Spectral Evolution Search: Efficient Inference-Time Scaling for Reward-Aligned Image Generation

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Ye, Jinyan, Duan, Zhongjie, Li, Zhiwen, Chen, Cen, Chen, Daoyuan, Li, Yaliang, Chen, Yingda
Format:	Preprint
Published:	2026
Subjects:	Machine Learning Computer Vision and Pattern Recognition
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

_version_	1866911418088947712
author	Ye, Jinyan Duan, Zhongjie Li, Zhiwen Chen, Cen Chen, Daoyuan Li, Yaliang Chen, Yingda
author_facet	Ye, Jinyan Duan, Zhongjie Li, Zhiwen Chen, Cen Chen, Daoyuan Li, Yaliang Chen, Yingda
contents	Inference-time scaling offers a versatile paradigm for aligning visual generative models with downstream objectives without parameter updates. However, existing approaches that optimize the high-dimensional initial noise suffer from severe inefficiency, as many search directions exert negligible influence on the final generation. We show that this inefficiency is closely related to a spectral bias in generative dynamics: model sensitivity to initial perturbations diminishes rapidly as frequency increases. Building on this insight, we propose Spectral Evolution Search (SES), a plug-and-play framework for initial noise optimization that executes gradient-free evolutionary search within a low-frequency subspace. Theoretically, we derive the Spectral Scaling Prediction from perturbation propagation dynamics, which explains the systematic differences in the impact of perturbations across frequencies. Extensive experiments demonstrate that SES significantly advances the Pareto frontier of generation quality versus computational cost, consistently outperforming strong baselines under equivalent budgets.
format	Preprint
id	arxiv_https___arxiv_org_abs_2602_03208
institution	arXiv
publishDate	2026
record_format	arxiv
spellingShingle	Spectral Evolution Search: Efficient Inference-Time Scaling for Reward-Aligned Image Generation Ye, Jinyan Duan, Zhongjie Li, Zhiwen Chen, Cen Chen, Daoyuan Li, Yaliang Chen, Yingda Machine Learning Computer Vision and Pattern Recognition Inference-time scaling offers a versatile paradigm for aligning visual generative models with downstream objectives without parameter updates. However, existing approaches that optimize the high-dimensional initial noise suffer from severe inefficiency, as many search directions exert negligible influence on the final generation. We show that this inefficiency is closely related to a spectral bias in generative dynamics: model sensitivity to initial perturbations diminishes rapidly as frequency increases. Building on this insight, we propose Spectral Evolution Search (SES), a plug-and-play framework for initial noise optimization that executes gradient-free evolutionary search within a low-frequency subspace. Theoretically, we derive the Spectral Scaling Prediction from perturbation propagation dynamics, which explains the systematic differences in the impact of perturbations across frequencies. Extensive experiments demonstrate that SES significantly advances the Pareto frontier of generation quality versus computational cost, consistently outperforming strong baselines under equivalent budgets.
title	Spectral Evolution Search: Efficient Inference-Time Scaling for Reward-Aligned Image Generation
topic	Machine Learning Computer Vision and Pattern Recognition
url	https://arxiv.org/abs/2602.03208