الملخص
Representation Autoencoders (RAEs) enable diffusion models to operate in the feature spaces of pretrained visual encoders. However, many off-the-shelf encoders are not optimized for faithful reconstruction, discarding fine-grained visual details. As expected, finetuning these encoders for image reconstruction recovers such details. However, perhaps counterintuitively, this procedure reduces the effective dimensionality of the resulting representation, and the altered geometry has downstream effects on generation. Specifically, we show that using the standard velocity prediction in flow matching in this high-dimensional space requires the model to fit orthogonal noise directions outside the low-dimensional signal manifold, making optimization inefficient. This motivates using the clean data parameterization ($\boldsymbol{x}_{0}$-prediction) instead, which focuses learning on the underlying signal manifold. Across experiments with multiple strong-reconstruction encoders, we show that $\boldsymbol{x}_{0}$-prediction consistently improves text-to-image generation performance.
الكلمات المفتاحية
الموضوع
بيانات النشر
- المجلة
- غير متاح
- وصول مفتوح
- وصول مفتوح أخضر
اقتبس هذه المقالة
APA 7
Feng, C., Xu, Z., Chen, B., Xiong, Y., Wang, X., Wang, J. H., Zhang, R., Lin, Z., Owens, A., & Li, Y. (2026). On the Diffusibility of High-Dimensional Latents. https://omanscience.com/ar/articles/on-the-diffusibility-of-high-dimensional-latents
MLA 9
Feng, Chao, et al. "On the Diffusibility of High-Dimensional Latents." https://omanscience.com/ar/articles/on-the-diffusibility-of-high-dimensional-latents.
شيكاغو (المؤلف–التاريخ)
Feng, Chao, Zhiyang Xu, Bowei Chen, Yuanjun Xiong, Xiyao Wang, Jui-Hsien Wang, Richard Zhang, Zhe Lin, Andrew Owens, and Yijun Li. 2026. "On the Diffusibility of High-Dimensional Latents." https://omanscience.com/ar/articles/on-the-diffusibility-of-high-dimensional-latents.
هارفارد
Feng, C., Xu, Z., Chen, B., Xiong, Y., Wang, X., Wang, J. H., Zhang, R., Lin, Z., Owens, A. and Li, Y. (2026) 'On the Diffusibility of High-Dimensional Latents', Available at: https://omanscience.com/ar/articles/on-the-diffusibility-of-high-dimensional-latents.
فانكوفر
Feng C, Xu Z, Chen B, Xiong Y, Wang X, Wang JH, et al. On the Diffusibility of High-Dimensional Latents. https://omanscience.com/ar/articles/on-the-diffusibility-of-high-dimensional-latents
IEEE
C. Feng, Z. Xu, B. Chen, Y. Xiong, X. Wang, J. H. Wang, R. Zhang, Z. Lin, A. Owens, and Y. Li, "On the Diffusibility of High-Dimensional Latents," https://omanscience.com/ar/articles/on-the-diffusibility-of-high-dimensional-latents.