Dataset distillation compresses large datasets into compact synthetic ones to reduce storage and computational costs. Among various approaches, distribution matching (DM)-based methods have attracted attention for their high efficiency. However, they often overlook the evolution of feature representations during training, which limits the expressiveness of synthetic data and weakens downstream performance. To address this issue, we propose Trajectory Guided Dataset Distillation (TGDD), which reformulates distribution matching as a dynamic alignment process along the model’s training trajectory. At each training stage, TGDD captures evolving semantics by aligning the feature distribution between the synthetic and original dataset. Meanwhile, it introduces a distribution constraint regularization to reduce class overlap. This design helps synthetic data preserve both semantic diversity and representativeness, improving performance in downstream tasks. Without additional optimization overhead, TGDD achieves a favorable balance between performance and efficiency. Experiments on ten datasets demonstrate that TGDD achieves state-of-the-art performance, notably a 5.0% accuracy gain on high-resolution benchmarks.
论文检索
输入标题、作者或关键词，从 11,272 篇学术成果中精准定位
Spatiotemporal forecasting is a fundamental task in areas such as traffic flow prediction, environmental sensing, and urban planning. Recent advances have shown that decomposing temporal signals into multiple frequencies and modeling them jointly with spatial structures can significantly enhance forecasting performance. However, existing multifrequency forecasting models still face two critical limitations. First, the importance of different temporal frequencies evolves over time, yet most models assume fixed or static frequency contributions. Second, spatial dependencies are inherently frequency-sensitive. For instance, low-frequency components often align with global spatial patterns, while highfrequency components tend to correspond to localized interactions. However, current approaches typically use a shared spatial information across all frequencies, introducing spatiotemporal inconsistency. To address these challenges, we propose a novel Adaptive Frequency Pathways (AdaFre) for spatiotemporal forecasting, which adaptively captures both dynamic frequency relevance and frequency-aligned spatial structures. AdaFre employs a multi-frequency routing mechanism to dynamically select and aggregate the most informative temporal frequency components, while associating each with its corresponding spatial representation derived from frequency-aware embeddings. Spatiotemporal backbones are then used to model each path independently before final aggregation. Extensive experiments on several real-world datasets demonstrate that AdaFre significantly outperforms state-of-the-art baselines.
Fake orders pose increasing threats to sequential recommender systems by misleading recommendation results through artificially manipulated interactions, including click farming, context-irrelevant substitutions, and sequential perturbations. Unlike injecting carefully designed fake users to influence recommendation performance, fake orders embedded within genuine user sequences aim to disrupt user preferences and mislead recommendation results, thereby manipulating exposure rates of specific items to gain competitive advantages. To protect users' authentic interest preferences and eliminate misleading information, this paper aims to perform precise and efficient rectification on compromised sequential recommender systems while avoiding the enormous computational and time costs of retraining existing models. Specifically, we identify that fake orders are not absolutely harmful—in certain cases, partial fake orders can even have a data augmentation effect. Based on this insight, we propose Dual-view Identification and Targeted Rectification (DITaR), which primarily identifies harmful samples to achieve unbiased rectification of the system. The core idea of this method is to obtain differentiated representations from collaborative and semantic views for precise detection, and then filters detected suspicious fake orders to select truly harmful ones for targeted rectification with gradient ascent. This ensures that useful information in fake orders is not removed while preventing bias residue. Moreover, it maintains the original data volume and sequence structure, thus protecting system performance and trustworthiness to achieve optimal unbiased rectification. Extensive experiments on three datasets demonstrate that DITaR achieves superior performance compared to state-of-the-art methods in terms of recommendation quality, computational efficiency, and system robustness.
WaveDiST: A Wavelet Diffusion Transformer for Spatio-Temporal Estimation on Unobserved Locations
PDF ↗Spatio-temporal estimation plays a vital role in numerous scientific and engineering tasks, particularly for novel or unobserved locations lacking historical references. Many areas remain unobserved by sensors due to their non-core location or pending development status. The states of these areas can only be estimated through similar nodes in the geospace, rather than through historical data with temporal trends. Estimating these unobserved node states is crucial for city-wide spatio-temporal sensing and urban development, extending beyond simple point or block data imputation. In this study, we introduce a diffusion point process in high-frequency space to develop a robust spatio-temporal diffusion transformer for urban estimation where partial historical reference data is lacking. Our approach decomposes spatio-temporal data into high and low-frequency components through wavelet transform, and trains a diffusion model of spatial temporal data with a transformer that operates on high frequency signals. We incorporate low-frequency signals as diffusion conditions in the transformer architecture to capture overall spatio-temporal profiles and gradual trends. To enhance the learning of each step, we design an diffusion model featuring a spatio-temporal attention module that adaptively captures interdependencies between time and space. Extensive experiments across diverse domains including traffic, economics, and environment demonstrate that our method significantly outperforms state-of-the-art baselines.
Despite substantial progress in anomaly synthesis, existing diffusion-based and coarse inpainting pipelines commonly suffer from structural deficiencies such as micro-structural discontinuities, limited semantic controllability, and inefficient generation. To overcome these limitations, we introduce ARAS, a language-conditioned, auto-regressive anomaly synthesis approach that precisely injects local, text-specified defects into normal images via token-anchored latent editing. Leveraging a hard-gated auto-regressive operator and a training-free, context-preserving masked sampling kernel, ARAS significantly enhances defect realism, preserves fine-grained material textures, and provides continuous semantic control over synthesized anomalies. Integrated within our Quality-Aware Re-weighted Anomaly Detection (QARAD) framework, we propose a dynamic weighting strategy that emphasizes high-quality synthetic samples by computing an image-text similarity score with a dual-encoder model. Extensive experiments across three datasets, MVTec AD, VisA, and BTAD, demonstrate that our QARAD outperforms SOTA methods in both image- and pixel-level anomaly detection tasks, achieving improved accuracy, robustness, and a 5× synthesis speedup compared to diffusion-based alternatives.
With the wide adoption of online education platforms, adaptive learning systems have become increasingly important. Learning Path Recommendation (LPR) aims to dynamically adjust learning content to optimize learning efficiency based on individual student needs. However, current LPR methods suffer from sparse reward for precise assessment and only focus on anonymous sessions that overlook more personalized and effective paths. To address these challenges, we propose UNO, UNified Offline Training Paradigm for Learning Path Recommendation. This approach introduces an offline training paradigm in RL-based LPR to provide dense process rewards by a personalized advantage based on a reward model, which can estimate the students' internal knowledge levels on the learning targets. Additionally, we propose UniLPR model, a personalized recommendation system that unifies modeling the implicit relationships between students' long-term accumulation and evolving requirements for questions, and refines through Group Relative Policy Optimization(GRPO). Finally, we design learning tasks that encompass historical reviewing, recent learning, and long-term exploratory learning to simulate the comprehensive and diverse learning needs of students. Our UNO achieves state-of-the-art performance across all tasks, demonstrating its effectiveness.
Think Wise, Collaborate Effectively: A Rationale-Aware LLM-Based Recommender with Reinforcement Learning from Collaborative Signals
PDF ↗Large Language Models (LLMs) have recently emerged as powerful reasoning engines in recommender systems, generating natural-language explanations that foster user engagement. However, their recommendation performance remains limited, as they lack exposure to collaborative user-item interaction patterns. In contrast, collaborative filtering (CF) models achieve strong performance by learning from these behavioral patterns at scale. To unify the strengths of both paradigms, we propose TWiCE-Rec (Think Wise, Collaborate Effectively), a rationale-aware LLM-based recommender that incorporates collaborative user-item interactions. In the first stage, we construct a rationale dataset by applying in-context learning with self-annotated curation. A state-of-the-art LLM is guided to generate persuasive rationales that explain the causal relationship between the user’s interaction sequence and the ground-truth next item, resulting in a curated post-hoc training dataset. In the second stage, we perform multi-task instruction-tuned adaptation—based on the rationale-augmented training dataset—comprising item description generation and both non-reasoning and reasoning-based sequential recommendation, to equip the LLM with the ability to generate rationales that reflect how user preferences align with item characteristics. Finally, we aim to enhance the LLM’s recommendation performance by incorporating user-item interaction patterns derived from the CF-Rec model. To achieve this, we propose a confidence-weighted reinforcement learning strategy that adjusts rewards in proportion to both the LLM’s prediction alignment with the ground-truth and the confidence from the pretrained CF-Rec model. Our method outperforms both CF- and LLM-Rec models on Amazon datasets in terms of recommendation performance and rationale quality. In an online A/B test, it achieved about 8% higher click-through rate than existing models, demonstrating practical value.
RoSA: Enhancing Parameter-Efficient Fine-Tuning via RoPE-aware Selective Adaptation in Large Language Models
PDF ↗Fine-tuning large language models is essential for task-specific adaptation, yet it remains computationally prohibitive. Parameter-Efficient Fine-Tuning (PEFT) methods have emerged as a solution, but current approaches typically ignore the distinct roles of model components and the heterogeneous importance across layers, thereby limiting adaptation efficiency. Motivated by the observation that Rotary Position Embeddings (RoPE) induce critical activations in the low-frequency dimensions of attention states, we propose RoPE-aware Selective Adaptation (RoSA), a novel PEFT framework that allocates trainable parameters in a more targeted and effective manner. RoSA comprises a RoPE-aware Attention Enhancement (RoAE) module, which selectively enhances the low-frequency components of RoPE-influenced attention states, and a Dynamic Layer Selection (DLS) strategy that adaptively identifies and updates the most critical layers based on LayerNorm gradient norms. By combining dimension-wise enhancement with layer-wise adaptation, RoSA achieves more targeted and efficient fine-tuning. Extensive experiments on fifteen commonsense and arithmetic benchmarks demonstrate that RoSA outperforms mainstream PEFT methods under comparable trainable parameters.
Targeting Borderline Fraudsters: Multi-View Hypergraph Fraud Detection with LLM-Guided Contrastive Learning
PDF ↗Graph fraud detection (GFD) on transaction networks is crucial for safeguarding financial systems. However, due to the limited perspective of existing graph neural networks (GNNs) in the single transaction view, sophisticated fraudsters can disguise themselves to exhibit weak fraud signals, appearing as borderline fraudsters. To address this challenge, we propose MH-LGC, a multi-view hypergraph fraud detection model with large language model (LLM) guided contrastive learning. MH-LGC tackles two key limitations of existing GNN-based GFD methods: (1) Due to the local aggregation mechanism, existing methods struggle to capture high-order trading patterns among distant fraudsters. MH-LGC introduces two temporal hyper-views as complements to the transaction view and employs a Temporal Hypergraph Attention Network (THAN) to integrate the three views. (2) Most GFD methods overlook the rich semantic cues embedded in transaction data. Although some general graph learning studies have explored LLM integration, the high computational overhead and task-specific fine-tuning make them impractical for GFD tasks. MH-LGC introduces a semantic view through a fine-tuning-free LLM-Guided Contrastive learning (LGC), adopting a novel paradigm for integrating GNN and LLM to reduce the computational overhead of LLM. Extensive experiments on three real-world datasets demonstrate that MH-LGC outperforms twelve state-of-the-art baselines, with AUC improvements ranging from 1.10% to 5.70%.
The rapid advancement of video generation models has made it increasingly challenging to distinguish AI-generated videos from real ones. This issue underscores the urgent need for effective AI-generated video detectors to prevent the dissemination of false information via such videos. However, the development of high-performance AI-generated video detectors is currently impeded by the lack of large-scale, high-quality datasets specifically designed for generative video detection. To this end, we introduce GenVidBench, a challenging AI-generated video detection dataset with several key advantages: 1) Large-scale video collection: The dataset contains 6.78 million videos and is currently the largest dataset for AI-generated video detection. 2) Cross-Source and Cross-Generator: The cross-source generation reduces the interference of video content on the detection. The cross-generator ensures diversity in video attributes between the training and test sets, preventing them from being overly similar. 3) State-of-the-Art Video Generators: The dataset includes videos from 11 state-of-the-art AI video generators, ensuring that it covers the latest advancements in the field of video generation. These generators ensure that the datasets are not only large in scale but also diverse, aiding in the development of generalized and effective detection models. Additionally, we present extensive experimental results with advanced video classification models. With GenVidBench, researchers can efficiently develop and evaluate AI-generated video detection models.
Dynamic LiDAR point cloud compression (LPCC) is crucial for the efficient transmission and storage of large-scale three-dimensional data in applications such as autonomous driving. However, many existing methods, which primarily focus on compressing geometric or motion information, face a fundamental limitation: they treat all points as equally important. This approach neglects the semantic priorities of a scene, resulting in inefficient bit allocation and particularly compromising the reconstruction quality of safety-critical regions, such as pedestrians and vehicles, which are vital to downstream perception tasks. To address these limitations, we propose R²D-LPCC, a relevance-ranking framework for region adaptive LPCC that prioritizes fidelity in semantically important regions. Central to our approach is the Adaptive Relevance Learning (ARL) module, which integrates semantic context with uncertainty to evaluate regional significance and guide compression. We also introduce a Multi-scale Region-Adaptive Transform (MRAT) module to enhance semantic feature modeling and preserve fine-grained details in key areas. Additionally, we develop an adaptive multi-modal motion estimation module to improve motion prediction in complex three-dimensional environments. Extensive experiments conducted on the SemanticKITTI benchmark demonstrate that R²D-LPCC significantly surpasses ten recent state-of-the-art methods, achieving a 45.48% BD-rate gain over the previous leading method, Unicorn, and a 98.58% gain over the GPCC standard, while ensuring superior reconstruction quality in semantically important regions.
Continual learning (CL) aims to enable models to incrementally learn from a sequence of tasks without forgetting previously acquired knowledge. While most prior work focuses on closed-world settings, where all test instances are assumed from the set of learned classes, real-world applications require models to handle both CL and out-of-distribution (OOD) samples. A key insight from recent studies on deep neural networks is the phenomenon of Neural Collapse (NC), which occurs in the terminal phase of training when the loss approaches zero. Under NC, class features collapse to their means, and classifier weights align with these means, enabling effective prototype-based strategies such as nearest class mean, for both classification and OOD detection. However, in CL, catastrophic forgetting (CF) prevents the model from naturally reaching this desirable regime. In this paper, we propose a novel method called Analytic Neural Collapse (AnaNC) that analytically creates the NC properties in the feature space of a frozen pre-trained model with no training, overcoming CF. Extensive experiments demonstrate that our approach outperforms state-of-the-art methods in continual OOD detection and learning, highlighting the effectiveness of our method in this challenging scenario.
Conversational search aims to satisfy users’ complex information needs via multiple-turn interactions. The key challenge lies in revealing real users’ search intent from the context-dependent queries. Previous studies achieve conversational search by fine-tuning a conversational dense retriever with relevance judgments between pairs of context-dependent queries and documents. However, this training paradigm encounters data scarcity issues. To this end, we propose ConvMix, a mixed-criteria framework to augment conversational dense retrieval, which covers more aspects than existing data augmentation frameworks. We design a two-sided relevance judgment augmentation schema in a scalable manner via the aid of large language models. Besides, we integrate the framework with quality control mechanisms to obtain semantically diverse samples and near-distribution supervisions to combine various annotated data. Experimental results on five widely used benchmarks show that the conversational dense retriever trained by our ConvMix framework outperforms previous baseline methods, which demonstrates our superior effectiveness.
Estimating the true prevalence of an epidemic outbreak is a key public health problem. This is challenging because surveillance is usually resource intensive and biased. In the network setting, prior work on cost sensitive disease surveillance has focused on choosing a subset of individuals (or nodes) to minimize objectives such as probability of outbreak detection. Such methods do not give insights into the outbreak size distribution which, despite being complex and multi-modal, is very useful in public health planning. We introduce TESTPREV, a problem of choosing a subset of nodes which maximizes the mutual information with disease prevalence, which directly provides information about the outbreak size distribution. We show that, under the independent cascade (IC) model, solutions computed by all prior disease surveillance approaches are highly sub-optimal for TESTPREV in general. We also show that TESTPREV is hard to even approximate. While this mutual information objective is computationally challenging for general networks, we show that it can be computed efficiently for various network classes. We present a greedy strategy, called GREEDYMI, that uses estimates of mutual information from cascade simulations and thus can be applied on any network and disease model. We find that GREEDYMI does better than natural baselines in terms of maximizing the mutual information as well as reducing the expected variance in outbreak size, under the IC model.
Data condensation techniques aim to synthesize a compact dataset from a larger one to enable efficient model training, yet while successful in unimodal settings, they often fail in multimodal scenarios where preserving intricate inter-modal dependencies is crucial. To address this, we introduce ImageBindDC, a novel data condensation framework operating within the unified feature space of ImageBind. Our approach moves beyond conventional distribution-matching by employing a powerful Characteristic Function (CF) loss, which operates in the Fourier domain to facilitate a more precise statistical alignment via exact infinite moment matching. We design our objective to enforce three critical levels of distributional consistency: (i) uni-modal alignment, which matches the statistical properties of synthetic and real data within each modality; (ii) cross-modal alignment, which preserves pairwise semantics by matching the distributions of hybrid real-synthetic data pairs; and (iii) joint-modal alignment, which captures the complete multivariate data structure by aligning the joint distribution of real data pairs with their synthetic counterparts. Extensive experiments highlight the effectiveness of ImageBindDC: on the NYU-v2 dataset, a model trained on just 5 condensed datapoints per class achieves lossless performance comparable to one trained on the full dataset, achieving a new state-of-the-art with an 8.2% absolute improvement over the previous best method and more than 4× less condensation time.
Imputing missing values in spatial-temporal traffic data is essential for intelligent transportation systems. Among advanced imputation methods, score-based diffusion models have demonstrated competitive performance. These models generate data by reversing a noising process, using observed values as conditional guidance. However, existing diffusion models typically apply a uniform guidance scale across both spatial and temporal dimensions, which is inadequate for nodes with high missing data rates. Sparse observations provide insufficient conditional guidance, causing the generative process to drift toward the learned prior distribution rather than closely following the conditional observations, resulting in suboptimal imputation performance. To address this, we propose FENCE, a spatial-temporal feedback diffusion guidance method designed to adaptively control guidance scales during imputation. First, FENCE introduces a dynamic feedback mechanism that adjusts the guidance scale based on the posterior likelihood approximations. The guidance scale is increased when generated values diverge from observations and reduced when alignment improves, preventing overcorrection. Second, because alignment to observations varies across nodes and denoising steps, a global guidance scale for all nodes is suboptimal. FENCE computes guidance scales at the cluster level by grouping nodes based on their attention scores, leveraging spatial-temporal correlations to provide more accurate guidance. Experimental results on real-world traffic datasets show that FENCE significantly enhances imputation accuracy.
Knowledge Graph (KG)-supported Graph Neural Network models are becoming crucial in recommendation systems due to their ability to mitigate the data sparsity challenge. However, these models remain suboptimal because they overlook the representation differences between the inherent user-item Bipartite Graph (BG) and the external head-relation-tail KG, leading to semantic misalignment. Moreover, they indiscriminately incorporate various types of relations from the KG, which may introduce noise information into the model, ultimately degrading recommendation performance. To address these challenges, we propose an end-to-end model named Multi-graph Fusion Cross-model Contrastive Learning (MFCCL). To uncover users' interest in items and explore the associations between items, we first construct a user-interest graph by integrating information from both the BG and KG, and an item-association graph derived from the BG. We devise a multi-graph representation learning module that incorporates rich semantics into user and item representations in parallel. Simultaneously, a classical collaborative filtering module is introduced to fully leverage user-item collaborative signals. Additionally, we design a novel free data-augmentation cross-model contrastive learning to facilitate the exchange of complementary information between different models. Empirical evaluations on three widely used benchmarks demonstrate that our MFCCL method achieved significant improvements over the baselines.
Hierarchical Frequency-Decomposition Graph Neural Networks for Road Network Representation Learning
PDF ↗Road networks are critical infrastructures underpinning intelligent transportation systems and their related applications. Effective representation learning of road networks remains challenging due to the complex interplay between spatial structures and frequency characteristics in traffic patterns. Existing graph neural networks for modeling road networks predominantly fall into two paradigms: spatial-based methods that capture local topology but tend to over-smooth representations, and spectral-based methods that analyze global frequency components but often overlook localized variations. This spatial-spectral misalignment limits their modeling capacity for road networks exhibiting both coarse global trends and fine-grained local fluctuations. To bridge this gap, we propose HiFiNet, a novel hierarchical frequency-decomposition graph neural network that unifies spatial and spectral modeling. HiFiNet constructs a multi-level hierarchy of virtual nodes to enable localized frequency analysis, and employs a decomposition–updating–reconstruction framework with a topology-aware graph transformer to separately model and fuse low- and high-frequency signals. Theoretically justified and empirically validated on multiple real-world datasets across four downstream tasks, HiFiNet demonstrates superior performance and generalization ability in capturing effective road network representations.
The sparsity of user–item interactions remains a fundamental obstacle in collaborative filtering, limiting the ability of Graph Neural Network (GNN)-based recommender systems to capture high-order user relationships without incurring over-smoothing and computational overhead. Existing social recommendation approaches mitigate this by incorporating social networks, yet most rely on explicit ties and fail to construct informative links in their absence. Meanwhile, contrastive learning (CL) has shown promise in improving representation quality, but current view generation strategies, augmentation-based for robustness and nonaugmentation-based for semantic fidelity, are seldom combined, leaving their complementary potential underexplored. We propose Social Generating with Multiview-guided Tuning (SGMT), a unified framework that addresses both challenges. First, an interest-aware social generation mechanism constructs synthetic user–user links from shared interaction patterns, theoretically shown to compress collaborative paths and uncover latent high-order relations. Second, we present two complementary CL modules, Noise-augmented View and Semantic-explored View, which we theoretically prove to preferentially enhance uniformity and alignment, respectively, two fundamental objectives in CL. Experiments on three real-world datasets show that SGMT outperforms state-of-the-art baselines, validating both the theoretical analysis and the practical efficacy of our model.
LMGL-WD: LLM-Guided Multi-Task Graph Learning for Category-Level Warehouse Demand Prediction in E-Commerce
PDF ↗In warehouse-based e-commerce, accurate category-level warehouse demand prediction is essential to ensure effective inventory management. Existing works mainly explore advanced time series models to capture the temporal dynamics, failing to mine cross-category and cross-warehouse correlations effectively. In this paper, we explore large language models to understand the semantic information and fuse multi-view knowledge to enhance demand prediction. However, it is not trivial due to: i) the inaccurate LLM’s understanding of the category-related and warehouse-related textual input; and ii) the complicated cross-warehouse knowledge utilization. To solve the above challenges, we propose an LLM-guided multi-task graph learning framework, LMGL-WD, for category-level warehouse demand prediction. Specifically, LMGL-WD includes three components: i) an LLM-guided category series encoding module to represent each category through contextual and series embedding; ii) a cross-warehouse category learning module to adaptively mine the informative knowledge from cross warehouses to enhance category representation; and iii) a cross-category multi-task learning module to adaptively capture cross-category correlations to improve demand prediction. Extensive evaluation results with real-world data collected from one of the largest e-commerce platforms in China demonstrate that LMGL-WD achieves superior performance, e.g., reduces MAPE by up to 31.59%, compared to state-of-the-art methods.