用无序令牌集动态编码图像，提升生成质量与鲁棒性
Tokenize Image as a Set
- 图像以无序令牌集形式表示，按语义复杂度分配编码资源
- 首次实现离散值、固定长度和求和不变性的统一建模
- 适合追求高质量图像生成与结构化表示的研究者
本文提出一种基于集合的图像生成新范式，通过无序令牌集表示动态分配编码容量，依据区域语义复杂度调整压缩策略。传统方法将图像序列化为固定位置的潜在码且压缩率均一，而本方法显著增强全局上下文聚合能力，并提升对局部扰动的鲁棒性。针对离散集合建模难题，提出双变换机制，可双射地将集合映射为带求和约束的固定长度整数序列。进一步设计了首个同时满足离散值、固定序列长度与求和不变性的固定和离散扩散框架，实现高效集合分布建模。实验表明，该方法在语义感知表示与生成质量上均优于现有方法。本工作在表示与建模策略上取得双重创新，推动视觉生成超越传统序列化令牌范式。代码与模型已公开于 https://github.com/Gengzigang/TokenSet。
原文摘要 · Abstract (English)
This paper proposes a fundamentally new paradigm for image generation through set-based tokenization and distribution modeling. Unlike conventional methods that serialize images into fixed-position latent codes with a uniform compression ratio, we introduce an unordered token set representation to dynamically allocate coding capacity based on regional semantic complexity. This TokenSet enhances global context aggregation and improves robustness against local perturbations. To address the critical challenge of modeling discrete sets, we devise a dual transformation mechanism that bijectively converts sets into fixed-length integer sequences with summation constraints. Further, we propose Fixed-Sum Discrete Diffusion--the first framework to simultaneously handle discrete values, fixed sequence length, and summation invariance--enabling effective set distribution modeling. Experiments demonstrate our method's superiority in semantic-aware representation and generation quality. Our innovations, spanning novel representation and modeling strategies, advance visual generation beyond traditional sequential token paradigms. Our code and models are publicly available at https://github.com/Gengzigang/TokenSet.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。