Multimodal content is crucial for click-through rate (CTR) prediction. However, directly incorporating continuous embeddings from pre-trained models into CTR models yields suboptimal results due to misaligned optimization objectives and convergence speed inconsistency during joint training. Discretizing embeddings into semantic IDs before feeding them into CTR models offers a more effective solution, yet existing methods suffer from limited codebook utilization, reconstruction accuracy, and semantic discriminability. We propose RQ-GMM (Residual Quantized Gaussian Mixture Model), which introduces probabilistic modeling to better capture the statistical structure of multimodal embedding spaces. Through Gaussian Mixture Models combined with residual quantization, RQ-GMM achieves superior codebook utilization and reconstruction accuracy. Experiments on public datasets and online A/B tests on a large-scale short-video platform serving hundreds of millions of users demonstrate substantial improvements: RQ-GMM yields a 1.502% gain in Advertiser Value over strong baselines. The method has been fully deployed, serving daily recommendations for hundreds of millions of users.

With the growing ubiquity of recommender systems (RSs), malicious manipulation through shilling attacks, where fake user profiles are injected to alter system outputs, poses increasing threats to system integrity. Existing attack methods often rely on simplified heuristics, require internal RS data, and most overlook user reviews, limiting their stealthiness, realism, and potential impact. Recently, LLM-based user agents are gaining traction in the RS community for their capabilities to simulate human behaviors like rating and review generation. In this context, we propose AgentSA, a low-knowledge shilling attack framework that employs such agents to manipulate recommendations through adversarial yet human-like interactions. We design targeted mechanisms to guide profile construction, memory retrieval, and action generation (including reviews) to maximize manipulation impact while maintaining behavioral camouflage. We evaluate the impact of these agents on various types of RSs and demonstrate that AgentSA consistently outperforms existing low-knowledge attack methods in both effectiveness and stealth. Our findings uncover a concerning new class of threats enabled by LLM-based agents, underscoring the pressing need to bolster RS security against such emerging risks.

As recommender systems become increasingly important components in online communities, studying their impact on these communities becomes ever more crucial. Facing the high costs and ethical risks of real-world social experiments, researchers construct recommendation simulators to study the interactions between recommender systems and users. However, existing simulators face challenges in providing universal, accurate, and scalable interaction modeling for various types of online communities involving millions of contents and users with diverse action types. To address these challenges, we propose RECOSIM, a simulation framework capable of offering efficient recommendation interaction simulations across a wide range of scenarios. RECOSIM decomposes the user agent into five fundamental modules: Encode Model, Decode Model, Activity Model, Scoring Model, and Generation Model, allowing for accurate and extensible modeling of user behavior and interaction dynamics. The recommender system agent adheres to established industry architectures, implementing three stages and four fundamental strategies, thereby improving generalizability across various platforms and the computational efficiency of simulation. Utilizing two real-world datasets (Weibo and Zhihu), we validate the accuracy and stability of each component and the overall framework of RECOSIM, demonstrating the reliability of RECOSIM as a simulation environment. Subsequently, we delve into analyzing the impact of the four fundamental recommendation strategies on online communities, providing design inspirations for enhancing user engagement and community growth.

Online news platforms have become the primary source of information consumption, with recommender systems serving as critical gateways that shape public discourse through their algorithmic power, necessitating rigorous evaluation methodologies. Traditional offline evaluation methods struggle with evolving user behavior and dynamic system adaptation, while online experiments are costly, time-consuming, and ethically challenging. To address these challenges, this paper introduces EvalAgent, a large language model agent system for simulating real-world online news recommender systems. EvalAgent employs Stable Memory (StM) to model users' exploration-exploitation dynamics, mitigating noise from irrelevant interactions by analyzing the distribution density of news articles within the short-term memory, and incrementally maintains the long-term memory to capture users' high-level preferences, thereby enabling a consistent and reliable simulation of sustained interactions. It further incorporates an Environment Interaction Framework (EIF) to enable seamless engagement with real-world recommender systems. This approach yields a precise, scalable, and ethically responsible evaluation framework for news recommender systems. Comprehensive experiments and user studies substantiate EvalAgent's efficacy, with publicly available code to support ongoing research in recommender system evaluation.

Recent recommender systems aim to provide not only accurate recommendations but also explanations that help users understand them better. However, most existing explainable recommendations only consider the importance of content in reviews, such as words or aspects, and ignore the ordering relationship among them. This oversight neglects crucial ordering dimensions in the human decision-making process, leading to suboptimal performance. Therefore, in this paper, we propose Aspect Order Tree-based (AOTree) explainable recommendation method, inspired by the Order Effects Theory from cognitive and decision psychology, in order to capture the dependency relationships among decisive factors. We first validate the theory in the recommendation scenario by analyzing the reviews of the users. Then, according to the theory, the proposed AOTree expands the construction of the decision tree to capture aspect orders in users’ decision-making processes, and use attention mechanisms to make predictions based on the aspect orders. Extensive experiments demonstrate our method's effectiveness on rating predictions, and our approach aligns more consistently with the user’s decision-making process by displaying explanations in a particular order, thereby enhancing interpretability.

Recommendation algorithms rely on user historical interactions to deliver personalized suggestions, which raises significant privacy concerns. Federated recommendation algorithms tackle this issue by combining local model training with server-side model aggregation, where most existing algorithms use a uniform weighted summation to aggregate item embeddings from different client models. This approach has three major limitations: 1) information loss during aggregation, 2) failure to retain personalized local features, and 3) incompatibility with parameter-free recommendation algorithms. To address these limitations, we first review the development of recommendation algorithms and recognize that their core function is to share collaborative information, specifically the global relationship between users and items. With this understanding, we propose a novel aggregation paradigm named collaborative information aggregation, which focuses on sharing collaborative information rather than item parameters. Based on this new paradigm, we introduce the federated collaborative information aggregation (FedCIA) method for privacy-preserving recommendation. This method requires each client to upload item similarity matrices for aggregation, which allows clients to align their local models without constraining embeddings to a unified vector space. As a result, it mitigates information loss caused by direct summation, preserves the personalized embedding distributions of individual clients, and supports the aggregation of parameter-free models. Theoretical analysis and experimental results on real-world datasets demonstrate the superior performance of FedCIA compared with the state-of-the-art federated recommendation algorithms. Code is available at https://github.com/Mingzhe-Han/FedCIA.

Recommender systems leverage extensive user interaction data to model preferences; however, directly modeling these data may introduce biases that disproportionately favor popular items. In this paper, we demonstrate that popularity bias arises from the influence of propensity factors during training. Building on this insight, we propose a fair sampling (FS) method that ensures each user and each item has an equal likelihood of being selected as both positive and negative instances, thereby mitigating the influence of propensity factors. The proposed FS method does not require estimating propensity scores, thus avoiding the risk of failing to fully eliminate popularity bias caused by estimation inaccuracies. Comprehensive experiments demonstrate that the proposed FS method achieves state-of-the-art performance in both point-wise and pair-wise recommendation tasks. The code implementation is available at https://github.com/jhliu0807/Fair-Sampling.

LLM-based user agents, which simulate user interaction behavior, are emerging as a promising approach to enhancing recommender systems. In real-world scenarios, users' interactions often exhibit cross-domain characteristics and are influenced by others. However, the memory design in current methods causes user agents to introduce significant irrelevant information during decision-making in cross-domain scenarios and makes them unable to recognize the influence of other users' interactions, such as popularity factors. To tackle this issue, we propose a dual-layer memory architecture combined with a two-step fusion mechanism. This design avoids irrelevant information during decision-making while ensuring effective integration of cross-domain preferences. We also introduce the concepts of interest groups and group-shared memory to better capture the influence of popularity factors on users with similar interests. Comprehensive experiments validate the effectiveness of AgentCF++. Our code is available at https://github.com/jhliu0807/AgentCF-plus.

Due to the lack of explicit reasoning modeling, existing LLM-powered recommendations fail to leverage LLMs' reasoning capabilities effectively. In this paper, we propose a pipeline called CoT-Rec, which integrates two key Chain-of-Thought (CoT) processes-user preference analysis and item perception analysis-into LLM-powered recommendations, thereby enhancing the utilization of LLMs' reasoning abilities. CoT-Rec consists of two stages: (1) personalized information extraction, where user preferences and item perception are extracted, and (2) personalized information utilization, where this information is incorporated into the LLM-powered recommendation process. Experimental results demonstrate that CoT-Rec shows potential for improving LLM-powered recommendations. The implementation is publicly available at https://github.com/jhliu0807/CoT-Rec.

Sequential recommendation methods can capture dynamic user preferences from user historical interactions to achieve better performance. However, most existing methods only use past information extracted from user historical interactions to train the models, leading to the deviations of user preference modeling. Besides past information, future information is also available during training, which contains the ''oracle'' user preferences in the future and will be beneficial to model dynamic user preferences. Therefore, we propose an oracle-guided dynamic user preference modeling method for sequential recommendation (Oracle4Rec), which leverages future information to guide model training on past information, aiming to learn ''forward-looking'' models. Specifically, Oracle4Rec first extracts past and future information through two separate encoders, then learns a forward-looking model through an oracle-guiding module which minimizes the discrepancy between past and future information. We also tailor a two-phase model training strategy to make the guiding more effective. Extensive experiments demonstrate that Oracle4Rec is superior to state-of-the-art sequential methods. Further experiments show that Oracle4Rec can be leveraged as a generic module in other sequential recommendation methods to improve their performance with a considerable margin.

Personalized algorithms can inadvertently expose users to discomforting recommendations, potentially triggering negative consequences. The subjectivity of discomfort and the black-box nature of these algorithms make it challenging to effectively identify and filter such content. To address this, we first conducted a formative study to understand users' practices and expectations regarding discomforting recommendation filtering. Then, we designed a Large Language Model (LLM)-based tool named DiscomfortFilter, which constructs an editable preference profile for a user and helps the user express filtering needs through conversation to mask discomforting preferences within the profile. Based on the edited profile, DiscomfortFilter facilitates the discomforting recommendations filtering in a plug-and-play manner, maintaining flexibility and transparency. The constructed preference profile improves LLM reasoning and simplifies user alignment, enabling a 3.8B open-source LLM to rival top commercial models in an offline proxy task. A one-week user study with 24 participants demonstrated the effectiveness of DiscomfortFilter, while also highlighting its potential impact on platform recommendation outcomes. We conclude by discussing the ongoing challenges, highlighting its relevance to broader research, assessing stakeholder impact, and outlining future research directions.

Performing user modeling on two or more social media platforms collaboratively and complementing each other (cross-site user modeling) has been a significant problem in the area of social media mining in recent years. The core of this problem is to get to know a person’s identities on multiple platforms and then train user models collaboratively among these platforms. However, for privacy protection, many people do not want their identities on different platforms to be linked and disclosed. For this problem, we set cross-site Content Preference Prediction as a task and propose a cross-site user modeling method without cross-site User Identity Linkage (UIL). The core thought borrowed from privacy-preserving recommender system research is to organize social media identities into groups to hide the identity linkage among platforms. Experiments on real-world datasets suggest that our method outperforms the existing cross-site user modeling methods with cross-site UIL regarding several metrics.

Danmu (or “bullet screen”), a popular feature on video sharing platforms, plays a crucial role in facilitating knowledge sharing and learning. In recent years, danmu has drawn attention to automatic generation methods. However, existing methods mostly utilize limited content sources, such as the video itself (e.g., subtitles) and neighboring danmus, while other valuable sources remain underexplored. To this end, this article proposes a Category-Guided Note-to-Danmu conversion model (CG-NTD) by leveraging user-generated notes. The model is designed to identify unique contents within the notes and convert them into danmus while also showing the source note categories. CG-NTD classifies the notes by fusing them with subtitle and neighboring danmu features. Then, it uses a cross-attention mechanism to integrate the note’s category feature with note, subtitle, and danmu contexts to identify three keywords from the notes as the generated danmus. Using Bilibili as the research site, we implement a plugin prototype named BNoteToDanmu. Automatic and human evaluations reveal that CG-NTD outperforms BiLSTM, mT5, and BERT baselines in Precision, Recall, and F1-score metrics and generates more understandable and relevant danmus than ChatGPT. Moreover, the plugin demonstrates promising applications, such as assisting users in viewing videos, posting danmus, and recognizing high-quality notes. These findings offer insights into leveraging user creations to generate danmu to enhance its learning value on video sharing platforms.

Temporal recommendation methods can achieve superior accuracy due to updating user/item embeddings continuously once obtaining new interactions. However, the randomness of user behaviors will introduce noises into the user interactions and cause the deviation in the modeling of user preference, resulting in sub-optimal performance. To this end, we propose NeuFilter, a robust temporal recommendation algorithm based on neural Kalman Filtering, to learn more accurate user and item embeddings with noisy interactions. Classic Kalman Filtering is time-consuming when applied to recommendation due to its covariance matrices. Thus, we propose a neural network solution to Kalman Filtering, so as to realize higher efficiency and stronger expressivity. Specifically, NeuFilter consists of three alternating units: 1) prediction unit, which predicts user and item embeddings based on their historical embeddings; 2) estimation unit, which updates user and item embeddings in a manner similar to Kalman Filtering; 3) correction unit, which corrects the updated user and item embeddings from estimation unit to ensure reliable estimation and accurate update. Experiments on two recommendation tasks show that NeuFilter can achieve higher accuracy compared with the state-of-the-art methods, while achieving high robustness. Moreover, our empirical studies on a node classification task further confirm the importance of handling noises in tasks on temporal graph, shedding a new light on temporal graph modeling.

Graph Signal Processing (GSP) has proven to be a highly effective and efficient tool for predicting user future interactions in recommender systems. However, current GSP methods recognize user interaction patterns based on the interactions of all users, so that the recognized interaction patterns are not fully user-matched and easily impacted by other users with different interaction behaviors, resulting in sub-optimal recommendation performance. To this end, we propose a hierarchical graph signal processing method (HiGSP) for collaborative filtering, which consists of two key modules: 1) the cluster-wise filter module that recognizes user unique interaction patterns merely from interactions of users with similar preferences, making the recognized patterns able to reflect user preference without being influenced by other users with different interaction behaviors, and 2) the globally-aware filter module that serves as a complementary to the cluster-wise filter module to recognize user general interaction patterns more effectively from all user interactions. By linearly combining these two modules, HiGSP can recognize user-matched interaction patterns, so as to model user preference and predict user future interactions more accurately. Extensive experiments on six real-world datasets demonstrate the superiority of HiGSP compared to other GCN-based and GSP-based recommendation methods in terms of efficacy and efficiency.

Influence maximization (IM) has been extensively researched in the information propagation field and applied in various domains. However, existing studies on the IM have primarily focused on network structure, and lack the in-depth exploration of online network complexities, like personal history or preference. In this paper, a heterogeneous cross-comparison improved Kshell algorithm (HCCKshell) is proposed to solve IM, applying users’ multi-dimensional attributes in the propagation, including social history and topological structure . Specifically, the model learns users’ potential representation of historical content preferences and topological structure based on the Encoder and GCN thoughts, then defines the heterogeneous similarity and the heterogeneous information entropy to measure users’ influence ability and provide reliability assurance on the propagation. To improve the performance, a cross-comparison improved K-shell heuristic algorithm based on the heterogeneous information entropy is proposed to find a valid influential seed set. Furthermore, the experiments on multiple real large-scale datasets and their results indicate that our HCCKshell algorithm is more effective than baseline algorithms on both effect and performance.

Usually generated by ordinary users and often not particularly designed for learning, the videos on video-sharing platforms are mostly not structured enough to support learning purposes, although they are increasingly leveraged for that. Most existing studies attempt to structure the video using video summarization techniques. However, these methods focus on extracting information from within the video and aiming to consume the video itself. In this article, we design and implement BNoteHelper, a note-based video outline prototype that generates outline titles by extracting user-generated notes on Bilibili, using the BART model fine-tuned on a built dataset. As a browser plugin, BNoteHelper provides users with video overview and navigation as well as note-taking template, via two main features: outline table and navigation marker. The model and prototype are evaluated through automatic and human evaluations. The automatic evaluation reveals that, both before and after fine-tuning, the BART model outperforms T5-Pegasus in BLEU and Perplexity metrics. Also, the results from user feedback reveal that the generation outline sourced from notes is preferred by users over that sourced from video captions due to its more concise, clear, and accurate characteristics but also too general with less details and diversities sometimes. Two features of the video outline are also found to have respective advantages, especially in holistic and fine-grained aspects. Based on these results, we propose insights into designing a video summary from the user-generated creation perspective, customizing it based on video types, and strengthening the advantages of its different visual styles on video-sharing platforms.

Cross-site user identity linkage (UIL) aims to link the identities of the same person across different social media platforms. Social media practitioners and service providers can construct composite user portraits based on cross-site UIL, which helps understand user behavior holistically and conduct accurate recommendations and personalization. However, many social media users expect each profile to stay within the platform where it was created and thus do not want the identities of different platforms to be linked. For this problem, we first investigate the approaches people would like to use to defend against cross-site UIL and the corresponding challenges. Based on the findings, we build an adversarial framework, DeLink, based on the thoughts of adversarial text generation to help people improve their social media screen names to defend against cross-site UIL. DeLink can support both Chinese and English languages and has good generalizability to the varying numbers of social media accounts and different cross-site user identity linkage models. Extensive evaluations validate DeLink’s better performance, including a higher success rate, higher efficiency, less impact on human perception, and capability to defend against different cross-site UIL models.

Sequential recommendation demonstrates the capability to recommend items by modeling the sequential behavior of users. Traditional methods typically treat users as sequences of items, overlooking the collaborative relationships among them. Graph-based methods incorporate collaborative information by utilizing the user-item interaction graph. However, these methods sometimes face challenges in terms of time complexity and computational efficiency. To address these limitations, this paper presents AutoSeqRec, an incremental recommendation model specifically designed for sequential recommendation tasks. AutoSeqRec is based on autoencoders and consists of an encoder and three decoders within the autoencoder architecture. These components consider both the user-item interaction matrix and the rows and columns of the item transition matrix. The reconstruction of the user-item interaction matrix captures user long-term preferences through collaborative filtering. In addition, the rows and columns of the item transition matrix represent the item out-degree and in-degree hopping behavior, which allows for modeling the user's short-term interests. When making incremental recommendations, only the input matrices need to be updated, without the need to update parameters, which makes AutoSeqRec very efficient. Comprehensive evaluations demonstrate that AutoSeqRec outperforms existing methods in terms of accuracy, while showcasing its robustness and efficiency.

Click Through Rate (CTR) prediction plays an essential role in recommender systems and online advertising.It is crucial to effectively model feature interactions to improve the prediction performance of CTR models.However, existing methods face three significant challenges.First, while most methods can automatically capture high-order feature interactions, their performance tends to diminish as the order of feature interactions increases.Second, existing methods lack the ability to provide convincing interpretations of the prediction results, especially for high-order feature interactions, which limits the trustworthiness of their predictions.Third, many methods suffer from the presence of redundant parameters, particularly in the embedding layer.This paper proposes a novel method called Gated Deep Cross Network (GDCN) and a Field-level Dimension Optimization (FDO) approach to address these challenges.As the core structure of GDCN, Gated Cross Network (GCN) captures explicit high-order feature interactions and dynamically filters important interactions with an information gate in each order.Additionally, we use the FDO approach to learn condensed dimensions for each field based on their importance.Comprehensive experiments on five datasets demonstrate the effectiveness, superiority and interpretability of GDCN.Moreover, we verify the effectiveness of FDO in learning various dimensions and reducing model parameters.The code is available on https://github.com/anonctr/GDCN.

In dynamic interaction graphs, user-item interactions usually follow heterogeneous patterns, represented by different structural information, such as user-item co-occurrence, sequential information of user interactions and the transition probabilities of item pairs. However, the existing methods cannot simultaneously leverage all three structural information, resulting in suboptimal performance. To this end, we propose øurs, a triple structural information modeling method for accurate, explainable and interactive recommendation on dynamic interaction graphs. Specifically, øurs consists of 1) a dynamic ideal low-pass graph filter to dynamically mine co-occurrence information in user-item interactions, which is implemented by incremental singular value ecomposition (SVD); 2) a parameter-free attention module to capture sequential information of user interactions effectively and efficiently; and 3) an item transition matrix to store the transition probabilities of item pairs. Then, we fuse the predictions from the triple structural information sources to obtain the final recommendation results. By analyzing the relationship between the SVD-based and the recently emerging graph signal processing (GSP)-based collaborative filtering methods, we find that the essence of SVD is an ideal low-pass graph filter, so that the interest vector space in øurs can be extended to achieve explainable and interactive recommendation, making it possible for users to actively break through the information cocoons. Experiments on six public datasets demonstrated the effectiveness of øurs in accuracy, explainability and interactivity.

Many Click-Through Rate (CTR) prediction works focused on designing advanced architectures to model complex feature interactions but neglected the importance of feature representation learning, e.g., adopting a plain embedding layer for each feature, which results in sub-optimal feature representations and thus inferior CTR prediction performance. For instance, low frequency features, which account for the majority of features in many CTR tasks, are less considered in standard supervised learning settings, leading to sub-optimal feature representations. In this paper, we introduce self-supervised learning to produce high-quality feature representations directly and propose a model-agnostic Contrastive Learning for CTR (CL4CTR) framework consisting of three self-supervised learning signals to regularize the feature representation learning: contrastive loss, feature alignment, and field uniformity. The contrastive module first constructs positive feature pairs by data augmentation and then minimizes the distance between the representations of each positive feature pair by the contrastive loss. The feature alignment constraint forces the representations of features from the same field to be close, and the field uniformity constraint forces the representations of features from different fields to be distant. Extensive experiments verify that CL4CTR achieves the best performance on four datasets and has excellent effectiveness and compatibility with various representative baselines.

The collaborative filtering (CF) problem with only user-item interaction information can be solved by graph signal processing (GSP), which uses low-pass filters to smooth the observed interaction signals on the similarity graph to obtain the prediction signals. However, the interaction signal may not be sufficient to accurately characterize user interests and the low-pass filters may ignore the useful information contained in the high-frequency component of the observed signals, resulting in suboptimal accuracy. To this end, we propose a personalized graph signal processing (PGSP) method for collaborative filtering. Firstly, we design the personalized graph signal containing richer user information and construct an augmented similarity graph containing more graph topology information, to more effectively characterize user interests. Secondly, we devise a mixed-frequency graph filter to introduce useful information in the high-frequency components of the observed signals by combining an ideal low-pass filter that smooths signals globally and a linear low-pass filter that smooths signals locally. Finally, we combine the personalized graph signal, the augmented similarity graph and the mixed-frequency graph filter by proposing a pipeline consisting of three key steps: pre-processing, graph convolution and post-processing. Extensive experiments show that PGSP can achieve superior accuracy compared with state-of-the-art CF methods and, as a nonparametric method, PGSP has very high training efficiency.

CTR prediction has been widely used in the real world. Many methods model feature interaction to improve their performance. However, most methods only learn a fixed representation for each feature without considering the varying importance of each feature under different contexts, resulting in inferior performance. Recently, several methods tried to learn vector-level weights for feature representations to address the fixed representation issue. However, they only produce linear transformations to refine the fixed feature representations, which are still not flexible enough to capture the varying importance of each feature under different contexts. In this paper, we propose a novel module named Feature Refinement Network (FRNet), which learns context-aware feature representations at bit-level for each feature in different contexts. FRNet consists of two key components: 1) Information Extraction Unit (IEU), which captures contextual information and cross-feature relationships to guide context-aware feature refinement; and 2) Complementary Selection Gate (CSGate), which adaptively integrates the original and complementary feature representations learned in IEU with bit-level weights. Notably, FRNet is orthogonal to existing CTR methods and thus can be applied in many existing methods to boost their performance. Comprehensive experiments are conducted to verify the effectiveness, efficiency, and compatibility of FRNet.

To meet different social purposes, users usually share content related to the same topic or event to multiple social media platforms (cross-platform content sharing). As the differences of social norms and audiences among these social ecosystems, there are differences in the use of words and expressions in different platforms, resulting in different language styles among different platforms. In reality, it is usually difficult for users to grasp the consistency between the language style of posts to be published and that of a platform as the problem of context collapse. To address this problem, firstly, we conduct an study to investigate users' content sharing practices across two Chinese popular social media platforms (Douban and Weibo). The results indicate that: 1) there are significant linguistic differences between different platforms; 2) users' content sharing practices are personalized, and the style of their newly shared content is correlated with their historical posts. Secondly, based on the above findings, we propose a personalized cross-platform post style transfer model. The model can automatically transfer users' posts from one platform's language style to the target platform's language style, while preserving the content and reflecting users' personalized characteristics as much as possible. Experiments on the datasets collected from Douban and Weibo show that our model generally outperforms other comparison models on both style transfer and personalization metrics.

Cross-modal learning is essential to enable accurate fake news detection due to the fast-growing multimodal contents in online social communities. A fundamental challenge of multimodal fake news detection lies in the inherent ambiguity across different content modalities, i.e., decisions made from unimodalities may disagree with each other, which may lead to inferior multimodal fake news detection. To address this issue, we formulate the cross-modal ambiguity learning problem from an information-theoretic perspective and propose CAFE — an ambiguity-aware multimodal fake news detection method. CAFE consists of 1) a cross-modal alignment module to transform the heterogeneous unimodality features into a shared semantic space, 2) a cross-modal ambiguity learning module to estimate the ambiguity between different modalities, and 3) a cross-modal fusion module to capture the cross-modal correlations. CAFE improves fake news detection accuracy by judiciously and adaptively aggregating unimodal features and cross-modal correlations, i.e., relying on unimodal features when cross-modal ambiguity is weak and referring to cross-modal correlations when cross-modal ambiguity is strong. Experimental studies on two widely used datasets (Twitter and Weibo) demonstrate that CAFE outperforms state-of-the-art fake news detection methods by 2.2-18.9% and 1.7-11.4% on accuracy, respectively.

User-generated contents (UGC) in social media are the direct expression of users’ interests, preferences, and opinions. User behavior prediction based on UGC has increasingly been investigated in recent years. Compared to learning a person’s behavioral patterns in each social media site separately, jointly predicting user behavior in multiple social media sites and complementing each other (cross-site user behavior prediction) can be more accurate. However, cross-site user behavior prediction based on UGC is a challenging task due to the difficulty of cross-site data sampling, the complexity of UGC modeling, and uncertainty of knowledge sharing among different sites. For these problems, we propose a Cross-Site Multi-Task (CSMT) learning method to jointly predict user behavior in multiple social media sites. CSMT mainly derives from the hierarchical attention network and multi-task learning. Using this method, the UGC in each social media site can obtain fine-grained representations in terms of words, topics, posts, hashtags, and time slices as well as the relevances among them, and prediction tasks in different social media sites can be jointly implemented and complement each other. By utilizing two cross-site datasets sampled from Weibo, Douban, Facebook, and Twitter, we validate our method’s superiority on several classification metrics compared with existing related methods.

Graph neural networks (GNN) recently achieved huge success in collaborative filtering (CF) due to the useful graph structure information. However, users will continuously interact with items, which causes the user-item interaction graphs to change over time and well-trained GNN models to be out-of-date soon. Naive solutions such as periodic retraining lose important temporal information and are computationally expensive. Recent works that leverage recurrent neural networks to keep GNN up-to-date may suffer from the "catastrophic forgetting'' issue, and experience a cold start with new users and items. To this end, we propose the incremental graph convolutional network (IGCN) --- a pure graph convolutional network (GCN) based method to update GNN models when new user-item interactions are available. IGCN consists of two main components: 1) a historical feature generation layer, which generates the initial user/item embedding via model agnostic meta-learning and ensures good initial states and fast model adaptation; 2) a temporal feature learning layer, which first aggregates the features from local neighborhood to update the embedding of each user/item within each subgraph via graph convolutional network and then fuses the user/item embeddings from last subgraph and current subgraph via incremental temporal convolutional network. Experimental studies on real-world datasets show that IGCN can outperform state-of-the-art CF algorithms in sequential recommendation tasks.

Many popular social network sites (SNSs) provide the post-syncing functionality, which allows users to synchronize posts automatically among different SNSs. Nowadays there exists divergence on this functionality from the view of sink SNS. The key to solving this problem is to understand the characteristics of users’ post-syncing practice and goals and evaluate whether they are consistent with an SNS’s norms, cultures, and goals. However, studying and understanding the characteristics of post-syncing practice and goal are challenging tasks as a result of the difficulty of data sampling and the complexity of post-syncing behavior. In this article, we focus on investigating this question by quantitative analysis in combination with qualitative analysis. In the quantitative study, by utilizing 211,233 synced-posts sampled from Weibo, we aim to investigate characteristics of post-syncing from three perspectives: user, content, and goal. The results suggest that post-syncing plays an important role in exhibiting one’s current activities, creations, and skills as well as advertisements but involves a risk of exhibiting personal sensitive profiles. To understand the results, we present an interview-based qualitative study based on thematic analysis. It indicates that the publicity, urgency, and remarkableness of contents and differences of social affordances and social circles between sink SNS and source SNS as well as the one-time consent of post-syncing authentication jointly account for the major role of post-syncing. Based on these results, we propose insights for post-syncing functionality’s adoption, design, and promotion.