Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Inch, Alex, Chaiyapattanaporn, Passawis, Zhu, Yuchen, Lu, Yuan, Ko, Ting-Wen, Paglieri, Davide
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2511.00686
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866917056350257152
author Inch, Alex
Chaiyapattanaporn, Passawis
Zhu, Yuchen
Lu, Yuan
Ko, Ting-Wen
Paglieri, Davide
author_facet Inch, Alex
Chaiyapattanaporn, Passawis
Zhu, Yuchen
Lu, Yuan
Ko, Ting-Wen
Paglieri, Davide
contents Text-to-image diffusion models, while proficient at generating high-fidelity images, often suffer from limited output diversity, hindering their application in exploratory and ideation tasks. Existing prompt optimization techniques typically target aesthetic fitness or are ill-suited to the creative visual domain. To address this shortcoming, we introduce WANDER, a novelty search-based approach to generating diverse sets of images from a single input prompt. WANDER operates directly on natural language prompts, employing a Large Language Model (LLM) for semantic evolution of diverse sets of images, and using CLIP embeddings to quantify novelty. We additionally apply emitters to guide the search into distinct regions of the prompt space, and demonstrate that they boost the diversity of the generated images. Empirical evaluations using FLUX-DEV for generation and GPT-4o-mini for mutation demonstrate that WANDER significantly outperforms existing evolutionary prompt optimization baselines in diversity metrics. Ablation studies confirm the efficacy of emitters.
format Preprint
id arxiv_https___arxiv_org_abs_2511_00686
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Evolve to Inspire: Novelty Search for Diverse Image Generation
Inch, Alex
Chaiyapattanaporn, Passawis
Zhu, Yuchen
Lu, Yuan
Ko, Ting-Wen
Paglieri, Davide
Computer Vision and Pattern Recognition
Artificial Intelligence
Text-to-image diffusion models, while proficient at generating high-fidelity images, often suffer from limited output diversity, hindering their application in exploratory and ideation tasks. Existing prompt optimization techniques typically target aesthetic fitness or are ill-suited to the creative visual domain. To address this shortcoming, we introduce WANDER, a novelty search-based approach to generating diverse sets of images from a single input prompt. WANDER operates directly on natural language prompts, employing a Large Language Model (LLM) for semantic evolution of diverse sets of images, and using CLIP embeddings to quantify novelty. We additionally apply emitters to guide the search into distinct regions of the prompt space, and demonstrate that they boost the diversity of the generated images. Empirical evaluations using FLUX-DEV for generation and GPT-4o-mini for mutation demonstrate that WANDER significantly outperforms existing evolutionary prompt optimization baselines in diversity metrics. Ablation studies confirm the efficacy of emitters.
title Evolve to Inspire: Novelty Search for Diverse Image Generation
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2511.00686