arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2503.02883cs.CV
收藏

ARINAR:双层自回归逐特征生成模型

ARINAR: Bi-Level Autoregressive Feature-by-Feature Generative Models

Australian National University(澳大利亚国立大学)

Qinyu Zhao, Stephen Gould, Liang Zheng

AI总结:

针对现有自回归图像生成模型在建模高维token分布时过于简单或速度慢的问题,提出双层自回归模型ARINAR,通过逐特征生成token并使用高斯混合模型建模单特征分布,在保持媲美MAR-B性能的同时实现五倍提速。

摘要:

现有的自回归(AR)图像生成模型采用逐token的生成模式。即,它们预测每个token的概率分布,并从该分布中采样下一个token。主要挑战在于如何建模高维token的复杂分布。先前的方法要么过于简单而无法拟合该分布,要么导致生成速度缓慢。我们没有去拟合整个token的分布,而是探索使用AR模型以逐特征的方式生成每个token,即将已生成的特征作为输入并生成下一个特征。基于此,我们提出了ARINAR(AR-in-AR),一种双层AR模型。外层AR层将先前的token作为输入,为下一个token预测一个条件向量z。内层在以z为条件的情形下,自回归地生成下一个token的特征。通过这种方式,内层只需对单个特征的分布进行建模,例如,使用简单的高斯混合模型。在ImageNet 256x256图像生成任务上,具有213M参数的ARINAR-B实现了2.75的FID,这与最先进的MAR-B模型(FID=2.31)相媲美,同时比后者快五倍。

英文摘要:

Existing autoregressive (AR) image generative models use a token-by-token generation schema. That is, they predict a per-token probability distribution and sample the next token from that distribution. The main challenge is how to model the complex distribution of high-dimensional tokens. Previous methods either are too simplistic to fit the distribution or result in slow generation speed. Instead of fitting the distribution of the whole tokens, we explore using a AR model to generate each token in a feature-by-feature way, i.e., taking the generated features as input and generating the next feature. Based on that, we propose ARINAR (AR-in-AR), a bi-level AR model. The outer AR layer take previous tokens as input, predicts a condition vector z for the next token. The inner layer, conditional on z, generates features of the next token autoregressively. In this way, the inner layer only needs to model the distribution of a single feature, for example, using a simple Gaussian Mixture Model. On the ImageNet 256x256 image generation task, ARINAR-B with 213M parameters achieves an FID of 2.75, which is comparable to the state-of-the-art MAR-B model (FID=2.31), while five times faster than the latter.

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。

↑