| Total: 99
Attributing model behavior to training data is an evolving research field. A common benchmark is data removal, which involves eliminating data instances with either low or high values, then assessing a model's performance trained on the modified dataset. Many existing studies leverage Shapley-based data values for this task. In this paper, we demonstrate that these data values are not optimally suited for pruning low-value data when only a limited amount of data remains. To address this limitation, we introduce the Constraint-Data-Value-Maximization (CDVM) approach, which effectively utilizes data attributions for pruning in low-data scenarios. By casting pruning as a constrained optimization that both maximizes total influence and penalizes excessive per-test contributions, CDVM delivers robust performance when only a small fraction of the data is retained. On the OpenDataVal benchmark, CDVM shows strong performance and competitive runtime.
Emerging foundation models have spurred growing interest in task-unspecific time series modeling, which can accommodate data from diverse domains and support various tasks. However, most existing methods still suffer from poor adaptability and generalization across cross-domain time series with varying feature dimensionalities. This study achieves unified time series modeling to transcend task and data boundaries, inspired by the recent advances in visual foundation models. We propose a universal cross-modal Time series modeling method named TimeIG, featuring Tailored Temporal Imaging and Dependent Channel Grouping. Specifically, TimeIG adopts the novel dependent channel grouping method to project heterogeneous time series into hierarchical spaces to identify representative features and learn inter-dependencies across variables. Then, a tailored temporal imaging method is proposed to generate carefully hand-crafted images from time series, facilitating complementary temporal-spatial feature extraction. Finally, a dual-branch module is designed to simultaneously model the sequential and visual data, enabling cross-modal alignment via the proposed Cross-Modal Attention and Dynamic Weighted-Averaging. Extensive experiments on real-world cross-domain time series datasets show that TimeIG consistently outperforms existing SOTA methods. The code of TimeIG is publicly available at https://github.com/missCmj/TimeIG.
Accurate user behavior modeling is fundamental to the prediction of click-through rates (CTR) in industrial recommendation systems and online advertising. Traditional discriminative models, which rely on isolated ID features, struggle to capture the evolving nature of user intents across multiple channels due to the inherent semantic gap. While Large Language Models (LLMs) offer rich semantic understanding to bridge this gap, their direct application faces two main challenges: modality misalignment between continuous semantic representations and discrete item IDs, and prohibitive computational costs that make them unsuitable for real-time inference. To address these issues, we propose the Hierarchical Semantic Interest Evolution Network (HSIEN), a novel generative-discriminative framework. HSIEN leverages an LLM to construct a Markovian state-space model that recursively captures and updates user interests from multi-channel behavior sequences. It then employs a Semantic Mixture-of-Experts (SMoE) strategy to aggregate multi-channel signals into hierarchical intent representations, which are encoded as compact dynamic semantic vectors for integration with downstream discriminative models. Extensive experiments on a public dataset and a large-scale industrial dataset demonstrate that HSIEN significantly alleviates modality misalignment and enhances CTR prediction performance through feature complementarity. This work provides a practical and efficient pathway for leveraging LLMs in large-scale recommendation scenarios.
Learning universal representations for time series is fundamental for diverse downstream tasks. However, current approaches largely rely on handcrafted data augmentations, which may distort intrinsic temporal dynamics and structural regularities. In addition, most static representation learning frameworks struggle to cope with the non-stationary nature of real-world time series. To address these issues, we propose Heterogeneous Hypergraph Structure-aware Contrastive Adaptive Network (H²SCAN), a novel augmentation-free framework that derives contrastive supervision directly from graph topology. Specifically, H²SCAN constructs a heterogeneous hypergraph with three node types to capture multi-scale temporal characteristics. Building upon this representation, a meta-adaptation network is introduced to dynamically reweight heterogeneous hyperedges, enabling the model to adapt to distribution shifts in real-time. Finally, a structure-aware contrastive learning objective is employed to align latent representation similarity with the intrinsic hypergraph topology. Experiments on multiple benchmarks and cloud Kafka cluster datasets demonstrate that H²SCAN outperforms existing methods by modeling high-order multi-domain dependencies and preserving the semantic integrity of time series data.
Recent generative AI renders multimodal misinformation structurally harder to detect, making reliable detection dependent on semantic verification grounded in verifiable evidence. However, current benchmarks often fail to isolate true semantic checking from superficial shortcut exploitation. We introduce Falsdo, a diagnostic benchmark designed to make robustness and auditability identifiable. Constructed via a heterogeneous web-mining pipeline, Falsdo provides instruction-conditioned grounding and formalizes a counterfactual artifact-control protocol. This stress test reveals a critical failure: when low-level generation traces are equalized, detector performance degrades, indicating reliance on artifacts rather than robust verification. To enable reproducible diagnosis, we propose DREA, a failure-aligned evidence-auditing baseline. DREA specifies evidence acquisition with explicit noise control and implements constrained channel-wise auditing with reliability-aware late fusion. Experiments demonstrate that Falsdo reliably exposes shortcut dependence, while DREA improves instruction-level grounding (Joint-F1) and minimizes degradation under artifact-controlled settings.
Reconstruction-based frameworks are widely adopted in Time Series Anomaly Detection (TSAD), assuming that models reconstruct normal behavior well but yield larger errors on anomalies. However, in unsupervised TSAD, minimizing reconstruction loss alone often breaks this assumption. Models tend to overfit local patterns for trivial reconstruction and fail to capture the global dependencies that characterize normal behavior. Consequently, anomalies that violate global dependencies can also be reconstructed well, leading to a misalignment between reconstruction and detection. To address this challenge, we propose AnoMamba, a novel TSAD framework that aligns reconstruction with anomaly detection by enhancing global dependency modeling. Specifically, AnoMamba employs patch embedding to reduce local redundancy and introduces a Mamba variant with global step-size reweighting to select meaningful global dependencies. The reweighting process is guided by multi-scale long-tail priors, which adaptively balance global and local dependencies to mitigate overfitting in the unsupervised setting. Extensive experiments on both univariate and multivariate benchmarks demonstrate that AnoMamba consistently outperforms state-of-the-art methods in both accuracy and efficiency, while offering interpretability through its hidden attention map.
Hyper-relational Knowledge Graphs (HKGs) extend traditional knowledge graphs by introducing high-order dependencies, where auxiliary qualifiers provide detailed information. However, modeling such intricate topologies (e.g., mixtures of hierarchies, cycles, and chains) is challenging. Existing methods suffer from structural distortion due to reliance on a single geometric space and overlook dynamic semantics across entity roles. To address these issues, we propose AdaGeM (Adaptive Geometric Learning on Manifolds), an innovative space-adaptive framework that integrates geometric learning with a role-aware Transformer. First, we propose an adaptive geometric hypergraph encoder that projects entities into a product manifold and design a topology-aware gating mechanism to dynamically select optimal geometric spaces for each entity. Second, we develop a role-aware Transformer equipped with role-specific projections and micro-structural bias injection to refine embeddings by distinguishing entity semantics across different roles and focusing on valid n-ary interactions, respectively. Extensive experiments on benchmark datasets demonstrate that AdaGeM outperforms state-of-the-art baselines in entity/relation prediction tasks, with ablation studies validating the necessity of multi-manifold modeling for heterogeneous HKG structures.
Inferring latent social ties from large-scale spatiotemporal mobility traces is a foundational AI task with broad applicability. Existing hypergraph-based methods often model higher-order relations by treating hyperedges as static snapshots, thus failing to capture the temporal dynamics and co-evolution of user interactions. Meanwhile, many approaches still struggle to distinguish stable social gatherings from transient noisy co-occurrences. To address these challenges, we propose a novel Dual-Channel Hybrid Graph Neural Network (HyGNN) that jointly models temporal dynamics and high-order structural dependencies. The framework consists of two complementary components: (1) a temporal meeting graph channel that decomposes multi-user interactions into ordered snapshots and aligns trajectory evolution through message propagation, and (2) a gathering hypergraph channel that uses a structure-aware encoder with homogeneity-guided aggregation to filter noise in high-order co-occurrences. A mutual-information-aware fusion module integrates both views while preserving their distinct semantics. Extensive experiments across three real-world datasets show that HyGNN consistently outperforms state-of-the-art baselines in PRAUC and ROCAUC metrics, and robustly captures the nonlinear interplay between mobility and sociality. The code is available at https://github.com/KunLiangChen/HyGNN.
Incomplete propagation data significantly hinders robust fake news detection. Recent approaches leverage large language models to simulate missing user interactions via role-playing, thereby enriching propagation with synthetic signals. However, such propagation data is intrinsically unreliable, and directly fusing it can lead to biased representations, leading to limited detection performance. In this paper, we alleviate the unreliability of synthetic propagation from the mutual information perspective and propose a novel information-theoretic propagation denoising and fusion (InfoPDF) framework to learn effective representations from both real and synthetic propagation. Specifically, we first generate attribute-specific synthetic propagation using large language models. Then we model each synthetic propagation graph as a probabilistic latent distribution to guide reliability-aware adaptive fusion with real propagation. During training, we design a mutual information-based objective to learn compressed and task-sufficient propagation representations. It jointly suppresses noisy signals across attribute-specific synthetic propagation, maintains consistency between real and synthetic propagation representations, and ensures task sufficiency for fake news detection and attribute prediction. Experiments on three real-world datasets show that InfoPDF consistently achieves superior performance across various fake news detection tasks. Further analysis demonstrates that InfoPDF can estimate attribute-level reliabilities and learn more discriminative propagation representations.
Self-supervised heterogeneous graph learning has achieved promising results in diverse applications but still faces two issues: (i) existing methods focus on either feature similarity or meta-path to capture homophily, neglecting their inherent complementarity; (ii) existing methods rely on meta-paths to capture interactions among same-type nodes, which may introduce noise and inherently exclude long-range cross-type interactions. To address these issues, we first propose a self-expressive solver that captures the complementary homophily between meta-paths and node features to obtain homophilous representations. Meanwhile, we design separate path encoders to model diverse interactions, thus explicitly including cross-type interactions while mitigating noise via adaptive fusion. Theoretical analysis verifies that homophilous representations exhibit a high-order grouping effect to capture complementary homophily, while path encoders possess adaptive smoothness capabilities to filter noise. Extensive experiments on diverse datasets, including a large-scale dataset, demonstrate the superiority of the proposed method.
Leveraging review texts to mine deep user preferences is vital for recommendation. However, existing methods neglect the positive-negative counteraction and rely on noisy hard sentiment thresholds. Furthermore, the feature density asymmetry causes dense semantic features to overwhelm sparse collaborative signals. To address these issues, we propose the SENSE model for fine-grained sentiment-aware and structurally aligned recommendation. We employ a confidence-aware soft sentiment assignment and a sentiment exchange mechanism to quantify fine-grained preferences and explicitly model sentiment interaction. For feature fusion, we construct a collaborative-semantic graph and incorporate adaptive data augmentation. Finally, via anchor-guided cross-view alignment, we use the reliable behavior view as a topological anchor to force semantic alignment, thereby preserving semantic richness while preventing collaborative signals from being overwhelmed. Through experiments on four datasets, we verify that SENSE outperforms state-of-the-art methods. It achieves average improvements of 19.26% in MSE for rating prediction, 13.41% in Recall@10 and 13.89% in NDCG@10 for Top-N recommendation. Our source code is available at https://github.com/yingjie160/SENSE.
Entity alignment (EA) is critical for knowledge graph (KG) fusion. Existing EA models lack transferability and are incapable of aligning unseen KGs without retraining. While using graph foundation models (GFMs) offer a solution, we find that directly adapting GFMs to EA remains largely ineffective. This stems from a critical "reasoning horizon gap": unlike link prediction in GFMs, EA necessitates capturing long-range dependencies across sparse and heterogeneous KG structures. To address this challenge, we propose a EA foundation model driven by a parallel encoding strategy. We utilize seed EA pairs as local anchors to guide the information flow, initializing and encoding two parallel streams simultaneously. This facilitates anchor-conditioned message passing and significantly shortens the inference trajectory by leveraging local structural proximity instead of global search. Additionally, we incorporate a merged relation graph to model global dependencies and a learnable interaction module for precise matching. Extensive experiments verify the effectiveness of our framework, highlighting its strong generalizability to unseen KGs.
The increasing availability of trajectory data is often hampered by sparsity and noise. Existing trajectory recovery methods are further limited by either information loss from coordinate discretization into location IDs, or the inefficiency of conventional diffusion models that require a lengthy denoising process from pure noise. To address these challenges, we propose DiffVec, a novel and efficient diffusion framework for free-space trajectory recovery that operates directly on continuous coordinate data. The backbone of our framework is MVformer, a Transformer-based architecture that models motion vectors—the differences between consecutive coordinates—to better capture the local motion patterns of movement. This model is trained within our ResTraj diffusion paradigm, which commences the denoising process from a structured, interpolated prior to significantly reduce sampling steps. Extensive experiments on the Geolife and Porto datasets demonstrate that DiffVec significantly outperforms state-of-the-art baselines across MAE, MSE, and NDTW.
Attribute-aware sequential recommendation entails predicting the next item a user will interact with based on a chronologically ordered history of past interactions, enriched with item attributes. Existing methods typically leverage self-attention mechanisms to aggregate the entire sequence into a unified representation used for next-item prediction. While effective, these models often suffer from high computational complexity and memory consumption, limiting their ability to process long user histories. This constraint restricts the model's capacity to fully capture long-term user preferences. In some scenarios, modeling item interactions purely through attention may also not be the most effective approach to extract sequential patterns. In this work, we propose ConvRec, an alternative method with linear computational and memory complexity that employs convolutional layers in a hierarchical, down-scaled fashion to generate compact, yet expressive sequence representations. To further enhance the model's ability to capture diverse sequential patterns, each layer aggregates the neighboring items gradually to reach a comprehensive sequence representation. Extensive experiments on four real-world datasets demonstrate that our approach outperforms state-of-the-art sequential recommendation models, highlighting the potential of convolution-based architectures for efficient and effective sequence modeling in recommendation systems. Our implementation code and datasets are available at https://github.com/ismll-research/ConvRec.
With the increasing application of graph learning advanced by deep learning, out-of-distribution (OOD) detection for graph-structured data has become an imperative challenge in the real world. Graph neural networks (GNNs) provide a promising solution for OOD detection. However, GNNs' core message-passing mechanism inherently relies on local neighborhood aggregation, and traditional graph structures only characterize pairwise node relations, failing to capture high-order associations and global patterns. Towards this end, in this paper, we propose a Hypergraph-enhanced graph contrastive learning framework for Graph Out-Of Distribution detection (termed HGOOD). Specifically, we construct two branches to mine graph semantics in a comprehensive manner. On the one hand, we employ a graph feature branch to encode neighborhood interactions via node-level and graph-level contrastive learning. On the other hand, we incorporate the hypergraph-global branch, which adaptively models graphs’ high-order global correlations. Building upon this, we introduce a cross-branch prototype contrast that aligns the captured graph patterns with their cross-branch clustering prototypes to enhance the semantic manifold of the in-distribution (ID) graph. Extensive experiments on benchmark datasets demonstrate that our HGOOD consistently outperforms prior methods. Our code is available at https://anonymous.4open.science/r/HGOOD-CFE2/.
Continuous-Time Dynamic Graphs (CTDGs) enable fine-grained modeling of evolving relational systems. However, most existing CTDG representation learning methods are tailored to in-distribution settings and exhibit limited robustness under out-of-distribution (OOD) shifts. Although recent causal approaches learn invariant representations via interventions, they are primarily designed for static or discrete-time graphs and become computationally prohibitive for CTDGs due to the combinatorial explosion of structural and temporal variations. To address these challenges, we propose CIR, a framework grounded in a novel structural causal model termed the ICCM. To avoid exhaustive interventions, we leverage the Normalized Weighted Geometric Mean (NWGM) to efficiently approximate interventional predictions. We further instantiate ICCM within a practical deep learning architecture that jointly captures invariant structural and temporal patterns through dedicated subgraph extractors, and maintains an environment memory bank to model distributional shifts across evolving contexts. Extensive experiments demonstrate that CIR consistently outperforms existing methods under diverse OOD scenarios.
With the rapid growth of digital data, real-world applications increasingly involve hierarchical information that combines static attributes with dynamic records. Modeling such heterogeneous data in a unified and generalizable manner remains challenging. Existing approaches often rely on extensive manual design, are tightly coupled to specific data schemas, and typically process static and dynamic attributes in isolation, thereby overlooking their implicit interactions. We propose UniSAGE, a unified framework for modeling data with both static and dynamic attributes. UniSAGE constructs a global attribute graph that represents hierarchical and temporal relationships in a unified structure. To ensure representational consistency, it introduces two orthogonal parameter subspaces that jointly support static aggregation and dynamic reasoning within a shared semantic space. Building on these unified representations, UniSAGE further enables task-specific interaction between static and dynamic attributes via a lightweight hyper-structure mechanism. UniSAGE is fully automated, robust to evolving data schemas, and capable of capturing complex cross-attribute dependencies. Extensive experiments on multiple public benchmarks and a real-world financial behavior dataset demonstrate that UniSAGE consistently outperforms existing methods, achieving performance improvements of over 10% on several tasks. Our code is available at https://github.com/zjunet/UniSAGE.
Asset pricing and portfolio recommendation are two closely related fundamental tasks in quantitative investment, for which machine learning methods have attracted significant attention in both academia and industry. In particular, nonlinear asset pricing models based on deep learning architectures that learn risk factors and risk exposures (betas) conditioned on high-dimensional asset characteristics have become widely used in the field. Despite their popularity, three challenges remain for portfolio recommendation: (i) their static risk pricing structure constrains predictive performance for expected returns; (ii) their representation learning is typically deterministic or fails to account for the inherent distributional uncertainty in the feature space arising from noisy returns and heterogeneous characteristics; and (iii) the sparse factor structure of asset returns and the diversity of characteristics make it difficult to identify incremental predictive information. To address these issues and bridge the gap to practical applications, we propose a novel multi-task dynamic factor model that jointly performs asset pricing and portfolio recommendation. Specifically, we introduce dual-adversarial trainers into the variational prior-posterior learning framework for Factor and Beta Networks, augmented with probabilistic equivariance regularization and dual adaptive spatio-temporal clustering. These components filter redundant information and enhance the model's ability to adapt to changing market conditions. Extensive experiments on a comprehensive open-source stock market dataset demonstrate that our model achieves strong and robust performance relative to baseline methods in the literature.
Medication recommendation plays a critical role in clinical decision-making by supporting personalized and safe treatment planning. Existing methods rely heavily on historical co-occurrence patterns and primarily optimize discrete prescription prediction objectives, limiting generalization in rare or emerging disease settings. We propose MGRec, a framework that shifts learning from discrete prescription prediction to condition-aware modeling in a continuous molecular representation space. MGRec treats molecular structure as the primary modeling target and employs a Therapeutic--Safety Factorized Conditional Variational Autoencoder to disentangle therapeutic and safety-related factors in a condition-aware molecular latent space. The model infers treatment-relevant molecular representations conditioned on current patient-specific clinical context, which are mapped to clinically approved medications for final recommendation. To improve clinical safety, we further introduce a DDI (drug-drug interaction)-guided latent regularization to integrate drug interaction knowledge at the representation level. Experiments on two real-world benchmarks demonstrate that MGRec achieves state-of-the-art accuracy and reduced interaction risk, particularly in data-sparse scenarios.
Network traffic analysis is crucial for maintaining the security of networks. Yet deploying accurate models on edge nodes remains challenging due to protocol diversity, complex traffic behaviors, and stringent resource constraints. Although recent deep learning models achieve strong performance, their dense architectures incur high inference latency, making them unsuitable for edge deployments. Mixture-of-Experts offers a promising solution through conditional computation, yet existing methods overlook latent inter-task relations and suffer from severe expert load imbalance, leading to expert collapse and suboptimal efficiency. To address these limitations, we propose TAMoE, a novel Task-Aware Mixture-of-Experts framework for multi-task network traffic analysis. TAMoE integrates task semantics into both routing and representation learning through a novel Task-Aware Routing, enabling the router to dynamically select experts conditioned on both traffic features and task information. Besides, to train stably usable and scalable multi-task models, we further design a collaborative multi-task training strategy that encompasses both a dataset mixed construction and a joint training process. Experiments on six public benchmarks show that TAMoE achieves state-of-the-art results. Deployment on an NVIDIA® Jetson AGX Orin edge node demonstrates that TAMoE reduces inference latency by 60.6% compared to dense models and lowers the Gini Coefficient from 0.391 to 0.275 relative to Multi-Router MoE, achieving more balanced expert utilization.
This research addresses the problem of adaptive modeling in time-series data streams with clear input-output relationships. This problem is challenging because rapid system changes (regime shifts) caused by environmental factors or input delay changes degrade model performance, and the trade-off among accuracy, robustness, and memory usage arises when using multiple small models for each time-series pattern. To address these issues, this paper presents an online framework/method that treats streaming time series as dynamic mixtures of time-delay systems. This framework maintains robustness of model tracking and reduces memory usage by summarizing past regimes using a fixed-length representation that captures both the system dynamics and input-output delays. Concretely, this approach constructs a summary system tensor using the system's Markov parameter series, capturing both dynamic behavior and delay characteristics. If necessary, a tensor decomposition algorithm extracts relevant past models from the tensor and helps select the system that best fits the current regime. This method enables rapid adaptation to environmental changes and is computationally efficient. Tests on real datasets show that DelayMix consistently outperforms other methods, achieving superior forecast accuracy and faster adaptation to delays, especially for highly non-stationary data.
Edge addition is commonly considered risky in Graph Anomaly Detection (GAD), as random edge addition may induce anomaly–normal connectivity. Consequently, most existing augmentation strategies focus on feature perturbation, edge removal, or subgraph sampling, leaving edge addition largely unexplored. In contrast, we empirically find that moderate and targeted structural completion among normal nodes consistently improves GAD performance, revealing guided edge addition as an overlooked yet effective augmentation dimension. Motivated by this observation, we introduce two adjacency-centered edge generation strategies with complementary mechanisms. One performs a training-free structural completion scheme via spectrum-aware Langevin dynamics, enriching graph connectivity while preserving node features. The other models the joint evolution of node features and graph structure through a stochastic differential equation–based diffusion process, producing structurally coherent and anomaly-aware complementary graphs. Extensive experiments on 12 benchmark datasets with 7 state-of-the-art GAD models demonstrate consistent and substantial improvements in both AUROC and AUPRC. Code and appendices are available at https://github.com/GaoHaokai222/LangGen-JointSDE
Accurate trajectory prediction of multiple road users at urban intersections--including motorized and nonmotorized vehicles and pedestrians--is critical for cooperative vehicle-infrastructure systems and intelligent transportation systems. This study focuses on multiple road users' trajectory prediction in urban intersections. Such predictions face two key challenges: modeling interactions among multiple road users in complex scenarios and achieving accurate long-term prediction. Existing solutions inadequately incorporate these interactions and suffer from error accumulation or overall bias when performing long-term prediction. To overcome these challenges, we propose a long-term Trajectory prediction multi-scale AutoRegressive framework (TrajAR), which follows an encoder-decoder structure. The encoder dynamically perceives road users' motion trends, types, signals, and implicit interactions for effective handling of complex scenarios. The decoder employs a multi-scale prediction strategy that progressively refines the predicted trajectory from coarse to fine scales, where coarse-scale predictions establish long-term path backbones and finer scales enhance precision. These designs allow TrajAR to achieve accurate long-term prediction in real-world scenarios, demonstrated through experiments on four real-world urban datasets from different intersections. Our results show an average improvement of 23.8%. We provide the code of TrajAR at https://github.com/LetianGong/TrajAR.
Trajectory representation learning transforms trajectory data into low-dimensional embeddings for downstream analytics. Although trajectory data inherently contains rich spatiotemporal information that remains to be more deeply explored, recent approaches have increasingly favored integrating external multimodal information rather than deeply mining internal trajectory patterns, resulting in the underutilization of the significant potential inherent in spatiotemporal features. To address these issues, we propose SimTRL, a novel unimodal framework specifically designed for road network-based trajectory data. Specifically, we design a bi-directional mamba-traj encoder that leverages the linear computational efficiency of State Space Models to capture global non-causal topological dependencies, overcoming the limitations of traditional causal models. Furthermore, we introduce a context-aware time encoder mechanism to explicitly model dynamic time embeddings. Extensive experiments conducted on two real-world datasets demonstrate that our model achieves average performance gains of 5.19%, 13.91%, 7.2%, and 2.69% across four tasks, respectively. Compared to multimodal models, our approach achieves an average improvement of 9.06% across all evaluation metrics, with a pre-training speedup of 11.2×. The source code of our model is available at https://github.com/sxgu1/SimTRL.
Explainable recommendation is crucial for building user trust, yet producing natural-language rationales that faithfully reflect the underlying decision process remains challenging. Most LLM-based explainable recommenders incorporate collaborative signals through shallow prompting or lightweight adapters, which often yields generic explanations that are only loosely grounded in preference evidence, particularly when interactions are sparse. To address this gap, we propose HyperXRec, a novel framework that unifies preference modeling and explanation generation by integrating hyperspherical latent clustering with a cluster-guided mixture-of-experts (MoE) inside an LLM. Specifically, HyperXRec employs a dual-VAE with a von Mises-Fisher mixture prior to learn robust hyperspherical user/item embeddings, resulting in stable and semantically coherent clusters even under sparse interaction settings. These clusters are subsequently leveraged to guide fine-grained expert routing in the LLM, encouraging explanations that are both personalized and explicitly aligned with the learned preference structure. Experiments on multiple benchmarks demonstrate that HyperXRec consistently outperforms strong baselines in explanation fidelity, personalization, and robustness in cold-start scenarios.