Purge-Gate: Backpropagation-Free Test-Time Adaptation for Point Clouds Classification via Token Purging
Abstract
Test-time adaptation (TTA) is crucial for mitigating performance degradation caused by distribution shifts in 3D point cloud classification. In this work, we introduce Token Purging (PG), a novel backpropagation-free approach that removes tokens highly affected by domain shifts before they reach attention layers. Unlike existing TTA methods, PG operates at the token level, ensuring robust adaptation without iterative updates. We propose two variants: PG-SP, which leverages source statistics, and PG-SF, a fully source-free version relying on CLS-token-driven adaptation. Extensive evaluations on ModelNet40-C, ShapeNet-C, and ScanObjectNN-C demonstrate that PG-SP achieves an average of +10.3% higher accuracy than state-of-the-art backpropagation-free methods, while PG-SF sets new benchmarks for source-free adaptation. Moreover, PG is 12.4x faster and 5.5x more memory efficient than our baseline, making it suitable for real-world deployment.
Cite
Text
Yazdanpanah et al. "Purge-Gate: Backpropagation-Free Test-Time Adaptation for Point Clouds Classification via Token Purging." International Conference on Computer Vision, 2025.Markdown
[Yazdanpanah et al. "Purge-Gate: Backpropagation-Free Test-Time Adaptation for Point Clouds Classification via Token Purging." International Conference on Computer Vision, 2025.](https://mlanthology.org/iccv/2025/yazdanpanah2025iccv-purgegate/)BibTeX
@inproceedings{yazdanpanah2025iccv-purgegate,
title = {{Purge-Gate: Backpropagation-Free Test-Time Adaptation for Point Clouds Classification via Token Purging}},
author = {Yazdanpanah, Moslem and Bahri, Ali and Noori, Mehrdad and Dastani, Sahar and Hakim, Gustavo Adolfo Vargas and Osowiechi, David and Ayed, Ismail Ben and Desrosiers, Christian},
booktitle = {International Conference on Computer Vision},
year = {2025},
pages = {27640-27649},
url = {https://mlanthology.org/iccv/2025/yazdanpanah2025iccv-purgegate/}
}