Metraipolis › archive › dc_00031de61c8a
Putting Registers to Work: Task Registers for Token Pruning in Vision Transformers — arXiv:2608.10989
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.10989
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-19T06:09:35.976Z
- SHA-256 of the archived text
f124de78e59b62f6e79393f63c76c82bcc31bea56fb3eab1124294e2a55568de- Stable URL of this page
https://metraipolis.com/documento/dc_00031de61c8a
How to cite
Putting Registers to Work: Task Registers for Token Pruning in Vision Transformers — arXiv:2608.10989 — arXiv (en), https://arxiv.org/abs/2608.10989 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_00031de61c8a (Metraipolis archive dc_00031de61c8a, sha256 f124de78e59b62f6e79393f63c76c82bcc31bea56fb3eab1124294e2a55568de)
BibTeX
@misc{dc00031de61c8a,
title = {Putting Registers to Work: Task Registers for Token Pruning in Vision Transformers — arXiv:2608.10989},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.10989},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_00031de61c8a; sha256 f124de78e59b62f6e79393f63c76c82bcc31bea56fb3eab1124294e2a55568de},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.10989. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Token-pruning policies are usually designed for a single recognition pipeline, but pretrained Vision Transformers are reused across tasks with different spatial demands. We ask which parts of a pruning policy transfer across image classification, semantic segmentation, and object detection. For each pipeline, controlled probes freeze the no-pruning checkpoint and apply a series of parameter-free reduction criteria at one eligible layer at a time without retraining. The probes reveal three differences: segmentation and detection rank the criteria differently, classification is especially sensitive to attention-based pruning in the earliest layers, and the dense tasks prefer opposite recovery endpoints. These findings motivate Task-Adaptive Pruning (TAP). Existing register tokens serve as task-agnostic storage for feature artifacts. TAP instead introduces one task register per task and activates only the current one. Its evolving state ranks tokens, distributes an exact removal budget over depth, and sets the recovery scale for dense features. At a final keep rate of $\rho=0.5$, our jointly adapted model, TAP-J, reaches $47.0$ mIoU at $1.30\times$ encoder throughput on ADE20K and $53.7$ box AP at $1.32\times$ encoder throughput on COCO while remaining competitive on ImageNet-1K. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-12 in cs.AI). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2608.10989 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-19T05:57:28.421Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-12?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).