prototype selection

This work addresses the high memory overhead and privacy concerns of conventional replay-based methods in online continual learning, which typically require storing large volumes of historical data. To mitigate catastrophic forgetting under stringent memory constraints, the authors propose a prototype-based compressed replay strategy that synthesizes a small set of representative prototype samples per class and augments them via a perturbation mechanism to generate diverse synthetic variants. By integrating prototype synthesis, feature extraction, and perturbation-based augmentation, the method drastically reduces storage requirements while preserving data privacy. Extensive experiments on multiple benchmarks and large-scale multitask settings demonstrate that the approach consistently outperforms existing replay techniques—even when retaining only a minimal number of samples per class—thereby achieving superior performance with significantly lower memory consumption.

This work addresses the low efficiency and poor structural preservation inherent in prototype selection for large-scale datasets. We propose TPS, a topology-aware prototype selection framework grounded in topological data analysis (TDA). TPS leverages persistent homology to characterize the intrinsic geometry and connectivity structure of data, enabling adaptive identification of topologically salient samples as prototypes; it further supports parallel implementation. Compared with conventional methods, TPS achieves substantial data compression—retaining only 5–15% of samples on multiple synthetic and real-world benchmarks—while maintaining or improving classification accuracy by 1.2–3.8 percentage points. The approach also exhibits strong interpretability and robustness. Its core innovation lies in the first systematic integration of TDA’s structural awareness into prototype selection, thereby unifying computational efficiency, structural fidelity, and interpretability.

This work addresses the challenge of effectively balancing sample representativeness and diversity in dynamic data selection to accelerate training while preserving model accuracy. The authors propose a novel framework that defines representativeness as coverage of high-frequency feature factors in the dataset, while diversity is achieved by progressively introducing complementary rare factors during training. Leveraging sparse autoencoders, the method identifies sparsely activated units in the feature space and estimates both sample-level and dataset-level factor distributions. A frequency-based penalty mechanism combined with a smooth scheduling strategy enables efficient, gradient-free data selection. Evaluated across five vision and language benchmarks, the approach matches or exceeds the performance of full-data training while achieving over 2× speedup in training time.

Current supervised fine-tuning (SFT) lacks efficient and interpretable data selection methods; blindly scaling data volume often fails to improve performance while incurring high preprocessing and validation costs. This paper proposes KAMIR, a data selection framework that evaluates sample familiarity based on internal model representations. Its core innovation lies in quantifying knowledge familiarity via layer-wise hidden-state similarity to the final output state—eliminating reliance on prompt engineering. A lightweight classifier identifies low-familiarity samples, enabling task-agnostic, few-shot-friendly training data filtering. Evaluated across diverse tasks—including machine reading comprehension and summarization—KAMIR achieves significant generalization gains using only a small subset of low-familiarity samples for fine-tuning. Empirical results validate the efficacy of the “less-but-better” data curation paradigm, demonstrating that targeted selection of semantically challenging instances yields superior performance over brute-force data expansion.

Automatic schema matching (ASM) suffers from low matching quality and excessive human intervention due to inherent complexity and uncertainty. Method: This study pioneers modeling ASM as a complex adaptive system (CAS) and introduces an agent-based modeling and simulation (ABMS) approach to construct a system-level matching framework exhibiting emergence and synergy. Departing from conventional local-rule-driven paradigms, the framework employs biologically inspired design and systems thinking to enable a paradigm shift—from atomic, isolated matching to global, self-organized coordination. Contribution/Results: The prototype tool Reflex-SMAS, built upon this framework, demonstrates significant improvements in matching accuracy and robustness across diverse scenarios. Empirical evaluation shows a reduction of over 62% in manual verification effort, confirming the dual advantages of the systemic approach: enhanced performance and substantial labor-cost savings.

This work addresses the degradation of embedding space coherence and model performance in federated learning caused by premature alignment of immature prototypes, particularly under highly non-IID data distributions. To mitigate this issue, the authors propose FedSAP, a novel framework that integrates a delayed-alignment curriculum scheduling mechanism with a geometry-driven proxy separation loss on the unit hypersphere. This approach enhances intra-class compactness and inter-class separability without increasing communication overhead. As the first study to formalize scheduled alignment as a general design principle in federated prototype learning, FedSAP stabilizes representation learning without introducing additional parameters and naturally extends to semi-supervised settings. Extensive experiments demonstrate consistent improvements, with up to a 4-percentage-point gain over state-of-the-art methods across three benchmark datasets, especially excelling in high-heterogeneity scenarios.

Existing prototype alignment methods in heterogeneous federated learning enforce clients with diverse architectures to align within a unified feature subspace, thereby constraining model expressiveness. This work proposes FedSAF, a novel structural alignment paradigm that shifts the alignment objective from coordinate-wise matching to preserving the consistency of inter-class relational structures. By decoupling semantic structure alignment from shared feature bases, FedSAF models class relationships through prototypes and integrates them into a distributed optimization framework. Extensive experiments demonstrate that FedSAF significantly outperforms current heterogeneous federated learning approaches across multiple benchmarks, achieving accuracy improvements of up to 3.52%.

This work proposes an unsupervised Example Partitioning (EP) method to efficiently construct interpretable and computationally lightweight feature dictionaries for analyzing the internal mechanisms of large language models. EP leverages real samples from streaming activation data as Voronoi region anchors, eliminating the need to predefine dictionary size, and uniquely employs observed activations directly as both intervention directions and region representatives. This enables feature alignment across layers, models, and training stages while inherently supporting out-of-distribution detection. By integrating distance-threshold leader clustering with causal interventions, EP achieves superior performance on Gemma-2-2B, surpassing GemmaScope SAE’s AxBench AUROC (0.881) at only one-thousandth of the computational cost, retaining 97% probe accuracy, and exhibiting high consistency with SAE features in 20% of its regions.

This work addresses a critical yet overlooked limitation of modern generative models: their tendency to produce class-typical samples at the expense of intra-class diversity, thereby diminishing the utility of synthetic data in downstream tasks. The study is the first to formally characterize this structural bias and introduces a novel post-hoc filtering mechanism that requires neither retraining nor generator-specific modifications. By partitioning real classes into homogeneous typical (HO) and heterogeneous non-redundant (HE) subsets, the method selects high-quality synthetic samples through a fidelity–diversity criterion that combines semantic alignment scores with redundancy penalties. Evaluated across multiple benchmarks, the approach consistently outperforms existing data selection strategies—achieving performance on par with real data using only 60% of the synthetic samples—and provides consistent gains even when applied to strong generative models in both classification and segmentation tasks.

This work addresses the lack of structured, verifiable, and reusable decision mechanisms in existing automated machine learning approaches for model selection. It proposes a semantic task profiling–based structured agent framework that leverages retrieval-augmented generation of historical cases and code modules to construct an intermediate representation blueprint encompassing modeling components, composition logic, and execution constraints. By integrating code execution feedback with a failure-aware reinforcement learning strategy, the framework enables memory-driven, traceable, multi-stage search optimization. Evaluated on financial time-series forecasting and generation tasks, the method significantly outperforms both conventional AutoML systems and current agent-based baselines, achieving consistent improvements in task performance, execution success rate, and decision interpretability.