Paper Detail
Shuhong Zheng, Michael Oechsle, Erik Sandström, Marie-Julie Rakotosaona, Federico Tombari, Igor Gilitschenski
Visual geometry transformers have become powerful architectures for multi-view 3D reconstruction, enabling joint prediction of multiple 3D attributes in a feed-forward manner. However, their computational cost grows quadratically with the input sequence length due to the global attention layers inside these models. This limits both their scalability and efficiency. In this work, we address this challenge with a simple yet general strategy: restricting the number of key/value tokens that each query interacts with during global attention. To achieve effective token selection, we introduce a two-stage framework. First, an inter-frame selection step operates at the frame level to identify frames that should be preserved. Second, an intra-frame selection step further discards more redundant tokens within the selected frames. Our analysis highlights the advantage of a diversity-based strategy for inter-frame selection, which ensures broad coverage of the scene. For intra-frame selection, we show that layer-aware sparsification is necessary, with the selection process guided by the entropy of the global attention pattern. Our approach offers a superior speed-accuracy trade-off compared to existing solutions. Extensive experiments show that it accelerates visual geometry transformers by over 85% for scenes with 500 images while maintaining, or even improving, baseline performance, which hints that how our token selection strategy can play a crucial role in future applications of visual geometry transformers. Our project website is available at https://zsh2000.github.io/good-token-hunting.github.io.
No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.
No ranking explanation is available yet.
No tags.
@article{zheng2026good,
title = {Good Token Hunting: A Hitchhiker's Guide to Token Selection for Visual Geometry Transformers},
author = {Shuhong Zheng and Michael Oechsle and Erik Sandström and Marie-Julie Rakotosaona and Federico Tombari and Igor Gilitschenski},
year = {2026},
abstract = {Visual geometry transformers have become powerful architectures for multi-view 3D reconstruction, enabling joint prediction of multiple 3D attributes in a feed-forward manner. However, their computational cost grows quadratically with the input sequence length due to the global attention layers inside these models. This limits both their scalability and efficiency. In this work, we address this challenge with a simple yet general strategy: restricting the number of key/value tokens that each que},
url = {https://arxiv.org/abs/2605.23892},
keywords = {cs.CV, cs.AI, cs.GR, cs.LG, cs.RO, visual geometry transformers, multi-view 3D reconstruction, global attention layers, token selection, inter-frame selection, intra-frame selection, layer-aware sparsification, entropy of global attention pattern, code available, huggingface daily},
eprint = {2605.23892},
archiveprefix = {arXiv},
}
{}