arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2512.01643cs.CV
收藏

ViT$^3$:解锁视觉中的测试时间训练

ViT$^3$: Unlocking Test-Time Training in Vision

Tsinghua University(清华大学);Alibaba Group(阿里巴巴集团)

Dongchen Han, Yining Li, Tianyu Li, Zixuan Cao, Ziming Wang, Jun Song, Yu Cheng, Bo Zheng, Gao Huang

AI总结:

本文提出ViT$^3$,一种线性复杂度的纯测试时间训练模型,通过系统研究揭示了视觉序列建模中TTT设计的六个实用原则,并在多个视觉任务中验证其性能。

摘要:

测试时间训练(TTT)最近在高效序列建模中崭露头角。TTT将注意力操作重新表述为在线学习问题,在测试时从键值对构建紧凑的内部模型。这种重新表述开辟了丰富灵活的设计空间,同时实现了线性计算复杂度。然而,构建强大的视觉TTT设计仍然具有挑战性:内部模块和内部训练的基本选择缺乏全面理解和实用指南。为此,本文系统地研究了视觉序列建模中的TTT设计。通过一系列实验和分析,我们提炼出六个实用洞察,确立了有效视觉TTT的设计原则,并揭示了未来改进的路径。这些发现最终导致了视觉测试时间训练(ViT$^3$)模型,一种纯TTT架构,实现了线性复杂度和可并行计算。我们评估了ViT$^3$在多种视觉任务中的表现,包括图像分类、图像生成、目标检测和语义分割。结果表明,ViT$^3$在多个任务中均能与先进的线性复杂度模型(如Mamba和线性注意力变体)持平或超越,并有效缩小了与高度优化视觉Transformer之间的差距。我们希望本研究和ViT$^3$基线能促进未来视觉TTT模型的研究。代码:github.com/LeapLabTHU/ViTTT。

英文摘要:

Test-Time Training (TTT) has recently emerged as a promising direction for efficient sequence modeling. TTT reformulates attention operation as an online learning problem, constructing a compact inner model from key-value pairs at test time. This reformulation opens a rich and flexible design space while achieving linear computational complexity. However, crafting a powerful visual TTT design remains challenging: fundamental choices for the inner module and inner training lack comprehensive understanding and practical guidelines. To bridge this critical gap, in this paper, we present a systematic empirical study of TTT designs for visual sequence modeling. From a series of experiments and analyses, we distill six practical insights that establish design principles for effective visual TTT and illuminate paths for future improvement. These findings culminate in the Vision Test-Time Training (ViT$^3$) model, a pure TTT architecture that achieves linear complexity and parallelizable computation. We evaluate ViT$^3$ across diverse visual tasks, including image classification, image generation, object detection, and semantic segmentation. Results show that ViT$^3$ consistently matches or outperforms advanced linear-complexity models (e.g., Mamba and linear attention variants) and effectively narrows the gap to highly optimized vision Transformers. We hope this study and the ViT$^3$ baseline can facilitate future work on visual TTT models. Code: github.com/LeapLabTHU/ViTTT.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑