Modular Embedding Recomposition for Incremental Learning
Panariello, Aniello, Frascaroli, Emanuele, Buzzega, Pietro, Bonicelli, Lorenzo, Porrello, Angelo, Calderara, Simone
–arXiv.org Artificial Intelligence
The advent of pre-trained Vision-Language Models (VLMs) has significantly transformed Continual Learning (CL), mainly due to their zero-shot classification abilities. Such proficiency makes VLMs well-suited for real-world applications, enabling robust performance on novel unseen classes without requiring adaptation. However, fine-tuning remains essential when downstream tasks deviate significantly from the pre-training domain. Prior CL approaches primarily focus on preserving the zero-shot capabilities of VLMs during incremental fine-tuning on a downstream task. We take a step further by devising an approach that transforms preservation into enhancement of the zero-shot capabilities of VLMs. Our approach, named MoDular Embedding Recomposition (MoDER), introduces a modular framework that trains multiple textual experts, each specialized in a single seen class, and stores them in a foundational hub. At inference time, for each unseen class, we query the hub and compose the retrieved experts to synthesize a refined prototype that improves classification. We show the effectiveness of our method across two popular zero-shot incremental protocols, Class-IL and MTIL, comprising a total of 14 datasets. The codebase is available at https://github.com/aimagelab/mammoth.
arXiv.org Artificial Intelligence
Oct-15-2025
- Country:
- Europe (0.28)
- Genre:
- Research Report (0.64)
- Industry:
- Health & Medicine (0.46)
- Technology:
- Information Technology > Artificial Intelligence
- Vision (1.00)
- Natural Language > Large Language Model (1.00)
- Machine Learning > Neural Networks (1.00)
- Information Technology > Artificial Intelligence