Research

Paper

TESTING March 04, 2026

ZipMap: Linear-Time Stateful 3D Reconstruction with Test-Time Training

Authors

Haian Jin, Rundi Wu, Tianyuan Zhang, Ruiqi Gao, Jonathan T. Barron, Noah Snavely, Aleksander Holynski

Abstract

Feed-forward transformer models have driven rapid progress in 3D vision, but state-of-the-art methods such as VGGT and $π^3$ have a computational cost that scales quadratically with the number of input images, making them inefficient when applied to large image collections. Sequential-reconstruction approaches reduce this cost but sacrifice reconstruction quality. We introduce ZipMap, a stateful feed-forward model that achieves linear-time, bidirectional 3D reconstruction while matching or surpassing the accuracy of quadratic-time methods. ZipMap employs test-time training layers to zip an entire image collection into a compact hidden scene state in a single forward pass, enabling reconstruction of over 700 frames in under 10 seconds on a single H100 GPU, more than $20\times$ faster than state-of-the-art methods such as VGGT. Moreover, we demonstrate the benefits of having a stateful representation in real-time scene-state querying and its extension to sequential streaming reconstruction.

Metadata

arXiv ID: 2603.04385
Provider: ARXIV
Primary Category: cs.CV
Published: 2026-03-04
Fetched: 2026-03-05 06:06

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.04385v1</id>\n    <title>ZipMap: Linear-Time Stateful 3D Reconstruction with Test-Time Training</title>\n    <updated>2026-03-04T18:49:37Z</updated>\n    <link href='https://arxiv.org/abs/2603.04385v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.04385v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Feed-forward transformer models have driven rapid progress in 3D vision, but state-of-the-art methods such as VGGT and $π^3$ have a computational cost that scales quadratically with the number of input images, making them inefficient when applied to large image collections. Sequential-reconstruction approaches reduce this cost but sacrifice reconstruction quality. We introduce ZipMap, a stateful feed-forward model that achieves linear-time, bidirectional 3D reconstruction while matching or surpassing the accuracy of quadratic-time methods. ZipMap employs test-time training layers to zip an entire image collection into a compact hidden scene state in a single forward pass, enabling reconstruction of over 700 frames in under 10 seconds on a single H100 GPU, more than $20\\times$ faster than state-of-the-art methods such as VGGT. Moreover, we demonstrate the benefits of having a stateful representation in real-time scene-state querying and its extension to sequential streaming reconstruction.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CV'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <published>2026-03-04T18:49:37Z</published>\n    <arxiv:comment>Project page: https://haian-jin.github.io/ZipMap</arxiv:comment>\n    <arxiv:primary_category term='cs.CV'/>\n    <author>\n      <name>Haian Jin</name>\n    </author>\n    <author>\n      <name>Rundi Wu</name>\n    </author>\n    <author>\n      <name>Tianyuan Zhang</name>\n    </author>\n    <author>\n      <name>Ruiqi Gao</name>\n    </author>\n    <author>\n      <name>Jonathan T. Barron</name>\n    </author>\n    <author>\n      <name>Noah Snavely</name>\n    </author>\n    <author>\n      <name>Aleksander Holynski</name>\n    </author>\n  </entry>"
}