Training-Free Rate-Distortion-Perception Traversal With Diffusion

Yuhan Wang, Suzhi Bi, Ying-Jun Angela Zhang · 2026

Read Training-Free Rate-Distortion-Perception Traversal With Diffusion by Yuhan Wang, Suzhi Bi, Ying-Jun Angela Zhang (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-04005.txt.

Description

Wang and others propose a training-free diffusion method for traversing the rate-distortion-perception tradeoff. The tradeoff captures lossy compression limits across bitrate, fidelity, and quality, yet neural methods usually sit at fixed points needing retraining. The method combines reverse channel coding with a score-scaled flow decoder built on pretrained diffusion models. Theory proves the decoder optimal for distortion-perception under white Gaussian noise, and the full method optimal in the Gaussian case. Experiments confirm flexible navigation of the tradeoff, offering a real route to adaptive compression.

Catalog record

CREATORYuhan Wang, Suzhi Bi, Ying-Jun Angela Zhang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE126211
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.04005

Browse the full library · About the Tower of Babel