通过球面约束提升连续令牌自回归图像生成质量
Hyperspherical Latents Improve Continuous-Token Autoregressive Generation
- 将自回归输入输出统一约束在固定半径超球面上，抑制方差崩溃
- ImageNet上943M模型达FID 1.34，超越同类模型和扩散模型
- 小模型也表现优异，适合追求高效生成的开发者
自回归（AR）模型在图像生成中前景广阔，但连续令牌版本常落后于潜在扩散和掩码生成模型。核心问题是VAE隐变量方差不均，在自回归解码时被放大，尤其在无分类器引导（CFG）下易导致方差坍缩。本文提出SphereAR，其核心设计是将所有自回归输入与输出（包括CFG后）约束在固定半径超球面（恒定ℓ₂范数）上，利用超球面VAE。理论分析表明，该约束消除了尺度分量（方差坍缩主因），从而稳定解码。实验显示，在ImageNet生成任务中，SphereAR-H（943M）以FID 1.34创下自回归模型新纪录。小型模型方面，SphereAR-L（479M）达FID 1.54，SphereAR-B（208M）达FID 1.92，性能匹配或超越更大规模基线如MAR-H（943M, FID 1.55）和VAR-d30（2B, FID 1.92）。据我们所知，这是首个在像素顺序下纯自回归图像生成器在可比参数规模上超越扩散与掩码生成模型的实例。
原文摘要 · Abstract (English)
Autoregressive (AR) models are promising for image generation, yet continuous-token AR variants often trail latent diffusion and masked-generation models. The core issue is heterogeneous variance in VAE latents, which is amplified during AR decoding, especially under classifier-free guidance (CFG), and can cause variance collapse. We propose SphereAR to address this issue. Its core design is to constrain all AR inputs and outputs -- including after CFG -- to lie on a fixed-radius hypersphere (constant $\ell_2$ norm), leveraging hyperspherical VAEs. Our theoretical analysis shows that hyperspherical constraint removes the scale component (the primary cause of variance collapse), thereby stabilizing AR decoding. Empirically, on ImageNet generation, SphereAR-H (943M) sets a new state of the art for AR models, achieving FID 1.34. Even at smaller scales, SphereAR-L (479M) reaches FID 1.54 and SphereAR-B (208M) reaches 1.92, matching or surpassing much larger baselines such as MAR-H (943M, 1.55) and VAR-d30 (2B, 1.92). To our knowledge, this is the first time a pure next-token AR image generator with raster order surpasses diffusion and masked-generation models at comparable parameter scales.
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。