Spamming activities (e.g., fake reviews, click farming, and deceptive content promotion) are increasingly conducted through collusive groups that exploit collective dynamics to manipulate platform metrics and mislead users, posing serious threats to the fairness, credibility, and functionality of online systems. To counteract these harmful behaviors, the task of spam detection has emerged as a critical area of research. However, existing detection methods generally remain limited in three key aspects: (i) They treat detection as a standard classification task, where representation learning and optimization are loosely coupled and suboptimal for capturing complex behaviors; (ii) They rely primarily on individual-level representation modeling, making it difficult to detect collective cheating strategies; (iii) They lack dedicated objective functions explicitly designed to characterize group-level spamming activities. To overcome these limitations, we introduce a collusion-aware Set-level learning framework (SetDet) that redefines the spam Detection task as a unified setwise optimization problem. Our approach offers three core advantages: (i) It enables end-to-end optimization by jointly learning representations and performing detection in a single, integrated process; (ii) It incorporates a model-level design for collusion representation, effectively capturing the temporal and relational patterns of coordinated spam; (iii) It pioneers a dedicated set-level optimization criterion that aligns closely with the structural characteristics of group-based cheating behaviors and accounts for class imbalance in real-world scenarios. Extensive experiments confirm the generalizability and superior performance of our framework across diverse spam scenarios and collusion strategies.

Existing S equential R ecommendation (SR) methods have conventionally viewed historical interactions as one-dimensional sequences, often overlooking the fact that user behaviors can be multi-faceted and uncertain. Such a straightforward perspective fails to account for varied behavior patterns embedded in the historical sequences. Moving beyond adhering to singular historical sequences, we treat augmented sequences as meaningful behavior patterns and jointly optimize all sequences (augmented and original sequences) to capture diverse patterns, intricate dependencies, and uncertainties. To acknowledge and distinguish new patterns derived from the original sequence, we develop a sequential order-enhanced method to calculate the edge weight, highlighting the unique dependency relationships inherent in each individual sequence. To prevent recommendations from becoming monotonous due to similarities in augmented sequences, we apply a repulsive mechanism to sequences with similar topics/categories, ensuring a broader spectrum of suggestions. Considering the potent expressive capability of the probabilistic model, Structured Determinantal Point Processes (SDPP), in representing structures, we perceive original and augmented sequences as such structures, leading to our generic learning framework S tructured S equential Rec ommendation (SSRec), which is theoretically proved to be a structured ranking optimization criterion . Comprehensive experiments on real-world datasets demonstrate SSRec’s distinct advantages over state-of-the-art models in terms of both diversity and accuracy.

Detecting small objects in UAV images is challenging due to their limited size and the complexity of background. This work introduces an improved YOLOv8nbased approach designed to enhance performance in such environments. Specifically, a feature modeling framework, named RMC (Routing-aware Modulated Context) is designed, which integrates a routing-aware attention mechanism with a multi-scale context modulation strategy to better capture and distinguish small object features. The backbone of YOLOv8n is restructured by introducing the Omni-dimensional Dynamic Convolution (ODConv) module and the Routing Attention Spatial Aggregator (RASA) module, both aimed at improving the model's sensitivity to target regions. Meanwhile, the original SPPF module is replaced by the Multi-Scale Context Modulation (MSCM) module to strengthen the modeling of complex contextual information. Further improvements are made in the neck by introducing an extra feature layer and replacing the detection head with the Adaptive Spatial Feature Fusion (ASFF) module, which is incorporated to enable effective multi-scale feature fusion. Validation on the VisDrone2019 dataset demonstrates that the RMC-YOLO method reaches 38.9%$\text{mAP}@0.5$, representing a 4.9% improvement over the original YOLOv8n.

To address key challenges in clothing quality inspection, such as short action duration, small amplitude, and difficulty in accurately locating boundaries, this paper propose a novel temporal action localization method based on multi-scale self-attention. This method employs the Transformer as the backbone network, integrating the VideoMAE feature extraction network to efficiently extract spatio-temporal features from videos, and incorporates a multi-scale selfattention mechanism. By dynamically allocating attention heads with varying frequencies and dilation rates, precise modeling of action sequences of different lengths is achieved. To enhance the localization accuracy of action boundaries, a Multi-Distribution Regression head has been designed to regress the action start, end, and center offset information separately. Simultaneously, the introduction of the Dynamic Feature Transformer (DFT) layer aims to mitigate the feature homogeneity issue and decrease the model's computational complexity. We evaluated our method on the public THUMOS14 dataset and a self-built clothing quality inspection dataset. Experiments show that our model achieved a competitive Average Precision (mAP) of 67.6 % at the current advanced level on THUMOS14. Critically, on our specialized dataset, it attains an mAP of 96.1 % with significantly lower computational overhead, demonstrating its practical effectiveness for the target application. This work presents an efficient and accurate solution for temporal action localization, paving the way for the automation and intelligence of quality control processes in the garment industry.

Sequential Recommendation (SR) focuses on personalizing user experiences by predicting future preferences based on historical interactions. Transformer models, with their attention mechanisms, have become the dominant architecture in SR tasks due to their ability to capture dependencies in user behavior sequences. However, traditional attention mechanisms, where attention weights are computed through query-key transformations, are inherently linear and deterministic. This fixed approach limits their ability to account for the dynamic and non-linear nature of user preferences, leading to challenges in capturing evolving interests and subtle behavioral patterns. Given that generative models excel at capturing non-linearity and probabilistic variability, we argue that generating attention distributions offers a more flexible and expressive alternative compared to traditional attention mechanisms. To support this claim, we present a theoretical proof demonstrating that generative attention mechanisms offer greater expressiveness and stochasticity than traditional deterministic approaches. Building upon this theoretical foundation, we introduce two generative attention models for SR, each grounded in the principles of Variational Autoencoders (VAE) and Diffusion Models (DMs), respectively. These models are designed specifically to generate adaptive attention distributions that better align with variable user preferences. Extensive experiments on real-world datasets show our models significantly outperform state-of-the-art in both accuracy and diversity.

Due to the increasing trend of performing spamming activities (e.g., Web spam, deceptive reviews, fake followers, etc.) on various online platforms to gain undeserved benefits, spam detection has emerged as a hot research issue. Previous attempts to combat spam mainly employ features related to metadata, user behaviors, or relational ties. These studies have made considerable progress in understanding and filtering spamming campaigns. However, this problem remains far from fully solved. Almost all the proposed features focus on a limited number of observed attributes or explainable phenomena, making it difficult for existing methods to achieve further improvement. To broaden the vision about solving the spam problem and address long-standing challenges (class imbalance and graph incompleteness) in the spam detection area, we propose a new attempt of utilizing signed latent factors to filter fraudulent activities. The spam-contaminated relational datasets of multiple online applications in this scenario are interpreted by the unified signed network. Two competitive and highly dissimilar algorithms of latent factors mining (LFM) models are designed based on multi-relational likelihoods estimation (LFM-MRLE) and signed pairwise ranking (LFM-SPR), respectively. We then explore how to apply the mined latent factors to spam detection tasks. Experiments on real-world datasets of different kinds of Web applications (social media and Web forum) indicate that LFM models outperform state-of-the-art baselines in detecting spamming activities. By specifically manipulating experimental data, the effectiveness of our methods in dealing with incomplete and imbalanced challenges is validated.

Diversifying recommendations to broaden user horizons and explore potential interests has become a prominent research area in recommender systems. Although numerous efforts have been made to enhance diverse recommendations, the trade-off between diversity and accuracy remains a significant challenge. The primary causes lie in the following two aspects: (i) the inherent goals of diversity-promoting recommendation, which are to simultaneously deliver accurate recommendations and cater to a broader spectrum of users’ interests, have not been adequately explored; and (ii) considering diversity only in the model training procedure cannot guarantee the provision of diversification services in recommender systems. In this work, we directly formulate the inherent goals of diversity-promoting recommendation as a dual-objective optimization problem by simultaneously minimizing the recommendation error and maximizing diversity. These proposed objectives are integrated into Generative Adversarial Nets (GANs) to guide the training process toward the orientation of boosting both diversification and accuracy. Additionally, we propose considering diversity in both training and serving phases. Experimental results demonstrate that our model outperforms others in both diversity and relevance. We extend DDPR to state-of-the-art CTR and re-ranking models, which also result in improved performance on these tasks, further demonstrating the applicability of our model in real-world scenarios.

Next Set Recommendation (NSRec), encompassing related tasks such as next basket recommendation and temporal sets prediction, stands as a trending research topic. Although numerous attempts have been made on this topic, there are certain drawbacks: (i) Existing studies are still confined to utilizing objective functions commonly found in Next Item Recommendation (NIRec), such as binary cross entropy and BPR, which are calculated based on individual item comparisons; (ii) They place emphasis on building sophisticated learning models to capture intricate dependency relationships across sequential sets, but frequently overlook pivotal dependency in their objective functions; (iii) Diversity factor within sequential sets is frequently overlooked. In this research, we endeavor to unveil a universal and Sets-level optimization framework for Next Set Recommendation (SNSRec), offering a holistic fusion of diversity distribution and intricate dependency relationships within temporal sets. To realize this, the following contributions are made: (i) We directly model the temporal set in a sequence as a cohesive entity, leveraging the Structured Determinantal Point Process (SDPP), wherein the probabilistic DPP distribution prioritizes collections of structures (sequential sets) instead of individual items; (ii) We introduce a co-occurrence representation to discern and acknowledge the importance of different sets; (iii) We propose a sets-level optimization criterion, which integrates the diversity distribution and dependency relations across the entire sequence of sets, guiding the model to recommend relevant and diversified set. Extensive experiments on real-world datasets show that our approach consistently outperforms previous methods on both relevance and diversity.

The key to personalized recommendation is to predict a personalized ranking on a catalog of items by modeling the user's preferences. There are many personalized ranking approaches for item recommendation from implicit feedback like Bayesian Personalized Ranking (BPR) and listwise ranking. Despite these methods have shown performance benefits, there are still limitations affecting recommendation performance. First, none of them directly optimize ranking of sets, causing inadequate exploitation of correlations among multiple items. Second, the diversity aspect of recommendations is insufficiently addressed compared to relevance. In this work, we present a new optimization criterion LkP based on set probability comparison for personalized ranking that moves beyond traditional ranking-based methods. It for-malizes set-level relevance and diversity ranking comparisons through a Determinantal Point Process (DPP) kernel decom-position. To confer ranking interpretability to the DPP set probabilities and prioritize the practicality of LkP, we condition the standard DPP on the cardinality$k$of the DPP-distributed set, known as k-DPP, a less-explored extension of DPP. The generic stochastic gradient descent based technique can be directly applied to optimizing models that employ LkP. We implement LkP in the context of both Matrix Factorization (MF) and neural networks approaches, on three real-world datasets, obtaining improved relevance and diversity performances. LkP is broadly applicable, and when applied to existing recommendation models it also yields strong performance improvements, suggesting that LkP holds significant value to the field of recommender systems.

Sequential Recommendation (SR) navigates users' dynamic preferences through modeling their historical interactions. The incorporation of the popular Transformer framework, which captures long relationships through pairwise dot products, has notably benefited SR. However, prevailing research in this domain faces three significant challenges: (i) Existing studies directly adopt the primary component of Transformer (i.e., the self-attention mechanism), without a clear explanation or tailored definition for its specific role in SR; (ii) The predominant focus on pairwise computations overlooks the global context or relative prevalence of item pairs within the overall sequence; (iii) Transformer primarily pursues relevance-dominated relationships, neglecting another essential objective in recommendation, i.e., diversity. In response, this work introduces a fresh perspective to elucidate the attention mechanism in SR. Here, attention is defined as dependency interactions among items, quantitatively determined under a global probabilistic model by observing the probabilities of corresponding item subsets. This viewpoint offers a precise and context-specific definition of attention, leading to the design of a distinctive attention mechanism tailored for SR. Specifically, we transmute the well-formulated global, repulsive interactions in Determinantal Point Processes (DPPs) to effectively model dependency interactions. Guided by the repulsive interactions, a theoretically and practically feasible DPP kernel is designed, enabling our attention mechanism to directly consider category/topic distribution for enhancing diversity. Consequently, the Probabilistic Attention mechanism (PAtt) for sequential recommendation is developed. Experimental results demonstrate the excellent scalability and adaptability of our attention mechanism, which significantly improves recommendation performance in terms of both relevance and diversity.

Transformer-based approaches have demonstrated remarkable success in various sequence-based tasks. However, traditional self-attention models may not sufficiently capture the intricate dependencies within items in sequential recommendation scenarios. This is due to the lack of explicit emphasis on attention weights, which play a critical role in allocating attention and understanding item-to-item correlations. To better exploit the potential of attention weights and improve the capability of sequential recommendation in learning high-order dependencies, we propose a novel sequential recommendation (SR) approach called attention weight refinement (AWRSR). AWRSR enhances the effectiveness of self-attention by additionally paying attention to attention weights, allowing for more refined attention distributions of correlations among items. We conduct comprehensive experiments on multiple real-world datasets, demonstrating that our approach consistently outperforms state-of-the-art SR models. Moreover, we provide a thorough analysis of AWRSR’s effectiveness in capturing higher-level dependencies. These findings suggest that AWRSR offers a promising new direction for enhancing the performance of self-attention architecture in SR tasks, with potential applications in other sequence-based problems as well.

Sequential recommendation is a popular task in academic research and close to real-world application scenarios, where the goal is to predict the next action(s) of the user based on his/her previous sequence of actions. In the training process of recommender systems, the loss function plays an essential role in guiding the optimization of recommendation models to generate accurate suggestions for users. However, most existing sequential recommendation tech- niques focus on designing algorithms or neural network architectures, and few efforts have been made to tailor loss functions that fit naturally into the practical application scenario of sequential recommender systems.

After exploring numerous observing system designs, the NASA aerosols, clouds, convection, and precipitation (ACCP) study team arrived at the top candidate architecture referred to as the atmosphere observing system (AOS) that is composed of a suite of spaceborne instruments in two orbital inclinations to characterize the complexity of hydrometeors and aerosols in the Earth’s atmosphere. This study proposes hybrid Bayesian retrieval algorithms that combine the Monte Carlo integration (MCI) and cost function optimization approaches to quantitatively evaluate the AOS architecture for skill in constraining the ice cloud microphysical properties. The remote sensor candidates under evaluation include multiple-frequency radars with W-, Ka-, and Ku-band channels and a submillimeter-wave radiometer. Two optimization techniques, the optimal estimation method (OEM) and Markov chain Monte Carlo (MCMC), are developed to maximize the posterior distribution function to retrieve ice cloud microphysical quantities with uncertainty estimates. Observing system simulation experiments are conducted using simulated synergistic radar and radiometer observations to determine the pixel-level retrieval accuracies by comparing the retrieved parameters to the true values. Results demonstrate that the low-frequency Ka-/Ku-band radar observations are complementary to the W-band channel since they provide more constraints on the condensed cloud scenes that are composed of large particles. The brightness temperature measurements exhibit sensitivities to the ice cloud layers with large water content (WC), and the synergistic active and passive observations improve the ice water path retrieval accuracies significantly. The scores measuring how well the AOS architecture satisfies the desired retrieval uncertainties for different ice cloud geophysical variables are also derived.

Spam activities on multifarious online platforms, such as the opinion spam and fake following relationships have been extensively studied for years. Existing works separately employ hand-crafted features --- mainly extracted from user behavior, text information, and relational network, to detect the specific spamming phenomenon on a certain kind of online platform. Although these attempts have made some headway, rapidly emerging spamming categories and frequently changing cheating strategies lead detection models to be subject to circumscribed usability and fragile effectiveness.

"Add to Favorites" is a popular function in online shopping sites which helps users to make a record of potentially interesting items for future purchases. It is usually regarded as a type of explicit feedback signal for item popularity and therefore also adopted as a ranking signal by many shopping search engines. With the increasing usage of crowdsourcing platforms, some malicious online sellers also organize crowdturfing activities to increase the numbers of "Add to Favorites" for their items. By this means, they expect the items to gain higher positions in search ranking lists and therefore boost sales. This kind of newly-appeared malicious activity proposes challenges to traditional search spam detection efforts because it involves the participation of many crowd workers who are normal online shopping users in most of the times, and these activities are composed of a series of behaviors including search, browse, click and add to favorites. To shed light on this research question, we are among the first to investigate this particular spamming activity by looking into both the task organization information in crowdsourcing platforms and the user behavior information from online shopping sites. With a comprehensive analysis of some ground truth spamming activities from the perspective of behavior, user and item, we propose a factor graph based model to identify this kind of spamming activity. Experimental results based on data collected in practical shopping search environment show that our model helps detect malicious "Add to Favorites" activities effectively.

Community Question Answering (CQA) portals provide rich sources of information on a variety of topics. However, the authenticity and quality of questions and answers (Q&As) has proven hard to control. In a troubling direction, the widespread growth of crowdsourcing websites has created a large-scale, potentially difficult-to-detect workforce to manipulate malicious contents in CQA. The crowd workers who join the same crowdsourcing task about promotion campaigns in CQA collusively manipulate deceptive Q&As for promoting a target (product or service). The collusive spamming group can fully control the sentiment of the target. How to utilize the structure and the attributes for detecting manipulated Q&As? How to detect the collusive group and leverage the group information for the detection task?

Query Auto Completion (QAC) aims to provide possible suggestions to Web search users from the moment they start entering a query, which is thought to reduce their physical and cognitive efforts in query formulation. However, the QAC has been misused by malicious users, being transformed into a new form of promotion campaign. These malicious users attack the search engines to replace legitimate auto-completion candidate suggestions with manipulated contents. Through this way, they provide a new malicious advertising service to promote their customers' products or services in QAC. To our best knowledge, we are among the first to investigate this new type of Promotion Campaign in QAC (PCQ). Firstly, we look into the causes of PCQ based on practical commercial search query logs. We found that various queries containing certain promotion intents are submitted multiple times to search engines to promote their rankings in QAC. Secondly, an effective promotion query detection framework is proposed by promotion intent propagation on query-user bipartite graph, which takes into account the behavioral characteristics of promotion campaigns. Finally, we extend the query detection framework to promotion target detection to identify the consistent promotion target which is the inherent goal of the promotion campaign. Large-scale manual annotations on practical data set convey both the effectiveness of our proposed algorithm, and an in-depth understanding of PCQ.

In the study of social networks, analyzing node influence and identifying influential nodes in social networks is of great theoretical and practical significance. To effectively evaluate node influence, a novel concept of influence label is introduced, which can measure node influence through two label attributes called influence level and node degree. Then, a novel influence model called LDM (Level and Degree Model) is proposed. LDM updates the influence label of each node iteratively by the quality of neighbors and the number of neighbors. A gain function that conforms to power-law distribution is also introduced to further optimize the accuracy of LDM. In the experiment part, to prove the effectiveness of LDM, LDM is compared with other typical methods by employing IC model to simulate the information diffusion process on four real-world social networks. Experimental results show that LDM can rank node influence more accurately than other methods.