Saved in:
Bibliographic Details
Main Authors: Oh, Gyutae, Bae, Jungwoo, Shin, Jitae
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.12816
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908883985891328
author Oh, Gyutae
Bae, Jungwoo
Shin, Jitae
author_facet Oh, Gyutae
Bae, Jungwoo
Shin, Jitae
contents Continual learning (CL) suffers from catastrophic forgetting, which is exacerbated in domain-incremental learning (DIL) where task identifiers are unavailable and storing past data is infeasible. While prompt-based CL (PCL) adapts representations with a frozen backbone, we observe that prompt-only improvements are often insufficient due to suboptimal prompt selection and classifier-level instability under domain shifts. We propose Residual SODAP, which jointly performs prompt-based representation adaptation and classifier-level knowledge preservation. Our framework combines $α$-entmax sparse prompt selection with residual aggregation, data-free distillation with pseudo-feature replay, prompt-usage--based drift detection, and uncertainty-aware multi-loss balancing. Across three DIL benchmarks without task IDs or extra data storage, Residual SODAP achieves state-of-the-art AvgACC/AvgF of 0.850/0.047 (DR), 0.760/0.031 (Skin Cancer), and 0.995/0.003 (CORe50).
format Preprint
id arxiv_https___arxiv_org_abs_2603_12816
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Residual SODAP: Residual Self-Organizing Domain-Adaptive Prompting with Structural Knowledge Preservation for Continual Learning
Oh, Gyutae
Bae, Jungwoo
Shin, Jitae
Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
Continual learning (CL) suffers from catastrophic forgetting, which is exacerbated in domain-incremental learning (DIL) where task identifiers are unavailable and storing past data is infeasible. While prompt-based CL (PCL) adapts representations with a frozen backbone, we observe that prompt-only improvements are often insufficient due to suboptimal prompt selection and classifier-level instability under domain shifts. We propose Residual SODAP, which jointly performs prompt-based representation adaptation and classifier-level knowledge preservation. Our framework combines $α$-entmax sparse prompt selection with residual aggregation, data-free distillation with pseudo-feature replay, prompt-usage--based drift detection, and uncertainty-aware multi-loss balancing. Across three DIL benchmarks without task IDs or extra data storage, Residual SODAP achieves state-of-the-art AvgACC/AvgF of 0.850/0.047 (DR), 0.760/0.031 (Skin Cancer), and 0.995/0.003 (CORe50).
title Residual SODAP: Residual Self-Organizing Domain-Adaptive Prompting with Structural Knowledge Preservation for Continual Learning
topic Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.12816