超越下一个Token：自回归视觉生成的Next-X预测
Beyond Next-Token: Next-X Prediction for Autoregressive Visual Generation
AI总结：
该研究提出通用自回归框架xAR，将预测单元从Token扩展为多粒度实体X，结合流匹配与带噪上下文学习缓解暴露偏差，在ImageNet-256基准上实现了更优生成效果与更快推理速度。
摘要：
自回归（AR）建模以下一个Token预测范式为核心，是当前最先进的语言和视觉生成模型的基础。传统上，“Token”被视为最小预测单元，通常是语言中的离散符号或视觉中的量化图像块。然而，针对2D图像结构的最优Token定义仍是一个悬而未决的问题。此外，自回归模型存在暴露偏差问题，即训练期间的教师强制会导致推理时的误差累积。\n本文提出xAR，这是一个通用的自回归框架，将Token的概念扩展为实体X，X可以表示单个图像块Token、一个单元（由k×k个相邻图像块组成的分组）、一个子样本（由远距离图像块组成的非局部分组）、一个尺度（从粗到细的分辨率），甚至是整幅图像。此外，我们将离散Token分类重新表述为连续实体回归，在每个自回归步骤中利用流匹配方法。该方法以带噪实体而非真实Token为训练条件，由此产生的带噪上下文学习（Noisy Context Learning）有效缓解了暴露偏差。\n因此，xAR具备两大关键优势：（1）它支持灵活的预测单元，能够捕捉不同的上下文粒度和空间结构；（2）它通过避免依赖教师强制来缓解暴露偏差。在ImageNet-256生成基准测试中，我们的基础模型xAR-B（1.72亿参数）性能优于DiT-XL/SiT-XL（6.75亿参数），同时推理速度快20倍。与此同时，xAR-H创下了1.24的FID新纪录，推理速度比此前性能最佳的模型快2.2倍，且无需依赖视觉基础模块（如DINOv2）或先进的引导间隔采样技术。
英文摘要：
Autoregressive (AR) modeling, known for its next-token prediction paradigm, underpins state-of-the-art language and visual generative models. Traditionally, a ``token'' is treated as the smallest prediction unit, often a discrete symbol in language or a quantized patch in vision. However, the optimal token definition for 2D image structures remains an open question. Moreover, AR models suffer from exposure bias, where teacher forcing during training leads to error accumulation at inference. In this paper, we propose xAR, a generalized AR framework that extends the notion of a token to an entity X, which can represent an individual patch token, a cell (a $k\times k$ grouping of neighboring patches), a subsample (a non-local grouping of distant patches), a scale (coarse-to-fine resolution), or even a whole image. Additionally, we reformulate discrete token classification as continuous entity regression, leveraging flow-matching methods at each AR step. This approach conditions training on noisy entities instead of ground truth tokens, leading to Noisy Context Learning, which effectively alleviates exposure bias. As a result, xAR offers two key advantages: (1) it enables flexible prediction units that capture different contextual granularity and spatial structures, and (2) it mitigates exposure bias by avoiding reliance on teacher forcing. On ImageNet-256 generation benchmark, our base model, xAR-B (172M), outperforms DiT-XL/SiT-XL (675M) while achieving 20$\times$ faster inference. Meanwhile, xAR-H sets a new state-of-the-art with an FID of 1.24, running 2.2$\times$ faster than the previous best-performing model without relying on vision foundation modules (e.g., DINOv2) or advanced guidance interval sampling.
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。
