Topological Data Analysis (TDA) provides artificial intelligence (AI) systems with mathematically rigorous geometric descriptors through Persistent Homology (PH), capturing essential shape characteristics in high-dimensional data. Yet, PH’s combinatorial complexity and sensitivity to outliers hinder its scalability and reliability, especially for Intrinsic PH (IPH) that relies on accurate geodesic distances. While stateof-the-art landmark-based subsampling methods, PH Landmarks, ameliorate computational costs and improve outlier robustness by selecting representative points based on local PH scores, it remain computationally intensive and at low sampling rates struggle to reconstruct the global topology. In this work, we introduce TOPOGRAPH, a simple yet powerful framework that preserves intrinsic topology. The resulting coarsened graph supports efficient IPH computations using Fermat distances. Experiments on both synthetic and realworld datasets show that TOPOGRAPH outperforms stateof-the-art sampling-based methods by achieving an order-ofmagnitude speedup and substantially improved topological fidelity in persistence diagrams, demonstrating its ability for robust and scalable topological data analysis.

Recent efforts have repurposed the Contrastive Language-Image Pre-training (CLIP) model for No-Reference Image Quality Assessment (NR-IQA) by measuring the cosine similarity between the image embedding and textual prompts such as "a good photo" or "a bad photo." However, this semantic similarity overlooks a critical yet underexplored cue: the magnitude of the CLIP image features, which we empirically find to exhibit a strong correlation with perceptual quality. In this work, we introduce a novel adaptive fusion framework that complements cosine similarity with a magnitude-aware quality cue. Specifically, we first extract the absolute CLIP image features and apply a Box-Cox transformation to statistically normalize the feature distribution and mitigate semantic sensitivity. The resulting scalar summary serves as a semantically-normalized auxiliary cue that complements cosine-based prompt matching. To integrate both cues effectively, we further design a confidence-guided fusion scheme that adaptively weighs each term according to its relative strength. Extensive experiments on multiple benchmark IQA datasets demonstrate that our method consistently outperforms standard CLIP-based IQA and state-of-the-art baselines, without any task-specific training.

Fine-grained image-text alignment is a pivotal challenge in multimodal learning, underpinning key applications such as visual question answering, image captioning, and vision-language navigation. Unlike global alignment, fine-grained alignment requires precise correspondence between localized visual regions and textual tokens, often hindered by noisy attention mechanisms and oversimplified modeling of cross-modal relationships. In this work, we identify two fundamental limitations of existing approaches: the lack of robust intra-modal mechanisms to assess the significance of visual and textual tokens, leading to poor generalization in complex scenes; and the absence of fine-grained uncertainty modeling, which fails to capture the one-to-many and many-to-one nature of region-word correspondences. To address these issues, we propose a unified approach that incorporates significance-aware and granularity-aware modeling and region-level uncertainty modeling. Our method leverages modality-specific biases to identify salient features without relying on brittle cross-modal attention, and represents region features as a mixture of Gaussian distributions to capture fine-grained uncertainty. Extensive experiments on Flickr30K and MS-COCO demonstrate that our approach achieves state-of-the-art performance across various backbone architectures, significantly enhancing the robustness and interpretability of fine-grained image-text alignment.

Knowledge tracing (KT) is a critical component of intelligent tutoring systems, which aim to track and predict students’ evolving knowledge states based on their interaction histories in online learning environments. However, most existing KT models fail to account for the periodic nature of student learning, in which knowledge acquisition occurs in concentrated study sessions, followed by periods of forgetting during breaks. To address this limitation, we propose periodic-aware knowledge tracing (PAKT), a novel model that explicitly captures these periodic learning and forgetting patterns. By leveraging Fourier transforms, PAKT identifies key periodic structures in the time intervals between student interactions, enabling the model to simulate both intraperiod learning gains and interperiod forgetting. Furthermore, we incorporate a causal self-attention mechanism to capture intricate dependencies among interactions, improving the accuracy of knowledge state estimation. Experiments on three public KT datasets demonstrate that PAKT outperforms existing methods in terms of prediction accuracy, highlighting the benefits of modeling periodicity for more effective and interpretable KT. Our results show that PAKT provides a more realistic and robust approach to personalized learning, making it a valuable tool for adaptive educational systems.

Understanding how students think—not merely what they know—remains the central challenge of knowledge tracing. Current approaches model knowledge as a flat collection of concepts and are fundamentally unable to capture the hierarchical reasoning processes that govern human problem-solving. We introduce SPARK-KT ( S tepwise P rerequisite- A ware R easoning for K nowledge T racing), a framework that reconceptualizes knowledge tracing through the lens of cognitive dependency structures. SPARK-KT leverages Large Language Models to automatically decompose each question into a Directed Acyclic Graph of reasoning steps, where nodes represent atomic cognitive operations, and edges encode prerequisite constraints. A novel dependency-aware mastery propagation mechanism ensures that a student’s competence on any step is bounded by their weakest prerequisite—mirroring the pedagogical reality that complex skills cannot be deployed without foundational mastery. Through a dual-stream architecture that disentangles conceptual knowledge from procedural structure, SPARK-KT improves predictive accuracy while providing fine-grained interpretability: it identifies not just whether a student will succeed, but which specific reasoning step constitutes their cognitive bottleneck. Experiments across three diverse educational domains demonstrate consistent improvements over sixteen state-of-the-art baselines, with particularly strong gains on multi-step reasoning problems. Beyond prediction, SPARK-KT enables actionable diagnostics—revealing latent potential masked by prerequisite gaps and informing precisely targeted interventions.

Multimodal Sentiment Analysis (MSA) faces two critical challenges: the lack of interpretability in the decision logic of multimodal fusion and modality imbalance caused by disparities in inter-modal information density. To address these issues, we propose KAN-MCP, a novel framework that integrates the interpretability of Kolmogorov-Arnold Networks (KAN) with the robustness of the Multimodal Clean Pareto (MCPareto) framework. First, KAN leverages its univariate function decomposition to achieve transparent analysis of cross-modal interactions. This structural design allows direct inspection of feature transformations without relying on external interpretation tools, thereby ensuring both high expressiveness and interpretability. Second, the proposed MCPareto enhances robustness by addressing modality imbalance and noise interference. Specifically, we introduce the Dimensionality Reduction and Denoising Modal Information Bottleneck (DRD-MIB) method, which jointly denoises and reduces feature dimensionality. This approach provides KAN with discriminative low-dimensional inputs to reduce the modeling complexity of KAN while preserving critical sentiment-related information. Furthermore, MCPareto dynamically balances gradient contributions across modalities using the purified features output by DRD-MIB, ensuring lossless transmission of auxiliary signals and effectively alleviating modality imbalance. This synergy of interpretability and robustness not only achieves superior performance on benchmark datasets such as CMU-MOSI, CMU-MOSEI, and CH-SIMS v2 but also offers an intuitive visualization interface through KAN's interpretable architecture. Our code is released on https://github.com/LuoMSen/KAN-MCP.

A critical component of fine-grained entity typing is the existence of precise relationship between entity types, such as type hierarchy. Previous approaches for fine-grained entity typing typically model the type hierarchy in vector space, which causes it extremely hard to precisely capture the complex relationship between entity types. To overcome the challenge of modeling type hierarchy in vector space, this article proposes for the first time to model the type hierarchy in high-dimensional box space. In addition, previous approaches focus more on the influence of the context of entity mentions, while neglecting the influence of entity mentions themselves. Based on the above challenges, we present a new approach called THBox, which not only successfully boosts the influence of entity mentions but also models the type hierarchy well. To verify the effectiveness of the method presented in this article, experimental results on three publicly available fine-grained entity typing benchmark datasets are provided to verify that the presented method is a new state-of-the-art solution for fine-grained entity typing.

Knowledge Tracing (KT) refers to inferring the students' knowledge mastery and predicting their future performance. KT serves as the foundation for personalized learning and enhances the effectiveness of educational interventions, becoming a crucial technology in intelligent tutoring systems. Recent approaches have demonstrated notable success by harnessing the potent representational capacities of deep learning. However, complex neural networks lead to entangled knowledge state embeddings, where the embedding dimensions are coupled, limiting their expressiveness and interpretability. In addition, the limitations of existing methods in Euclidean space result in distortions when capturing complex relationships among knowledge states. This distortion manifests as the distance and geometric structures among knowledge states being deformed during the embedding process. To address the challenges, in this paper, we propose a hyperbolic hypergraph transformer with knowledge stateDisentanglement forKnowledgeTracing, named DisenKT. We construct the students' response sequences into the hypergraph, projected into the hyperbolic space to alleviate the representation distortion problem of questions and knowledge states. The embeddings of hierarchical knowledge states are refined through message passing between questions and students based on the proposed hyperbolic hypergraph transformer. Moreover, we are the first to disentangle knowledge states via a contrastive clustering auxiliary task, which enhances the expressiveness and interpretability of knowledge state embeddings. Extensive experimental results on three public datasets demonstrate that DisenKT outperforms state-of-the-art methods on student performance prediction and interpretability.

Knowledge tracing (KT) refers to predicting learners' performance in the future according to their historical responses, which has become an essential task in intelligent tutoring systems. Most deep learning-based methods usually model the learners' knowledge states via recurrent neural networks (RNNs) or attention mechanisms. Recently emerging graph neural networks (GNNs) assist the KT model to capture the relationships such as question-skill and question-learner. However, non-pairwise and complex higher-order information among responses is ignored. In addition, a single-channel encoded hidden vector struggles to represent multigranularity knowledge states. To tackle the above problems, we propose a novel KT model named dual-channel adaptive scale hypergraph encoders with cross-view contrastive learning (HyperKT). Specifically, we design an adaptive scale hyperedge distillation component for generating knowledge-aware hyperedges and pattern-aware hyperedges that reflect non-pairwise higher-order features among responses. Then, we propose dual-channel hypergraph encoders to capture multigranularity knowledge states from global and local state hypergraphs. The encoders consist of a simplified hypergraph convolution network and a collaborative hypergraph convolution network. To enhance the supervisory signal in the state hypergraphs, we introduce the cross-view contrastive learning mechanism, which performs among state hypergraph views and their transformed line graph views. Extensive experiments on three real-world datasets demonstrate the superior performance of our HyperKT over the state-of-the-art (SOTA).

Graph Structure Learning (GSL), with its ability to simultaneously optimize the graph structure and learn the suitable parameters of Graph Neural Networks (GNN), has attracted considerable attention. Although existing methods are capable of learning optimal graph structure from single or multiple information sources, they exhibit certain limitations such as disregarding intricate relational information in the original graph structure and bias introduced by a small number of labels. In this study, we introduce a novel contrastive multi-view method for graph structure learning, named CMVGSL, which estimates graph structure suited for GNN properties from a broader range of perspectives. Specifically, we extract a k truss subgraph as an augmented view for contrastive learning. The representations learned from both semi-supervised and contrastive learning are utilized to construct observations of the optimal graph, and then the estimated graph structure is obtained by employing a Bayesian inference-based method. Furthermore, we design a post-processing method to impose sparsity and internode distance constraints on the estimated graph. The joint optimization of graph structure and GNN parameters is achieved through multiple iterations. Extensive experimental results on benchmark datasets with different homophily demonstrate the significant effectiveness of our proposed CMVGSL.

Different from traditional vector-based fine-grained entity typing methods, the box-based method is more effective in capturing the complex relationships between entity mentions and entity types. The box-based fine-grained entity typing method projects entity types and entity mentions into high-dimensional box space, where entity types and entity mentions are embedded asd-dimensional hyperrectangles. However, the impacts of entity types are not considered during classification in high-dimensional box space, and the model cannot be optimized precisely when two boxes are completely separated or overlapped in high-dimensional box space. Based on the above shortcomings, anImprovedBoxEmbeddings (IBE) method for fine-grained entity typing is proposed in this work. The IBE not only introduces the impacts of entity types during classification in high-dimensional box space, but also proposes a distance based module to optimize the model precisely when two boxes are completely separated or overlapped in high-dimensional box space. Experimental results on four fine-grained entity typing datasets verify the effectiveness of the proposed IBE, demonstrating that IBE is a state-of-the-art method for fine-grained entity typing.