Class-Incremental Learning

Overview

Class-incremental learning trains a model on a sequence of classes, often without task labels at test time, and aims to limit catastrophic forgetting as new categories arrive. Methods include replaying stored or generated examples, regularizing or distilling old knowledge, using class prototypes, and isolating updates with adapters, prompts, or experts; some use pretrained models or analytic updates to reduce training and memory costs. Research spans standard image benchmarks as well as few-shot, long-tailed, online, federated, and structured tasks such as detection and segmentation. A central challenge is balancing adaptation to new classes with retention of old ones under limited memory, changing data distributions, and longer, less predictable task sequences.

Latest papers

171 papers13 papers in the last four weeks, down 54% on the four weeks before. 0.1% of all new papers.

Nov 16, 2026

Network Intrusion Detection Adaptability under Spurious Labels: A Noise-Aware Few-Shot Class-Incremental Learning Framework

Network Intrusion Detection Systems (NIDSs) must incrementally adapt to emerging attacks, even when only a limited amount of labeled network traffic is available. Consequently, Few-Shot Class-Incremental Learning (FSCIL)-based NIDSs have gained growing attention. However, existing FSCIL-based NIDSs are highly vulnerable to label noise in newly-observed attack classes. To address this limitation, we propose NAFN, a noise-aware FSCIL-based NIDS framework that reduces the impact of mislabeled data during incremental adaptation. NAFN consists of two components: (i) Noise Detection, which identifies and removes mislabeled biflows using noise-detection algorithms, and (ii) Feature Sampling, which estimates class-conditional embeddings from clean biflows, calibrates them with base-class statistics, and resamples synthetic features to recover the original biflow count. The framework follows the Rethink Few-Shot (RFS) teacher-student paradigm with lightweight nearest-neighbor or logistic regression classifiers, and various pretraining & memory selection strategies. Experiments on CSE-CIC-IDS2018, Edge-IIoT, and IoT-NID datasets show that NAFN effectively preserves prior knowledge and maintains strong incremental performance even under 50% label noise, consistently improving new-class F1-scores over non-denoised baselines, up to +52%, while being time-efficient and robust in cross-network and high-noise incremental settings. The implementation is publicly available to foster reproducibility.
Nov 16, 2026

Learning What the True Labels Would Teach: Posterior Analytic Class-Incremental Learning with Unlabeled Data

In class-incremental learning with unlabeled data, each task brings only class names and unlabeled images, so CLIP's zero-shot predictions replace labels. Existing methods train on hard pseudo-labels, but a wrong one gives the true class zero probability, which self-training often confirms. We argue that each image should carry its label posterior, as an analytic classifier trained on posteriors equals the true-label one in expectation. The zero-shot prediction is not this posterior, as its class proportions confound name bias with true imbalance and it ignores other images. We propose Posterior Analytic Class-Incremental Learning (PACIL), which corrects the class prior with a prior selection rule and refines each distribution by a fixed-prior von Mises-Fisher mixture on whitened features. These soft labels train the classifier and replace labels to select its fusion weight with zero-shot CLIP by closed-form leave-one-out cross-validation. We prove that posterior targets minimize the mean-square gap to the true-label prediction and that their argmax widens this gap only through ambiguous images. The fusion weight selected with soft labels loses at most their distance to the posteriors in expected correct decisions. PACIL has the best average incremental accuracy among methods trained without labels on all 18 settings.
Nov 16, 2026

From Joint Risk to Decision Invariance: Decision-Preserving Regularization for Exemplar-Free Class-Incremental Learning

Exemplar-free continual learning seeks to minimize joint risk over all seen tasks, yet historical data are unavailable for direct optimization. We observe that, for classification, preserving the model's decisions on previously learned classes helps retain old knowledge. This motivates Decision-Preserving Regularization (DPR). Penalizing changes in all pairwise old-class logit margins reduces algebraically to a centered relative-logit penalty. We further apply spectral whitening to remove the classifier's spectral scale bias. Because the global classifier supplies the decision constraints for the next stage, it must remain calibrated across all seen classes despite the drift of stored old-class statistics during representation updates. We therefore fit a closed-form affine map from paired current-task features to transport old-class Gaussian means and covariances into the updated feature space, then calibrate the global classifier using these distributions and current-task features. Across CIFAR-100, Tiny-ImageNet, ImageNet-Subset, and ImageNet-1K under cold-start protocols, comparisons with state-of-the-art methods show that DPR achieves higher accuracy in most settings while remaining highly efficient. Code is available at https://github.com/bnufw/dpr.
Nov 11, 2026

Federated Class Incremental Learning: A Pseudo Feature Based Approach Without Exemplars

Federated learning often assumes that data is fixed in advance, which is unrealistic in many real-world scenarios where new data continuously arrives, causing catastrophic forgetting. To address this challenge, we propose FCLPF (Federated Class Incremental Learning with Pseudo Features), a method that uses pseudo features generated from prototypes to mitigate catastrophic forgetting. Our approach reduces communication costs and improves efficiency by eliminating the need for past data and avoiding computationally heavy models like GANs. Experimental results on CIFAR-100 show that FCLPF achieves an average accuracy of 51.87% and an average forgetting of 9.62%, significantly outperforming existing baselines with an average accuracy of 47.72% and forgetting of 20.46%. On TinyImageNet, FCLPF achieves 37.56% accuracy and 3.14% forgetting, compared to the baselines' 27.69% accuracy and 24.46% forgetting, demonstrating the superior performance of FCLPF.
Nov 4, 2026

PRISM-AMC: Replay-Free Continual Learning for Automatic Modulation Classification

Automatic modulation classification (AMC) is a key component of cognitive wireless systems, enabling spectrum awareness and adaptive communication. Most deep learning approaches to AMC assume a fixed set of modulation classes and access to previously observed training data. In practice, modulation classes and channel conditions may evolve over time, while storage or data-retention constraints may preclude replay of raw baseband I/Q samples. We propose PRISM-AMC, a replay-free continual learning framework that mitigates catastrophic forgetting by preserving task-relevant subspaces of intermediate representations via gradient projection memory. We further analyze projected stochastic-gradient updates under an SNR-dependent gradient-noise model. PRISM-AMC is evaluated in both class-incremental and domain-incremental settings. Experiments on the MIGOU and DeepSig 2018.01A datasets show substantially reduced forgetting relative to sequential fine-tuning and elastic weight consolidation, while remaining competitive with or outperforming replay-based RSD-AMC in class-incremental settings, without storing raw I/Q samples.
Nov 3, 2026

P-SER: Physical Consistency-Constrained Scattering Expandable Representation for Class-Incremental HRRP Target Recognition

High-resolution range profiles (HRRPs) provide compact radar target representations but exhibit strong aspect sensitivity, complicating class-incremental recognition when historical data are unavailable. Changes in viewing aspect shift scattering-peak locations and amplitudes, while reflector groups may produce responses resembling ships. Existing dynamic expansion methods can also incur substantial parameter and memory overhead. To address these challenges, we propose P-SER, a Physical Consistency-Constrained Scattering Expandable Representation framework for exemplar-free class-incremental HRRP recognition. P-SER reconstructs dominant scattering points from complex-valued sum- and azimuth-difference-channel HRRPs and observation geometry. The reconstructed spatial structure, target extent, and energy distribution are used to derive class-level scattering prototypes and sample-level saliency priors, providing physically grounded constraints for feature learning and old-class knowledge retention. A primitive scatterer dictionary further characterizes local strong-scattering mechanisms and calibrates uncertain ship--reflector predictions at inference. Meanwhile, lightweight residual adapters are incrementally introduced into a shared HRRP backbone to learn new classes with limited model growth. Experiments on simulated and measured HRRP datasets show that P-SER delivers competitive recognition performance without retaining old-class exemplars, while maintaining physical consistency, interference robustness, knowledge retention, and model scalability.
Nov 1, 2026

Hippocampal consolidation without replay in fully fine-tuned networks

Networks that keep learning overwrite what they know. Prevailing remedies replay stored data, distil from old models or freeze the backbone. The brain suggests a different division of labour: the hippocampus indexes where experience lies in cortex while the cortex keeps learning. We show that this division lets a fully trainable network keep learning without cross-task replay, a teacher or inference-time memory. After each task, a training-only hippocampal memory indexes the input geometry of every group of synapses sharing an input, residual energy included, and limits each synapse's movement per direction before and after the optimizer. This class-free geometry carries most of the protection; with conflict detection on top, the method cuts absorption of established classes by near-duplicates several-fold against every compared method. To our knowledge, this is the first demonstration of one principle keeping whole networks learning across four architecture families and models up to 8 billion parameters. Methods that freeze the backbone hold up only as far as its representation already separates the classes; here the whole network keeps learning to tell near-duplicates apart, leading the frozen-backbone baseline by 17.5 points on TACRED. Without the memory, much old knowledge stays decodable but beyond the network's reach.
Oct 29, 2026

Class-aware Latent Attention Replay (CLAR): Class Incremental Hazard Diagnosis for Resilient AI in the Manufacturing Industrial Internet

Artificial Intelligence (AI) systems provide critical computation services in the Manufacturing Industrial Internet (MII), and their resilience depends on timely diagnosis of hazards across the data, AI pipeline, and cyber-physical layers. During operation, new hazard types keep emerging with only a few samples each, hazards co-occur, and the diagnosis model runs on memory-constrained computation nodes. Replay-based class-incremental learning (CIL) can learn new hazard classes while revisiting a small memory of past samples, but existing methods assume single-label, balanced streams, store high-dimensional raw samples, and select them independently of the classifier. We propose Class-aware Latent Attention Replay (CLAR), which combines a dual buffer that stores mostly compact latent representations with a few raw samples, an Attention Leverage Score (ALS) that upper-bounds the classifier logit and guides class-balanced reservoir sampling, and an exponential-moving-average teacher that slows latent-space drift. For an idealized buffer model, we prove that the buffer is class-balanced and follows an ALS-weighted class distribution, and that below a critical per-class capacity, ALS-guided replay has a lower replay-gradient error than class-balanced uniform replay. In the case study, On hazard diagnosis data from an MII testbed, CLAR achieves the highest mean average weighted F1 score among the evaluated replay methods with the same buffer size
Oct 27, 2026

Generative Replay for Equilibrium Propagation: Toward Edge-Friendly Continual Learning in Hopfield Networks

Equilibrium Propagation (EqProp) offers a biologically grounded and hardware-aligned alternative to backpropagation for training Hopfield Networks. However, its behavior in continual learning settings remains largely unexplored. This work confronts catastrophic forgetting in EqProp-trained Hopfield Networks from the inside out, eliminating the need for external data buffers or architectural overhead that standard Experience Replay (ER) and Generative Replay (Gen-Replay) inevitably require. We introduce an Energy-Based Generative Replay strategy (EB-Gen) which exploits the intrinsic generative dynamics of Hopfield Networks, allowing a single model to act as both classifier and replay samples generator. Samples capturing past task knowledge are generated by relaxing a frozen copy of the network, enabling rehearsal solely through internally generated pseudo-data. We evaluate the proposed method in the Class-Incremental setting, using the split versions of the MNIST and CIFAR10 benchmarks and compare it with FineTune, Experience Replay, Generative Replay, Sleep-Like Replay Consolidation (SRC) and Joint training. Our results show that EB-Gen significantly improves stability while preserving plasticity, achieving a 75% and 64% average accuracy on Split MNIST and Split CIFAR10 respectively. These findings show that Equilibrium Propagation can be adapted to alleviate catastrophic forgetting without forgoing the traits which made it compelling in the first place.
Oct 26, 2026

TRACE: Memory-Efficient Time Series Class-Incremental Learning via Temporal-Spectral Reconstruction

Web-connected sensing applications continuously generate multivariate time series with evolving behavioral categories, making Time Series Class-Incremental Learning (TSCIL) essential for adapting to new classes without forgetting previously acquired knowledge. However, preserving diverse and evolving class distributions under strict privacy and memory constraints remains a fundamental challenge. Existing approaches often face a trade-off between retaining sufficient historical knowledge for accurate recognition and minimizing the memory required for continual adaptation. To address these limitations, we propose TRACE, a memory-efficient, replay-free framework that preserves class-specific knowledge through generative modeling rather than sample rehearsal. TRACE employs a compact, frozen temporal-spectral representation and lightweight class-specific autoencoders to capture complex intra-class variations. By independently preserving each class model, TRACE avoids recursive generation and catastrophic forgetting, while enabling direct classification through reconstruction-based scoring without retaining historical samples or large statistical matrices. Extensive experiments across five TSCIL benchmarks demonstrate that TRACE achieves competitive classification accuracy while reducing mean retained memory by approximately 98.26% compared with TS-ACL, the strongest evaluated analytical baseline, and achieving an approximately 33.43% relative improvement in accuracy on GrabMyo.
Oct 25, 2026

Cross-User Adaptation for sEMG Gesture Recognition via Continual Learning

Prosthetic and assistive interfaces increasingly depend on decoding hand gestures from surface electromyography (sEMG). Because the signal is shaped by person-specific factors, the recurrent neural networks (RNNs) commonly used to model these temporal streams must adapt to new wearers and new gesture classes after deployment. Continual learning (CL) enables this adaptation without retraining from scratch, yet applying it to recurrent sEMG models is not straightforward. Standard recurrent training practices can destabilize learning, and which CL strategy holds up under realistic cross-user drift is unestablished. We evaluate three representative CL strategies under domain- and class-incremental drift on NinaPro DB1, and find experience replay to outperform elastic weight consolidation (EWC) and naive fine-tuning. It reaches 83.3% final accuracy for class-incremental drift while limiting forgetting to 1.8-5.8% accuracy drop, as against 16.2-28.5% for EWC and naive training. We further show these models must be trained on shuffled, independently-windowed segments. Streaming windows in temporal order collapses the network into predicting a single dominant gesture, while a shuffled, stateless regime trains stably. Finally, hand-crafted time-domain features raise per-task accuracy by up to about 40% over raw signals, but only replay preserves that gain under class-incremental drift.
Oct 23, 2026

CARES: Coverage-Aware Replay Exemplar Selection for Continual Learning

Replay-based Continual Learning mitigates catastrophic forgetting by keeping a buffer of past samples, making buffer composition critical under a fixed memory budget. Recent coreset-based replay methods improve buffer quality using model-derived gradients, losses, or auxiliary optimization, but incur substantial computational overhead as memory is repeatedly updated. We propose CARES, a continual model-optimization-free approach for buffer construction that instead exploits the geometry of a frozen pretrained feature space using class-aware, density-weighted coverage. We evaluate CARES across balanced, imbalanced, and label-noisy streams under online class- and task-incremental settings. On Tiny-ImageNet, CARES is 2.3 to 8.8 times faster than existing coreset-based selectors. Although CARES achieves this speedup using no training signal from the continual learner, it ranks among the state of the art across balanced, imbalanced, and label-noisy streams under online class- and task-incremental settings, outperforming the strongest baseline by up to 7.58 percentage points under class imbalance while also attaining higher average accuracy on Tiny-ImageNet. Code will be released upon publication
Oct 22, 2026

REDAR: Reinforcement-Driven Adaptive Regularisation for Class Incremental Learning

Regularisation-based class-incremental learning (CIL) methods control the stability-plasticity trade-off through fixed coefficients tuned offline, independently of the continual learning objective and blind to phase-specific variations in data volume, task difficulty, and class similarity. We argue this is a fundamental mismatch and propose REDAR: Reinforcement-Driven Adaptive Regularisation for Class Incremental Learning, which treats regularisation strength selection as a sequential decision problem. At each incremental phase, a lightweight reinforcement learning (RL) agent observes a compact state of current performance metrics and incoming data statistics, and outputs the distillation coefficients for that phase. To respect CIL's constraint on historical data access, the policy is trained entirely on pseudo-CIL tasks constructed from the initial phase, then deployed on the incremental sequence without further adaptation. Experiments on CIFAR-100 and Tiny-ImageNet demonstrate that REDAR reduces forgetting and improves forward transfer over fixed and heuristic baselines, and that policies trained on short pseudo-task horizons transfer effectively to longer incremental runs.
Oct 19, 2026

Generalizing Multimodal Foundational Models in Context-Driven Class Incremental Learning

Class-Incremental Learning (CIL) aims to learn new classes continuously without forgetting the previously learned ones. Recent works in CIL have significantly advanced across various modalities; however, at the intersection of audio-visual modality, this problem remains underexplored. While representations from multimodal foundational models (MFMs) encapsulate rich static priors, our preliminary analysis shows that these representations struggle in incremental settings. Motivated by this, we propose a framework that generalizes frozen MFMs to audio-visual CIL. Specifically, we leverage their dense audio and visual representations and employ an attention-based contextual guidance module, in which the audio representations serve as task-stable contextual guidance for the visual representations, enabling effective class recognition in the incremental setting. To further mitigate catastrophic forgetting and preserve stability across tasks, we employ feature- and logit-level distillation, which are primarily responsible for the reduced forgetting. Extensive evaluations across various audio-visual CIL benchmarks demonstrate that our approach consistently outperforms state-of-the-art methods. Code is available at https://github.com/avigupta2798/MFM-CIL.
Oct 19, 2026

LSA-PAST: Layer-Specific Adaptation and Prototype Transport for Compositional Incremental Learning

Class-Incremental Learning (CIL) enables models to recognize new classes over time while retaining prior knowledge, but many real-world problems are compositional, where visual concepts arise from combinations of primitives such as states (i.e., red, shiny) and objects (i.e., shoe, car). Standard CIL methods treat each composition as an independent class, ignoring shared primitives and providing no mechanism to stabilize their representations as new compositions arrive. This causes parameter updates to entangle primitives and induce drift that degrades performance on earlier compositions. Compositional Incremental Learning (composition-IL) formalizes this setting by introducing new state-object compositions across tasks. In this paper, we propose Layer-Specific Adaptation with Prototype-Anchored Adapter-Space Transport (LSA-PAST), a parameter-efficient method tailored to composition-IL. LSA integrates Low-Rank Adaptation (LoRA) into a frozen ViT with depth-specialized adapters for states, objects, and compositions. PAST then maps past, anchored prototypes into the current adapter space through a rehearsal-free affine transform fit on current-task data, enabling old prototypes to remain compatible with new adapters. Extensive experiments demonstrate that LSA-PAST achieves state-of-the-art (SOA) accuracy-forgetting trade-off performance for composition-IL.
Oct 19, 2026

MT-CC: Multi-Group Temperature Scaling for Asymmetric Calibration Behavior in Class-Incremental Learning

As deep neural networks are required to continuously adapt to evolving data, class-incremental learning (CIL) has become an active research area. However, most existing works primarily focus on improving accuracy while overlooking confidence calibration, which measures the trustworthiness of the model's predictions. Although temperature scaling (TS) is effective for static calibration, it has rarely been explored in continual settings where data distributions evolve over time. In this paper, we introduce Multi-Group Temperature Scaling for Continual Calibration (MT-CC), a novel post-hoc calibration framework that addresses the asymmetric calibration behavior in CIL, where the rate of confidence changes does not align with the rate of accuracy degradation across tasks, leading to different levels of miscalibration. Our key contribution is to design multiple temperature parameters through a lightweight grouping network that separates samples by confidence statistics and task-aware signals in CIL. Furthermore, a Groupwise Difference-between-Confidence-and-Accuracy (GDCA) regularization is incorporated to promote intra-group consistency. Experiments demonstrate that MT-CC significantly improves calibration performance while maintaining accuracy, offering a principled framework for reliable continual learning.
Oct 18, 2026

Learning without Overwriting: Structural Knowledge Fusion for Continual Learning

Catastrophic forgetting in sequential deep learning arises from a fixed-capacity architectural constraint: parameter updates encoding new tasks perturb weight subspaces supporting prior representations. We introduce the Knowledge Fusion Network (KFN), which mitigates the stability-plasticity trade-off in class-incremental learning through novelty-gated structural expansion rather than parameter negotiation within a shared capacity budget. KFN decouples task acquisition from knowledge consolidation: each incoming task is learned in an isolated lightweight specialist network (2.8M parameters) and integrated into the global model via a Structural Fusion Module that allocates dedicated channel capacity conditioned on a quantitative novelty score. Dual structural isolation--channel-space separation with temporally phased gradient unfreezing--protects consolidated knowledge. On ten-task Split CIFAR-100, KFN achieves 11.94% minimum task accuracy and 71.16% task-10 accuracy, whereas shared-parameter baselines degrade severely: LwF to 9.07% average accuracy (0.00% min), Replay++ to 8.91% (85.08% forgetting), and Replay to 16.21% (73.44% forgetting); KFN's forgetting (57.92%) is 15.5 pp lower than Replay. Furthermore, KFN prevents logit-scale drift, maintaining near-symmetric task performance. Across CIFAR-100 and EuroSAT satellite imagery, KFN's class-incremental retention (72.7% and 42.5%) substantially exceeds the best baselines (29.6% and 0.94%). Crucially, KFN achieves parameter-efficient linear scaling, adding 25% of the base parameter count per step and requiring 25.1% of naive backbone duplication over ten tasks (28.1M vs 112.0M). With a 37.1% peak GPU memory reduction over full retraining, KFN establishes novelty-gated expansion as a practical path beyond fixed-capacity architectures.
Oct 17, 2026

An Impossibility Result for ETF Target Geometry, and an Equinorm Remedy for Imbalanced Analytic Continual Learning

Analytic continual learning (ACL) methods such as ACIL and GACL fit a ridge-regression classifier head on frozen features via the Sherman-Morrison-Woodbury identity, achieving exemplar-free, exact updates. A separate line of work, Progressive Neural Collapse (ProNC), shows standard continual classifiers benefit from a progressively expanding Simplex Equiangular Tight Frame (ETF) target. It is natural to ask whether fusing the two ideas, exact analytic updates and ETF targets, would yield a better closed-form continual learner. We show this is not the case as the fusion turns out to be identical to plain one-hot ACIL. Further, Simplex ETFs are not nested across class counts, so the literal "progressively expand the ETF" construction stops yielding an ETF as new tasks arrive, and it introduces catastrophic forgetting (4×4\times that of a classic ACIL head). We therefore separate an ETF's two defining properties, equiangularity and equinormality, and apply each on its own to the classifier weights rather than the targets, where the invariance above does not hold. Equinorm post-processing of the fitted weights recovers both tail-class and overall average accuracy across the long-tailed regime from 5:1 to 100:1 imbalance (e.g. +32.0 tail-accuracy points and +16.0 average-accuracy points at 100:1 on CIFAR-100/ViT), transferring in sign across two backbones and two datasets, at a mild cost of 5.4 average-accuracy points on CIFAR-100/ViT on near-balanced streams (below ≈3:1\approx 3{:}1). Equiangular geometry, by contrast, is exactly inert on the target side but actively harmful on the weight side, usually losing to plain one-hot. Hence, the generally useful part of Neural Collapse geometry in Analytic Continual Learning is equinorm post-processing of the weights, becoming more useful as class imbalance grows; and equiangularity is at best a no-op and induces catastrophic forgetting at worst.
Oct 15, 2026

Co2In: a Bi-level Memory Incremental Learning Framework with Knowledge Encoding, Consolidation, and Integration

Incremental learning (IL) aims to continually acquire new knowledge (plasticity) while retaining previously learned information (stability). However, striking a balance between plasticity and stability remains a significant challenge for intelligent systems. The human brain achieves exceptional balance, owing to various memory units that collaboratively encode and store information. Inspired by human memory mechanisms, this paper introduces an IL framework with knowledge enCoding, Consolidation, and Integration (Co2In). Co2In is designed as a bi-level memory architecture: a working memory for adaptive knowledge acquisition and a long-term memory dedicated to the persistent retention of information. The two memory modules work cooperatively during IL. The working memory first learns from new data to encode knowledge into parameters. Subsequently, Co2In performs a consolidation process to identify underlying patterns in the learned parameters and re-express them into a compact knowledge representation. Next, the knowledge representation and the identified patterns are transformed into separate network layers and integrated into the long-term memory. These designs empower Co2In to accumulate knowledge with high plasticity and stability. We evaluate Co2In on CIFAR-10, CIFAR-100, and Tiny-ImageNet under exemplar-free Class-IL and Task-IL settings. Experimental results show that Co2In achieves state-of-the-art performance with efficient memory consumption.
Oct 14, 2026

MAEM: Mitigating Recency Bias via a Replay-Free, Capacity-Controlled Dynamic Mixture of Experts for Class-Incremental Graph Learning

Graph Neural Networks (GNNs) excel on graph-structured data but suffer catastrophic forgetting when classes arrive sequentially. In class-incremental graph learning, predictions are often dominated by recently learned classes, causing severe recency bias. Dynamic expansion can isolate knowledge across sessions, but many methods assume task-incremental settings and require task identities at test time to select branches or heads---information unavailable in class-incremental learning. Learnable routers offer an alternative, yet continual fine-tuning the router on new data can bias routing toward recent components. We propose MAEM, a replay-free, capacity-controlled dynamic mixture-of-experts framework for class-incremental graph learning. MAEM builds representations with Simplified Graph Convolution (SGC) and trains a dedicated expert for each new session, then freezes it to protect prior knowledge. For routing without task IDs or a shared trainable router, each expert uses a masked autoencoder (MAE); test nodes are assigned to the expert with the lowest reconstruction error, reducing routing-level recency bias. When the expert pool reaches a fixed capacity, MAEM merges the least-used experts via Git Re-Basin alignment and further distills them into a fused expert using topology-aware pseudo data synthesized from MAE distributions (DeepInversion-style), without storing or revisiting historical graph data. On four class-incremental graph benchmarks, MAEM mitigates recency bias, outperforms strong continual graph learning baselines, and keeps model capacity bounded under strict replay-free learning.
Oct 14, 2026

SPARS: A Stability-Plasticity-Aware Region Selection Framework for Continual Learning

Continual learning aims to acquire new knowledge while retaining previously learned knowledge. However, updating shared parameters when learning new tasks often overwrites existing knowledge, causing catastrophic forgetting. Memory replay methods mitigate catastrophic forgetting but rely on previous samples, raising privacy, memory, and scalability concerns. On the other hand, parameter isolation-based methods preserve learned knowledge but may limit knowledge transfer, whereas regularization promotes adaptability but remains susceptible to interference between old and new tasks. To maintain the stability-plasticity trade-off, this paper proposes Stability-Plasticity-Aware Region Selection (SPARS), a replay-free hybrid framework combining global parameter isolation with localized parameter regularization. SPARS estimates parameter importance using either the Fisher Information Matrix (FIM) or a novel Confidence Score (CS) estimator that leverages logit sensitivity and achieves comparable performance to FIM without requiring labelled data. Based on parameter importance, SPARS partitions parameters into three importance-aware regions namely, winning, neighboring, and non-important parameter regions. Each region receives a different adaptation strategy: selective freezing, soft regularization, or unrestricted learning based on its role in the stability-plasticity trade-off. To mitigate classifier bias in class-incremental learning, we employ Quadratic Discriminant Analysis to improve performance on previously learned classes .Experiments on CIFAR-10 and TinyImageNet demonstrate consistent improvement in average incremental accuracy over existing isolation- and regularization-based methods across task- and class-incremental learning. On TinyImageNet, SPARS achieves the best performance with 6% improvement for class-incremental learning while retaining over 60% of parameters for future tasks.
Oct 7, 2026

Continual Learning without Continual Training

Continual learning requires models to adapt to new domains and new classes while retaining prior knowledge. Many existing methods rely on continued optimization, using regularization, replay, or parameter expansion to prevent new updates from overwriting previously learned knowledge. Instead, we propose replacing continual training with continual inference: a PFN-based model that is meta-trained, and then frozen, adapting to new classes only by extending an in-context evidence set. Our model, Latent Concept PFN, performs in-context Bayesian inference over a latent concept space that captures semantic structure shared across domains and classes. As each new domain or class arrives, exemplars are added to the memory; adaptation reflects updated posterior beliefs over latent concepts rather than gradient updates. No parameters are changed, reducing forgetting. The same method handles both domain and class incremental continual learning without task identity. Concept annotations are only used during meta-training, acting as a soft anchor on the latent space rather than a fixed bottleneck. Unlike fixed-vocabulary concept methods, the model also handles noisy, ambiguous, or incomplete annotations by combining concept labels with raw input evidence to discover distinctions beyond the predefined concept set. Experiments on class and domain incremental learning datasets demonstrate competitive continual learning performance while learning interpretable latent concepts.
Oct 5, 2026

AccentCL: Robust Accent Classification with Incremental Expansion

Accent classifiers are typically trained with a fixed label inventory and cannot accommodate new accent categories as new data becomes available. Moreover, accented speech corpora often exhibit substantial class imbalance and/or domain shift due to differences in recording conditions across corpora. We present AccentCL, a class-incremental learning framework for English accent classification that is robust to class imbalance and cross-corpus domain shift. AccentCL extracts multi-layer representations from a frozen Whisper-Large-v3 encoder, optimized with an imbalance-aware cross-entropy loss to reduce bias toward the majority accent classes and a domain mean alignment loss that minimizes distributional mean shift across training corpora. The label space is then expanded via replay-based continual learning, using the frozen base model for knowledge retention and an old-to-new margin loss to reduce overprediction on newly added classes. On a five-class accent classification task, AccentCL achieves 77.1% balanced accuracy and a 76.9% macro-averaged F1 score. We further evaluate the model's ability to incrementally incorporate two new accent categories: Spanish-accented and Chinese-accented English. When adding Spanish-accented English to the pretrained model, AccentCL attains an F1 of 83.3% on the new class while retaining 77.3% balanced accuracy on the base classes. When subsequently adding Chinese-accented English, it achieves 61.8% F1 on the new class while preserving 77.6% balanced accuracy on the previously learned classes. These results show that AccentCL enables robust regional accent classification while allowing new accent categories to be added without full retraining.
Oct 5, 2026

What Must Replay Preserve? Separating Correctable Bias from Class Correspondence

Class-incremental learning must recognize all classes seen so far without task labels. Logit replay methods such as DER and DER++ mitigate forgetting by matching the model's past predictions on stored examples. Deleting this matching reveals its benefit, but the resulting accuracy cost cannot show whether the stored scores themselves are needed, or whether the cost survives correction of the classifier's bias toward recent classes. We propose a diagnostic framework that treats a cached prediction as temporally heterogeneous supervision: it separates classes known when an example was stored from classes learned afterward, edits each group, and evaluates every model before and after a task-level offset that leaves within-task predictions unchanged. On CIFAR-100 with DER++, suitable fixed constants replace the unrefreshed stored scores of later-learned classes within an equivalence margin of 1 percentage point, and the offset reduces the cost of deleting their matching from 14.9 to 1.8 points. Reassigning the non-gold scores of classes known at storage, which preserves their values and each task's target probability, costs 4.3 points before and 4.0 after the offset, and a parallel cost persists in image distillation. In the tested fixed-head setting, the large cost of deleting later-class matching is thus mostly correctable by this offset, whereas the smaller cost of disrupting class correspondence persists. Code and data are available at anonymous.4open.science/r/replay-preserve-E22B.
Oct 4, 2026

When the Cross-Silo Federation Goes Offline: Continual Learning for Site Onboarding with Limited Unlabeled Data

An organization often holds too little labeled data to train a model that generalizes, and the records that would supply the rest sit with organizations that cannot release them. Cross-silo federated learning offers a way through, since participants exchange model parameters rather than records, but it ordinarily settles two aspects of the arrangement in advance, the participating sites and the classes the model can predict, and deployment can breach both. A new site joins after training, once the established sites have finished their engagement and gone offline, and its records arrive unlabeled, mixing conditions the model already recognizes with conditions no participant has observed. We present an autonomous three-stage procedure that expands the model entirely at the joining site: reconstruction experts screen for novelty, clustering separates the flagged records into candidate conditions, and class means describe the old classes, all inside one shared representation. Those classes were learned from records that never leave their owners, so the usual defenses against forgetting are unavailable, and the procedure supplies the evidence they would have carried from either of two dissimilar sources, prototypes held by the federation or records held by the joining site. On a real industrial condition-monitoring dataset, run end to end with no label consulted, either source holds old-class accuracy at 0.868 or above with forgetting at most 0.063, and the two differ by 0.021, so a configuration can be chosen by the disclosure it permits rather than the accuracy it delivers. Both keep old- and new-class accuracy in balance where every alternative we measure gives up one for the other, and both retain more of the old classes than distillation- and regularization-based baselines. The balance still holds with only 6 labeled records per arriving condition and 3 retained per old class.
Oct 4, 2026

Learning to Curate What You Generate for Generalizable Few-Shot Class-Incremental Learning

Few-shot class-incremental learning (FSCIL) aims to learn novel classes from limited annotations while preserving prior knowledge. Existing methods typically assume a sufficiently large base session, but this assumption fails when both base and incremental data are scarce, leading to weak initial representations, semantic drift, and unstable boundaries. We study this underexplored yet realistic setting, termed Generalizable FSCIL (G-FSCIL), where the base session itself contains only a few classes. Although synthetic data can alleviate supervision scarcity, naively mixing generated samples often introduces semantic noise and exacerbates old-new boundary conflicts. To address this, we propose a framework that curates trustworthy synthetic knowledge for stable G-FSCIL. Specifically, we first construct class-specific synthetic candidate pools using a frozen latent diffusion model, where class inversion is performed at the first observation and the resulting condition embeddings are reused for on-demand generation. Building on these candidates, we learn a knowledge curation strategy that selects samples with both semantic consistency and visual diversity, and distill this process into a transferable selection policy during the base session, which is then reused without further optimization. Leveraging the curated synthetic data, we further design a boundary-stable incremental adaptation scheme, including synthetic-informed prototype initialization and bidirectional boundary calibration to mitigate old-new conflicts. Extensive experiments demonstrate that our method consistently outperforms existing FSCIL baselines, with reduced forgetting and improved balance between old and new classes. Code is available at https://github.com/NiHaoWoJiaoYYC/G-FSCIL.
Oct 4, 2026

MAGIC: Topology-Aware Analytic Graph Few-Shot Class-Incremental Learning

Graph few-shot class-incremental learning (GFSCIL) requires a model to continually recognize emerging classes from only a few labeled nodes while preserving previously acquired knowledge. Beyond the catastrophic forgetting inherited from conventional graph continual learning, GFSCIL presents two distinctive challenges: extremely limited novel-class supervision causes severe overfitting, while cross-session edges---edges connecting newly arriving nodes with historical nodes---alter historical propagation neighborhoods and thereby induce representation drift. We propose MAGIC, a replay-free GFSCIL framework that combines a frozen graph representation backbone (e.g., an intrinsically parameter-free backbone such as SGC or a pretrained graph foundation model) with closed-form analytic continual learning. To alleviate novel-session overfitting, MAGIC learns a topological prior from the base graph that can characterize both homophilous and heterophilous relations, and injects this prior through Potts Markov random field inference to refine supervision for novel classes. To mitigate representation drift, MAGIC transfers previous predictions from the old representations of affected historical nodes to their updated representations through drift-aware analytic distillation. Experiments across five datasets and eight baselines demonstrate the effectiveness of MAGIC. Under the 5-shot setting, MAGIC improves Mean Accuracy and Final Accuracy by 5.48 percentage points and 9.33 percentage points on average, and reduces Performance Drop by 10.78 percentage points on average compared with the best baselines. MAGIC also shows clear advantages under the 1- and 3-shot settings, with larger gains as the number of supports increases. Moreover, MAGIC requires substantially less training time.
Oct 3, 2026

Gated Target Propagation for Compositional Generalization in Continual Learning

Continual learning is typically framed as acquiring new knowledge without catastrophically forgetting previous tasks. However, a flexible continual learner should also be able to reuse and recombine previously acquired knowledge to rapidly solve novel task compositions. We introduce Gated Target Propagation (GaTaP), a continual learning algorithm in which task-specific gating variables---learned through a closed-form inner loop update---selectively suppress or enhance network modules. Network parameters are learned in a slower timescale outer loop, using the same local difference target propagation error signal as is used for adapting gating variables. We provide tractable experiments on class-incremental learning scenarios for both multilayer perceptron and convolutional network architectures. We show strong performance retention on previously learned tasks, as well as compositional generalization to unseen tasks, achieved through few-shot gain adaptation at inference. We analyze learned gating patterns and find that related tasks exhibit similar gating patterns, suggesting that inferred gates capture meaningful, reusable task structure. Overall, GaTaP provides a powerful framework for jointly ameliorating catastrophic forgetting and enabling few-shot compositional generalization in neural network models.
Oct 2, 2026

Selective Backpropagation for Efficient Few-Shot Class-Incremental Learning

Few-Shot Class-Incremental Learning (FSCIL) requires models to continuously learn new classes from limited samples while retaining prior knowledge, under strict constraints on compute and memory. Existing approaches lie along a difficult trade-off: simple fine-tuning is computationally efficient but suffers from catastrophic forgetting, replay-based methods mitigate forgetting at the cost of substantial compute and memory, and exemplar-free methods often reduce forgetting by freezing most of the backbone, improving efficiency at the expense of adaptability. We propose Selective Backpropagation (SBP), a deterministic parameter budgeting framework that bridges this gap. SBP restricts gradient updates to a pre-allocated subset of network parameters, freezing past knowledge and preserving unbiased capacity for future learning, enabling rapid adaptation without costly mask optimization. We show that SBP achieves strong performance across standard FSCIL benchmarks while requiring training time close to that of naive fine-tuning and substantially lower training time than prior SOTA methods. Crucially, our experiments expose a limitation of standard FSCIL evaluation: performance on short, distribution-consistent benchmarks does not necessarily predict behavior under distribution shift or over substantially longer learning horizons. We therefore evaluate FSCIL methods in cross-domain settings and over an 80-session ImageNet-1K stream. SBP remains strong across these regimes while maintaining low training cost, providing a favorable stability-plasticity-efficiency trade-off. Our code is available at https://github.com/PaInt-Lab/sbp-main-public.
Oct 1, 2026

Diagnosing Prototype Estimation in Few-Shot Class-Incremental Learning: Training-Free Geometry Selection and Unlabeled-Query Prototype Rectification

Frozen visual encoders make prototype heads competitive in few-shot class-incremental learning (FSCIL), but it remains unclear how much of the residual head gap is sensitive to noisy novel-class means rather than to the subsequent changes introduced by classifier refitting. On identical features, we construct a retrospective matched counterfactual: first replace each five-shot novel mean with its complete-data counterpart under fixed cosine scoring, then fit a linear probe. Across sixteen strong encoder dataset pairs, mean replacement accounts for the majority of this ordered gap. Shot and classwise diagnostics support a finite-sample mechanism, whereas a weaker base-only encoder does not satisfy the same criterion. This diagnosis motivates ShrinkSel, a support-only selector over prototype radius and covariance shrinkage whose family contains shared LDA as an exact fallback. Across sixteen evaluated pairs, it improves LDA on fifteen by 0.04-1.52 points and selects the exact LDA endpoint on Aircraft. We further introduce Query Prototype Rectification (QPR), a batch-transductive extension that uses gated unlabeled queries to update novel directions while preserving the selected geometry. On CLIP ViT-B/16, QPR adds 1.21-2.03 points; exact query-wise self-exclusion retains the benefit, showing that the gains do not require self-inclusion. The update repairs scarce novel prototypes at a measurable base-class cost and weakens with plentiful supports or very small cohorts. Thus, the evidence available for prototype estimation is a central constraint in frozen-feature FSCIL.