The Prism Hypothesis: Harmonizing Semantic and Pixel Representations via Unified Autoencoding

TL;DR

This paper introduces the Prism Hypothesis and Unified Autoencoding (UAE), a model that harmonizes semantic and pixel representations by leveraging spectral characteristics, leading to improved image modeling performance.

Contribution

It uncovers a spectral correspondence between semantic and pixel encoders and proposes UAE to unify these representations through frequency-band modulation.

Findings

UAE achieves state-of-the-art performance in unifying semantic and pixel representations.

UAE improves FID and IS scores over baseline models.

Spectral analysis reveals semantic encoders focus on low-frequency components, while pixel encoders retain high-frequency details.

Abstract

Deep representations across modalities are inherently intertwined. In this paper, we systematically analyze the spectral characteristics of various semantic and pixel encoders. Interestingly, our study uncovers a highly inspiring and rarely explored correspondence between an encoder's feature spectrum and its functional role: semantic encoders primarily capture low-frequency components that encode abstract meaning, whereas pixel encoders additionally retain high-frequency information that conveys fine-grained detail. This heuristic finding offers a unifying perspective that ties encoder behavior to its underlying spectral structure. We define it as the Prism Hypothesis, where each data modality can be viewed as a projection of the natural world onto a shared feature spectrum, just like the prism. Building on this insight, we propose Unified Autoencoding (UAE), a model that harmonizes semantic structure and pixel details via an innovative frequency-band modulator, enabling their seamless coexistence. Extensive experiments demonstrate that UAE effectively unifies semantic abstraction and pixel-level fidelity within a single latent space, achieving state-of-the-art performance. Moreover, we show that UAE can be directly applied to pixel-space modeling, significantly improving both FID and IS over the vanilla JIT baseline. Our code is avaliable at: https://github.com/WeichenFan/UAE.