This is a curated list of "A Survey on Efficient Vision-Language Action Models" research.

To the best of our knowledge, this work presents the first comprehensive survey specifically dedicated to the realm of Efficient VLAs that covers the entire "data-model-training" process. We will continue to UPDATE this repository to provide you with the latest cutting-edge developments, so stay tuned!😘 We hope that our work will bring some inspiration to you~😉

News 🔥

• 2026.09.10: Our paper has been accepted by IEEE TPAMI！
• 2026.08.23: Several recent papers on efficient VLA have been updated
• 2026.06.18: Refreshed the paper lists and added recent robot-efficient-VLA papers, with an emphasis on model efficiency, adaptation efficiency, and data efficiency for robotic manipulation.
• 2025.09.12: This repository was initialized.

Overview

Fig. 1: The Organization of Our Survey. We systematically categorize efficient VLAs into three core pillars: (1) Efficient Model Design, encompassing efficient architectures and model compression techniques; (2) Efficient Training, covering efficient pre-training and post-training strategies; and (3) Efficient Data Collection, including efficient data collection and augmentation methods. The framework also reviews VLA foundations, key applications, challenges, and future directions, establishing the groundwork for advancing scalable embodied intelligence.

Table of Contents

• 🔍 Table of Contents
• 🚀Efficient Model Design
  • Efficient Architectures
    • Efficient Attention
    • Transformer Alternatives
    • Efficient Action Decoding
    • Lightweight Component
    • Mixture-of-Experts
    • Hierarchical Systems
  • Model Compression
    • Layer Pruning
    • Quantization
    • Token Optimization
• 🤖Efficient Training
  • Efficient Pre-Training
    • Data-Efficient Pre-training
    • Efficient Action Representation
    • Other Pre-training Strategies
  • Efficient Post-Training
    • Supervised Fine-tuning
    • RL-Based Method
• 📖Efficient Data Collection
  • Human-in-the-Loop Data Collection
  • Simulation Data Collection
  • Internet-Scale and Cross-Domain Data Utilization
  • Self-Exploration Data Collection
  • Data Augmentation
• 🔖 Citation
• 📧 Contact Us

Efficient VLAs

Efficient Model Design

Efficient Architectures

Fig. 2: Key strategies for Efficient Architectures in VLAs. We illustrate six primary approaches: (a) Efficient Attention, mitigating the O(n^2) complexity of standard self-attention; (b) Transformer Alternatives, such as Mamba; (c) Efficient Action Decoding, advancing from autoregressive generation to parallel and generative methods; (d) Lightweight Components, adopting smaller model backbones; (e) Mixture-of-Experts, employing sparse activation via input routing; and (f) Hierarchical Systems, which decouple high-level VLM planning from low-level VLA execution.

Efficient Attention

Transformer Alternatives

Efficient Action Decoding

Lightweight Component

Mixture-of-Experts

Hierarchical Systems

Model Compression

Fig. 3: Key strategies for Model Compression in VLAs. We illustrate three primary approaches: (a) Layer Pruning, which removes redundant layers to reduce model depth and computational cost; (b) Quantization, which reduces the numerical precision of model parameters to decrease memory footprint and accelerate inference; and (c) Token Optimization, which minimizes the number of processed tokens via token compression (merging tokens), token pruning (dropping non-essential tokens), and token caching (reusing static tokens).

Layer Pruning

Quantization

Token Optimization

Efficient Training

Fig. 4: Key strategies for Efficient Training in VLAs, divided into two main stages. (a) Efficient Pre-Training migrates general-purpose VLMs into the embodied domain to create an initial, action-aware policy, encompassing Data-Efficient Pre-training, Efficient Action Representation, and Other Pre-training Strategies. (b) Efficient Post-Training subsequently specializes this policy for specific tasks, leveraging Supervised Fine-tuning and RL-Based Methods.

Efficient Pre-Training

Data-Efficient Pre-training

Efficient Action Representation

Other Pre-training Strategies

Efficient Post-Training

Supervised Fine-tuning

RL-Based Method

Efficient Data Collection

Fig. 5: Taxonomy of Efficient Data Collection strategies in VLAs. This figure illustrates the primary approaches, encompassing human-in-the-loop, simulated, reusability-oriented, self-driven, and augmentative techniques for scalable acquisition of high-quality robotic datasets while minimizing resource overhead.

Human-in-the-Loop Data Collection

Simulation Data Collection

Internet-Scale and Cross-Domain Data Utilization

Self-Exploration Data Collection

Data Augmentation

Citation

If you find this survey helpful for your research or applications, please consider citing it using the following BibTeX entry:

@article{yu2025survey,
  title={A survey on efficient vision-language-action models},
  author={Yu, Zhaoshu and Wang, Bo and Zeng, Pengpeng and Zhang, Haonan and Zhang, Ji and Wang, Zheng and Gao, Lianli and Song, Jingkuan and Sebe, Nicu and Shen, Heng Tao},
  journal={arXiv preprint arXiv:2510.24795},
  year={2025}
}

Contact Us

For any questions or suggestions, please feel free to contact us at:

Email: yuzhaoshu@gmail.com

Star History