📚 Introduction

We propose Task-Relevant Parameter and Token Selection (TR-PTS), a novel framework that unifies task-driven parameter selection and token refinement. We evaluate TR-PTS on benchmark datasets, including FGVC and VTAB-1k, where it achieves state-of-the-art performance, surpassing full fine- tuning by 3.40% and 10.35%, respectively.

🔥 News

[2025-06-26] 🎉🎉🎉 Our TR-PTS is accepted by ICCV 2025! 🎉🎉🎉

📝 Open-source Plan

• Training & Inference code
• Paper

🚀 Quick Start

Coming Soon...

📖 BibTeX

@article{luo2025trpts,
  title={TR-PTS: Task-Relevant Parameter and Token Selection for Efficient Tuning},
  author={Luo, Siqi and Yang, Haoran and Xin, Yi and Yi, Mingyang and Wu, Guangyang and Zhai, Guangtao and Liu, xiaohong},
  journal={Proceedings of the IEEE International Conference on Computer Vision (ICCV)},
  year={2025}
}