From Inference Efficiency to Embodied Efficiency: Revisiting Efficiency Metrics for Vision-Language-Action Models

TL;DR

This paper argues that traditional efficiency metrics for vision-language-action models do not reflect real-world embodied performance, proposing system-level embodied efficiency metrics for better evaluation.

Contribution

It introduces system-level embodied efficiency metrics and demonstrates their importance over conventional metrics in evaluating VLA models.

Findings

Conventional metrics can misrepresent real-world efficiency.

Embodied efficiency metrics reveal hidden performance differences.

Trade-offs exist between computational savings and motion quality.

Abstract

Vision-Language-Action (VLA) models have recently enabled embodied agents to perform increasingly complex tasks by jointly reasoning over visual, linguistic, and motor modalities. However, we find that the prevailing notion of ``efficiency'' in current VLA research, characterized by parameters, FLOPs, or token decoding throughput, does not reflect actual performance on robotic platforms. In real-world execution, efficiency is determined by system-level embodied behaviors such as task completion time, trajectory smoothness, cumulative joint rotation, and motion energy. Through controlled studies across model compression, token sparsification, and action sequence compression, we make several observations that challenge common assumptions. (1) Methods that reduce computation under conventional metrics often increase end-to-end execution cost or degrade motion quality, despite maintaining task success rates. (2) System-level embodied efficiency metrics reveal performance differences in the learned action policies that remain hidden under conventional evaluations. (3) Common adaptation methods such as in-context prompting or supervised fine-tuning show only mild and metric-specific improvements in embodied efficiency. While these methods can reduce targeted embodied-efficiency metrics such as jerk or action rate, the resulting gains may come with trade-offs in other metrics, such as longer completion time. Taken together, our results suggest that conventional inference efficiency metrics can overlook important aspects of embodied execution. Incorporating embodied efficiency provides a more complete view of policy behavior and practical performance, enabling fairer and more comprehensive comparisons of VLA models.