VLA Models and Efficiency

Vision-language-action modeling with stronger spatial understanding and more efficient inference.

This line of work studies how to make vision-language-action models more spatially aware while keeping them practical to deploy.

Representative public papers:

(Lin et al., 2025)
(missing reference)
(Liu et al., 2025)

The emphasis is on balancing semantic alignment, action capability, and computational efficiency.

References

2025

Evo-0: Vision-language-action model with implicit spatial understanding

T Lin, Gen Li, Y. Zhong, and 5 more authors

arXiv, 2025

Bib Paper

@article{lin2025evo0,
  title = {Evo-0: Vision-language-action model with implicit spatial understanding},
  author = {Lin, Tao and Li, Gen and Zhong, Y. and Zou, Y. and Du, Yuxin and Liu, J. and Gu, E. and Zhao, B.},
  journal = {arXiv preprint arXiv:2507.00416},
  year = {2025},
}

VLA-Pruner: Temporal-Aware Dual-Level Visual Token Pruning for Efficient Vision-Language-Action Inference

Z. Liu, Y. Chen, H. Cai, T Lin, and 3 more authors

arXiv, 2025

Bib Paper

@article{liu2025vlapruner,
  title = {VLA-Pruner: Temporal-Aware Dual-Level Visual Token Pruning for Efficient Vision-Language-Action Inference},
  author = {Liu, Z. and Chen, Y. and Cai, H. and Lin, Tao and Yang, S. and Liu, Z. and Zhao, B.},
  journal = {arXiv preprint arXiv:2511.16449},
  year = {2025},
}