Saved in:
Bibliographic Details
Main Authors: Zhang, Yunzhe, Liu, Hongfu, Hong, Pengyu
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.19532
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917510943604736
author Zhang, Yunzhe
Liu, Hongfu
Hong, Pengyu
author_facet Zhang, Yunzhe
Liu, Hongfu
Hong, Pengyu
contents Text-to-image diffusion models can synthesize high-quality images, yet the outcome is notoriously sensitive to the random seed: different initial seeds often yield large variations in image quality and prompt-image alignment. We revisit this "seed effect" and show that attention dynamics over prompt core tokens, the content-bearing words, measured during the first few denoising steps, strongly predict final generation quality. Building on this observation, we introduce Attention-Based Seed Selection (ABSS), a training-free, plug-and-play method that ranks seeds for a given prompt by leveraging cross-attention to core tokens during the denoising process. ABSS requires no finetuning and does not alter the initial noise; it scores and ranks all candidate seeds, keeps only the top-k for full generation, and discards the rest, without relying on a fixed accept/reject threshold. Operating purely at inference time, ABSS can serve as a lightweight pre-selection add-on for existing seed-optimization pipelines, enabling additional gains. Across three benchmarks, extensive experiments show that ABSS enables consistent improvements in text-image alignment and visual quality for Stable Diffusion variants, as corroborated by human preference and alignment metrics.
format Preprint
id arxiv_https___arxiv_org_abs_2605_19532
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Boosting Text-to-Image Diffusion Models via Core Token Attention-Based Seed Selection
Zhang, Yunzhe
Liu, Hongfu
Hong, Pengyu
Computer Vision and Pattern Recognition
Machine Learning
Text-to-image diffusion models can synthesize high-quality images, yet the outcome is notoriously sensitive to the random seed: different initial seeds often yield large variations in image quality and prompt-image alignment. We revisit this "seed effect" and show that attention dynamics over prompt core tokens, the content-bearing words, measured during the first few denoising steps, strongly predict final generation quality. Building on this observation, we introduce Attention-Based Seed Selection (ABSS), a training-free, plug-and-play method that ranks seeds for a given prompt by leveraging cross-attention to core tokens during the denoising process. ABSS requires no finetuning and does not alter the initial noise; it scores and ranks all candidate seeds, keeps only the top-k for full generation, and discards the rest, without relying on a fixed accept/reject threshold. Operating purely at inference time, ABSS can serve as a lightweight pre-selection add-on for existing seed-optimization pipelines, enabling additional gains. Across three benchmarks, extensive experiments show that ABSS enables consistent improvements in text-image alignment and visual quality for Stable Diffusion variants, as corroborated by human preference and alignment metrics.
title Boosting Text-to-Image Diffusion Models via Core Token Attention-Based Seed Selection
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2605.19532