arXiv 2504.104342025-08-09cs.CV

Anchor Token Matching：用于免训练 AR 图像编辑的隐式结构锁定

Anchor Token Matching: Implicit Structure Locking for Training-free AR Image Editing

VCIP, College of Computer Science, Nankai University（VCIP，计算机科学学院，南开大学）；NKIARI, Shenzhen Futian（NKIARI，深圳福田）；Computer Vision Center, Universitat Autònoma de Barcelona（计算机视觉中心，巴塞罗那自治大学）；City University of Hong Kong (Dongguan)（香港城市大学（东莞））

Taihang Hu, Linxuan Li, Kai Wang, Yaxing Wang, Jian Yang, Ming-Ming Cheng

备注：: Accepted by ICCV2025. Code will be released in https://github.com/hutaiHang/ATM

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

针对自回归模型图像编辑中注意力空间稀疏与结构误差累积问题，提出首个免训练编辑策略 ISLock，通过 Anchor Token Matching 协议在隐空间动态对齐自注意力以锁定结构，实现高质量结构感知编辑。

摘要：

文本到图像生成在扩散模型推动下取得突破性进展，通过交叉注意力操控实现了高保真合成与精确图像编辑。近期，自回归（AR）模型作为强大替代方案重新崛起，利用下一 token 生成来媲美扩散模型。然而，由于结构控制上的根本差异，为扩散模型设计的现有编辑技术无法直接应用于 AR 模型。具体而言，AR 模型在图像编辑时面临注意力图空间稀疏以及结构误差的序列累积问题，破坏了物体布局与全局一致性。本研究提出 ISLock，首个面向 AR 视觉模型的免训练编辑策略。ISLock 不依赖显式注意力操控或微调，而是通过 Anchor Token Matching（ATM）协议将自注意力模式与参考图像动态对齐，从而保留结构蓝图。该方法通过在隐空间隐式强制结构一致性，实现了结构感知编辑并保持生成自主性。大量实验表明，ISLock 无需额外训练即可实现高质量、结构一致的编辑，且优于或媲美常规编辑技术。本研究为高效灵活的 AR 图像编辑开辟了道路，进一步弥合了扩散与自回归生成模型间的性能差距。代码将公开于 https://github.com/hutaiHang/ATM

英文摘要：

Text-to-image generation has seen groundbreaking advancements with diffusion models, enabling high-fidelity synthesis and precise image editing through cross-attention manipulation. Recently, autoregressive (AR) models have re-emerged as powerful alternatives, leveraging next-token generation to match diffusion models. However, existing editing techniques designed for diffusion models fail to translate directly to AR models due to fundamental differences in structural control. Specifically, AR models suffer from spatial poverty of attention maps and sequential accumulation of structural errors during image editing, which disrupt object layouts and global consistency. In this work, we introduce Implicit Structure Locking (ISLock), the first training-free editing strategy for AR visual models. Rather than relying on explicit attention manipulation or fine-tuning, ISLock preserves structural blueprints by dynamically aligning self-attention patterns with reference images through the Anchor Token Matching (ATM) protocol. By implicitly enforcing structural consistency in latent space, our method ISLock enables structure-aware editing while maintaining generative autonomy. Extensive experiments demonstrate that ISLock achieves high-quality, structure-consistent edits without additional training and is superior or comparable to conventional editing techniques. Our findings pioneer the way for efficient and flexible AR-based image editing, further bridging the performance gap between diffusion and autoregressive generative models. The code will be publicly available at https://github.com/hutaiHang/ATM

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。