Semantic Transferability of Generative Rate-Distortion-Perception Systems
Abstract
Generative lossy compression combines a rate-limited representation with a stochastic decoder that preserves source-dependent information while generating unresolved detail. We ask what remains reusable when a rate-distortion-perception (RDP) system selected for one operating point is transferred to a prescribed conditional-generation task. We distinguish the single-letter RDP factorization from its block-code implementation and compare Full Redesign with three increasingly restrictive transfer strategies. Renderer-Only Transfer retains only the renderer and redesigns semantic steering. Representation-Induced Transfer additionally retains the semantic latent law induced by the original representation. Full-Model Transfer retains a compatible RDP code, marginalizing over the task conditional law to form its semantic encoder. For finite alphabets, we characterize Renderer-Only Transfer's exact compression-common-randomness rate region, renderer-hull controllability condition, and compression-rate penalty relative to Full Redesign. Under unlimited common randomness, this penalty is the minimum residual semantic information I(S;U|X) forced by the inherited renderer. Representation-Induced Transfer is exact if and only if the semantic condition depends on the source only through the recurrent-class label C of the end-to-end RDP test channel. Every compatible RDP code reused under Full-Model Transfer preserves C and supports every task satisfying X−C−S. Any additionally transferable within-class semantic dependence must satisfy I(S;X|C)≤ R0−H(C), where R0 is the inherited compression rate. The bound is exact when R0=H(C). A scalar Gaussian specialization yields a renderer-noise controllability threshold, a closed-form transfer penalty, and examples of identical Gaussian RDP test channels with different Renderer-Only Transfer capabilities.
Related
- Generative Rate-Distortion-Perception Theory on Riemannian Manifolds
- An Overview of Rate-Distortion-Perception Theory
- Rate-Distortion-Perception Tradeoff for Vector Gaussian Sources with Limited Common Randomness: Exact Characterization and Approximate Universality
- A Self-Coupling Gaussian Extremal Inequality for Information-Constrained Optimal Transport with Applications to Rate-Distortion-Perception
- Cross-Modal Semantic Successive Refinement for Generative Image Communication over Underwater Acoustic Channels
BibTeX
@misc{jeon2026semantic,
title = {{Semantic Transferability of Generative Rate-Distortion-Perception Systems}},
author = {Jinwan Jeon and Seoyeon Hwang and Sung Hoon Lim},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.34697},
eprint = {2610.34697},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.34697},
}
RIS
TY - UNPB
TI - Semantic Transferability of Generative Rate-Distortion-Perception Systems
AU - Jinwan Jeon
AU - Seoyeon Hwang
AU - Sung Hoon Lim
PY - 2026
DA - 2026/10/26
PB - arXiv
DO - 10.48550/arxiv.2610.34697
UR - https://arxiv.org/abs/2610.34697
AB - Generative lossy compression combines a rate-limited representation with a stochastic decoder that preserves source-dependent information while generating unresolved detail. We ask what remains reusable when a rate-distortion-perception (RDP) system selected for one operating point is transferred to a prescribed conditional-generation task. We distinguish the single-letter RDP factorization from its block-code implementation and compare Full Redesign with three increasingly restrictive transfer strategies. Renderer-Only Transfer retains only the renderer and redesigns semantic steering. Representation-Induced Transfer additionally retains the semantic latent law induced by the original representation. Full-Model Transfer retains a compatible RDP code, marginalizing over the task conditional law to form its semantic encoder. For finite alphabets, we characterize Renderer-Only Transfer's exact compression-common-randomness rate region, renderer-hull controllability condition, and compression-rate penalty relative to Full Redesign. Under unlimited common randomness, this penalty is the minimum residual semantic information I(S;U|X) forced by the inherited renderer. Representation-Induced Transfer is exact if and only if the semantic condition depends on the source only through the recurrent-class label C of the end-to-end RDP test channel. Every compatible RDP code reused under Full-Model Transfer preserves C and supports every task satisfying X−C−S. Any additionally transferable within-class semantic dependence must satisfy I(S;X|C)≤ R_0−H(C), where R_0 is the inherited compression rate. The bound is exact when R_0=H(C). A scalar Gaussian specialization yields a renderer-noise controllability threshold, a closed-form transfer penalty, and examples of identical Gaussian RDP test channels with different Renderer-Only Transfer capabilities.
ER -