A Universal Avoidance Method for Diverse Multi-branch Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Park, Kyeongman, Jhang, Minha, Jung, Kyomin
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918454655713280
author Park, Kyeongman
Jhang, Minha
Jung, Kyomin
author_facet Park, Kyeongman
Jhang, Minha
Jung, Kyomin
contents Modern generative models still lack human-level creativity, particularly in multi-branch diversity. Prior approaches to address this problem often incur heavy computation or strong dependency on model architecture. Therefore, we introduce UAG(Universal Avoidance Generation), a model-agnostic and computationally efficient generation strategy that penalizes similarity among previously generated outputs. Thus, UAG can enhance multi-branch diversity across both diffusion and transformer models, with minimal additional computation. In experiments, our method achieves up to 1.9 times higher diversity, runs 4.4 times faster, and requires only 1/64 of the FLOPs compared to state-of-the-art methods. The full code is https://anonymous.4open.science/r/2026_ACL_Universal/.
format Preprint
id arxiv_https___arxiv_org_abs_2604_17323
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle A Universal Avoidance Method for Diverse Multi-branch Generation
Park, Kyeongman
Jhang, Minha
Jung, Kyomin
Computation and Language
Machine Learning
Modern generative models still lack human-level creativity, particularly in multi-branch diversity. Prior approaches to address this problem often incur heavy computation or strong dependency on model architecture. Therefore, we introduce UAG(Universal Avoidance Generation), a model-agnostic and computationally efficient generation strategy that penalizes similarity among previously generated outputs. Thus, UAG can enhance multi-branch diversity across both diffusion and transformer models, with minimal additional computation. In experiments, our method achieves up to 1.9 times higher diversity, runs 4.4 times faster, and requires only 1/64 of the FLOPs compared to state-of-the-art methods. The full code is https://anonymous.4open.science/r/2026_ACL_Universal/.
title A Universal Avoidance Method for Diverse Multi-branch Generation
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2604.17323