R^2MoE: Redundancy-Removal Mixture of Experts for Lifelong Concept Learning

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Guo, Xiaohan, Cai, Yusong, Liu, Zejia, Wang, Zhengning, Pan, Lili, Li, Hongliang
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866918096192667648
author Guo, Xiaohan
Cai, Yusong
Liu, Zejia
Wang, Zhengning
Pan, Lili
Li, Hongliang
author_facet Guo, Xiaohan
Cai, Yusong
Liu, Zejia
Wang, Zhengning
Pan, Lili
Li, Hongliang
contents Enabling large-scale generative models to continuously learn new visual concepts is essential for personalizing pre-trained models to meet individual user preferences. Existing approaches for continual visual concept learning are constrained by two fundamental challenges: catastrophic forgetting and parameter expansion. In this paper, we propose Redundancy-Removal Mixture of Experts (R^2MoE), a parameter-efficient framework for lifelong visual concept learning that effectively learns new concepts while incurring minimal parameter overhead. Our framework includes three key innovative contributions: First, we propose a mixture-of-experts framework with a routing distillation mechanism that enables experts to acquire concept-specific knowledge while preserving the gating network's routing capability, thereby effectively mitigating catastrophic forgetting. Second, we propose a strategy for eliminating redundant layer-wise experts that reduces the number of expert parameters by fully utilizing previously learned experts. Third, we employ a hierarchical local attention-guided inference approach to mitigate interference between generated visual concepts. Extensive experiments have demonstrated that our method generates images with superior conceptual fidelity compared to the state-of-the-art (SOTA) method, achieving an impressive 87.8\% reduction in forgetting rates and 63.3\% fewer parameters on the CustomConcept 101 dataset. Our code is available at {https://github.com/learninginvision/R2MoE}
format Preprint
id arxiv_https___arxiv_org_abs_2507_13107
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle R^2MoE: Redundancy-Removal Mixture of Experts for Lifelong Concept Learning
Guo, Xiaohan
Cai, Yusong
Liu, Zejia
Wang, Zhengning
Pan, Lili
Li, Hongliang
Computer Vision and Pattern Recognition
Enabling large-scale generative models to continuously learn new visual concepts is essential for personalizing pre-trained models to meet individual user preferences. Existing approaches for continual visual concept learning are constrained by two fundamental challenges: catastrophic forgetting and parameter expansion. In this paper, we propose Redundancy-Removal Mixture of Experts (R^2MoE), a parameter-efficient framework for lifelong visual concept learning that effectively learns new concepts while incurring minimal parameter overhead. Our framework includes three key innovative contributions: First, we propose a mixture-of-experts framework with a routing distillation mechanism that enables experts to acquire concept-specific knowledge while preserving the gating network's routing capability, thereby effectively mitigating catastrophic forgetting. Second, we propose a strategy for eliminating redundant layer-wise experts that reduces the number of expert parameters by fully utilizing previously learned experts. Third, we employ a hierarchical local attention-guided inference approach to mitigate interference between generated visual concepts. Extensive experiments have demonstrated that our method generates images with superior conceptual fidelity compared to the state-of-the-art (SOTA) method, achieving an impressive 87.8\% reduction in forgetting rates and 63.3\% fewer parameters on the CustomConcept 101 dataset. Our code is available at {https://github.com/learninginvision/R2MoE}
title R^2MoE: Redundancy-Removal Mixture of Experts for Lifelong Concept Learning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2507.13107