When Classes Evolve: A Benchmark and Framework for Stage-Aware Class-Incremental Learning

2026-01-31

AI Summary

This work addresses the limitation of conventional class-incremental learning (CIL), which overlooks the phenomenon of intra-class morphological evolution over time and thus struggles to adapt to changes within categories. The study formally introduces the problem of intra-class morphological evolution and proposes a novel paradigm termed Stage-aware Class-Incremental Learning (Stage-CIL). To facilitate systematic evaluation, the authors construct Stage-Bench—a benchmark spanning ten domains and two temporal stages—that assesses models’ performance in balancing inter-class forgetting and intra-class evolution. The proposed STAGE method explicitly learns evolutionary patterns by dynamically decoupling semantic identity from morphological transformation within a fixed-capacity memory buffer, enabling prediction of future morphologies. Experiments demonstrate that STAGE significantly outperforms existing approaches on Stage-Bench, effectively achieving both inter-class discriminability and intra-class adaptability.

Technology Category

• Machine Learning: Evolutionary Learning
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior
• Search and Optimization: Evolutionary Computation

Application Category

• Web Mining and Content Analysis: Models for Web evolution
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Class-Incremental Learning (CIL) aims to sequentially learn new classes while mitigating catastrophic forgetting of previously learned knowledge. Conventional CIL approaches implicitly assume that classes are morphologically static, focusing primarily on preserving previously learned representations as new classes are introduced. However, this assumption neglects intra-class evolution: a phenomenon wherein instances of the same semantic class undergo significant morphological transformations, such as a larva turning into a butterfly. Consequently, a model must both discriminate between classes and adapt to evolving appearances within a single class. To systematically address this challenge, we formalize Stage-Aware CIL (Stage-CIL), a paradigm in which each class is learned progressively through distinct morphological stages. To facilitate rigorous evaluation within this paradigm, we introduce the Stage-Bench, a 10-domain, 2-stages dataset and protocol that jointly measure inter- and intra-class forgetting. We further propose STAGE, a novel method that explicitly learns abstract and transferable evolution patterns within a fixed-size memory pool. By decoupling semantic identity from transformation dynamics, STAGE enables accurate prediction of future morphologies based on earlier representations. Extensive empirical evaluation demonstrates that STAGE consistently and substantially outperforms existing state-of-the-art approaches, highlighting its effectiveness in simultaneously addressing inter-class discrimination and intra-class morphological adaptation.

Problem

• Class-Incremental Learning
• Catastrophic Forgetting
• Intra-class Evolution
• Morphological Transformation
• Stage-Aware Learning

Innovation

• Stage-Aware Class-Incremental Learning
• Intra-class Evolution
• Morphological Transformation
• Catastrophic Forgetting
• Memory-Efficient Learning