning is a typical acceleration technique for compute-bound models by removing computation on unimportant values. Recently, it has been applied to accelerate Vision-Language-Action (VLA) model inference. However, existing acceleration methods focus on local information from the current action step and ignore the global context, leading to $>$20\% success rate drop and limited speedup in some scenarios. In this paper, we point out \textbf{spatial-temporal consistency} in VLA tasks: input images in consecutive steps exhibit high similarity, and propose the key insight that token selection should combine local information with global context of the model. Based on this, we propose ***SpecPrune-VLA***, a training-free, two-level pruning method with heuristic control. **(1) Action-level static pruning.** We leverage global history and local attention to statically reduce visual tokens per action. **(2) Layer-level dynamic pruning.** We prune tokens adaptively per layer based on layer-wise importance. **(3) Lightweight action-aware controller:** We classify actions as coarse- or fine-grained by the speed of the end effector and adjust pruning aggressiveness accordingly. Extensive experiments show that SpecPrune-VLA achieves up to 1.57× speedup in LIBERO simulation and 1.70× on real-world tasks, with negligible success rate degradation.

(1) Problem: Vision-Language-Action (VLA) models are computationally heavy, hindering real-time robotic control on standard hardware. Existing acceleration methods often compromise accuracy or ignore task-specific patterns. (2) Solution: We propose SpecPrune-VLA, a training-free method that accelerates inference by pruning redundant visual data. It leverages temporal consistency across frames and adapts computation intensity based on action granularity (coarse vs. fine movements). (3) Impact: Our method achieves up to 1.70× speedup in real-world tasks with negligible accuracy loss. This enables efficient, real-time deployment of advanced VLA models on accessible hardware.