Defending large language models (LLMs) against jailbreak attacks is crucial for ensuring their safe deployment. Existing defense strategies typically rely on predefined static criteria to differentiate between harmful and benign prompts. However, such rigid rules fail to accommodate the inherent complexity and dynamic nature of real-world jailbreak attacks. In this paper, we focus on the novel challenge of adaptive defense against diverse jailbreaks. We propose a new concept "mirror'', which is a dynamically generated prompt that reflects the syntactic structure of the input while ensuring semantic safety. The discrepancies between input prompts and their corresponding mirrors serve as guiding principles for defense. A novel defense model, MirrorShield, is further proposed to detect and calibrate risky inputs based on the crafted mirrors. Evaluated on multiple benchmark datasets and compared against ten state-of-the-art attack methods, MirrorShield demonstrates superior defense performance and promising generalization capabilities.

Large language model-based multi-agent systems (LLM-MAS) effectively accomplish complex and dynamic tasks through inter-agent communication, but this reliance introduces substantial safety vulnerabilities. Existing attack methods targeting LLM-MAS either compromise agent internals or rely on direct and overt persuasion, which limit their effectiveness, adaptability, and stealthiness. In this paper, we propose MAST, a Multi-round Adaptive Stealthy Tampering framework designed to exploit communication vulnerabilities within the system. MAST integrates Monte Carlo Tree Search with Direct Preference Optimization to train an attack policy model that adaptively generates effective multi-round tampering strategies. Furthermore, to preserve stealthiness, we impose dual semantic and embedding similarity constraints during the tampering process. Comprehensive experiments across diverse tasks, communication architectures, and LLMs demonstrate that MAST consistently achieves high attack success rates while significantly enhancing stealthiness compared to baselines. These findings highlight the effectiveness, stealthiness, and adaptability of MAST, underscoring the need for robust communication safeguards in LLM-MAS.

Large Reasoning Models (LRMs) can exhibit step-by-step reasoning, reflection, and backtracking, but these behaviors are often unregulated, leading to overthinking.As a result, LRMs continue generating redundant reasoning even after reaching high-confidence conclusions.This increases inference cost and latency, limiting practical deployment.The root cause is the absence of an intrinsic mechanism to monitor the reasoning state and decide when to continue, backtrack, or stop.We propose MERA, a meta-cognitive reasoning framework that decouples reasoning from control to enable independent optimization of control strategies.MERA constructs high-quality reasoning-control supervision data via a takeoverbased pipeline, and transforms long-horizon traces into structured reasoning-control alternating sequences for training.The model is trained with supervised fine-tuning to internalize the structured separation, and further optimized with Control-Segment Policy Optimization (CSPO), which combines segment-wise GRPO with control masking to focus learning on control segments.Experiments across reasoning benchmarks show that MERA improves both efficiency and accuracy.

Defending large language models (LLMs) against jailbreak attacks is essential for their safe and reliable deployment.Existing defenses often rely on shallow pattern matching, which struggles to generalize to novel and unseen attack strategies.To address this challenge, we propose the Cognitive-Driven Defense (CDD) framework, which targets the underlying structure of jailbreak prompts by applying metaoperations, defined as basic manipulations that conceal harmful intent.CDD emulates human cognitive reasoning through a structured reasoning chain.It begins with a global perception of the prompt and follows with a localized analysis to uncover hidden manipulations.By applying supervised fine-tuning on this structured chain, the model learns to identify and reason about known manipulation patterns.To enhance generalization to unseen threats, an entropy-guided reinforcement learning algorithm (EG-GRPO) is introduced to encourage exploration of new types and variants of meta-operations.Experiments demonstrate that CDD can achieve state-of-the-art defense performance and exhibit strong generalization to unseen jailbreak attacks.

Bingyu Yan, Xiaoming Zhang, JinYu Hou, Chaozhuo Li, Ziyi Zhou, Yiming Hei, Litian Zhang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Graph neural networks (GNNs) have achieved significant success in graph representation learning and have been applied to various domains. However, many real-world graphs contain sensitive personal information, such as user profiles in social networks, raising serious privacy concerns when graph learning is performed using GNNs. To address this issue, locally private graph learning protocols have gained considerable attention. These protocols leverage the privacy advantages of local differential privacy (LDP) and the effectiveness of GNN's message-passing in calibrating noisy data, offering strict privacy guarantees for users' local data while maintaining high utility (e.g., node classification accuracy) for graph learning. Despite these advantages, such protocols may be vulnerable to data poisoning attacks, a threat that has not been considered in previous research. Identifying and addressing these threats is crucial for ensuring the robustness and security of privacy-preserving graph learning frameworks. This work introduces the first data poisoning attack targeting locally private graph learning protocols. The attacker injects fake users into the protocol, manipulates these fake users to establish links with genuine users, and sends carefully crafted data to the server, ultimately compromising the utility of private graph learning. The effectiveness of the attack is demonstrated both theoretically and empirically. In addition, several defense strategies have also been explored, but their limited effectiveness highlights the need for more robust defenses.

WiFi sensing advancements facilitate the capture of human gestures from wireless signals, ensuring both privacy preservation and robustness under low-light conditions. Deep learning-based WiFi Human Gesture Recognition (HGR) demonstrates remarkable performance in handling complex gestures. To reduce labeling efforts, recent years have seen the emergence of semi-supervised WiFi HGR, leveraging massive amounts of unlabeled data. However, existing semi-supervised schemes often assume a balanced class distribution and utilize a fixed threshold for selecting pseudo-labels of unlabeled samples, leading to low performance for minority classes and decreased model generalization on real-world imbalanced datasets. To address this issue, we propose a novel semi-supervised WiFi HGR approach with dynamic pseudo-labeling thresholds to handle imbalanced class distribution, incorporating Spatial-Temporal Attention (STA) networks. Unlike using a fixed threshold for all unlabeled samples, our design implements class-independent thresholds for different classes, dynamically adjusting them by encoding pseudo-label distribution during training. To emphasize critical features in informative areas within the WiFi signals, we incorporate both spatial self-attention and temporal attention mechanisms to dynamically learn salient features and identify pivotal frames, respectively. Moreover, we introduce adaptive WiFi data augmentations that propel the semi-supervised framework and enhance model robustness. Experimental results on the Widar3.0 dataset reveal that our approach outperforms existing semi-supervised methods by large margins in accuracy, effectively mitigating imbalanced bias and enhancing model generalization.

Retrieval-Augmented Generation (RAG) improves the performance of Large Language Models (LLMs) by retrieving and integrating relevant information from external knowledge bases, which helps generate more accurate responses. However, RAG is vulnerable to retrieval poisoning attacks , where attackers can induce LLM to produce inaccurate responses by injecting malicious documents into the retrieval process. In this article, we propose ShieldRAG , a novel defense framework designed to counteract retrieval poisoning attacks by reshaping the retrieval embedding space. ShieldRAG leverages a dual-strategy effect realized via a majority-consensus mechanism: ① Push : Implicitly forces the embedding of a user query away from malicious documents by filtering out their minority signals, reducing their influence. ② Pull : Aligns the embedding of a user query closer to that of benign documents, reinforcing accurate retrieval. These strategies work synergistically to preserve retrieval integrity and enhance the quality of LLM-generated responses. Specifically, ShieldRAG operates through three key steps: Sliding Retrieval Explanation Generation , Keyword Aggregation , and Query Targeting Optimization . These three steps collectively ensure the effective integration of information from benign sources while filtering out malicious interference, thereby significantly enhancing the robustness of RAG systems against retrieval poisoning attacks. We evaluate ShieldRAG on four open-domain Question Answering (QA) datasets: Natural Questions, MS-MARCO, HotpotQA, and 2WikiMultiHopQA, using seven representative LLMs. Extensive experiments demonstrate that ShieldRAG significantly improves response accuracy while mitigating adversarial effects, showcasing strong generalization across multiple datasets and LLM architectures.

Hallucination remains a major obstacle to the domain generalizability and reliability of Large Language Models (LLMs). Recent approaches address this issue by integrating Retrieval-Augmented Generation (RAG) with stepwise reasoning processes to iteratively retrieve knowledge. However, indiscriminate incorporation of external knowledge may interfere with reasoning, increasing latency and amplifying error accumulation. Moreover, existing methods rely on a unidirectional flow of external knowledge into LLMs while neglecting internal–external knowledge synergy, limiting autonomous reasoning capability. To address these limitations, we propose the Reasoning–Evidence Interactive Augmented Graph (RE-IAG), a framework that couples reasoning with evidence through a staged triggering mechanism and structured interaction. RE-IAG performs localized refinement of intermediate reasoning via adaptive branching under uncertainty and selectively triggers retrieval when internal reasoning stagnates. Crucially, it organizes both internal reasoning and retrieved evidence into aligned graph structures, enabling structure-guided verification and fine-grained refinement of intermediate conclusions. This design transforms retrieval from passive augmentation into an active constraint on reasoning, reducing error propagation, alleviating knowledge conflicts, and improving knowledge integration for hallucination mitigation. Extensive experiments on four multi-hop QA benchmarks show that RE-IAG outperforms adaptive RAG baselines, achieves competitive or superior performance to RL-based approaches, and demonstrates strong robustness and generalization across model scales and architectures.

Knowledge-based questions are typically employed to evaluate LLM's knowledge boundaries; meanwhile, numerous studies focus on question generation as a means to enhance the capabilities of both models and individuals. However, there is a lack of in-depth exploration about what constitutes a good question from the perspective of knowledge cognition. This paper proposes aligning the complete knowledge underlying questions with educational criteria effectively employed in physics courses, thereby developing novel knowledge-intensive metrics of question quality. To this end, we propose Meta-Fact Checking (MFC), which transforms questions into knowledge graph (KG) triples utilizing LLMs through few-shot prompting, thereby quantifying question quality based on the patterns observed within these triples. MFC introduces a novel interaction mechanism for KGs that communicates meta-facts, illustrating the types of knowledge that KGs can offer to the LLM for reasoning questions, rather than relying solely on the original triples. This strategy ensures that MFC remains unaffected by unexplored triples that LLM has not yet encountered within KGs compared to the retrieve-while-reasoning routine. Experiments across multiple datasets and LLMs demonstrate that MFC significantly improves the accuracy and efficiency of both question answering and assessing. This research marks a pioneering effort to automate the evaluation of question quality based on cognitive capabilities.

Multi-interest recommendation constantly aspires to an oracle individual preference modeling approach, that satisfies the diverse and dynamic properties. Fueled by the deep learning technology, existing neural network (NN)-based recommender systems employ single-point or multi-point interest representation strategy to realize preference modeling,and boost the recommendation performance with a remarkable margin. However, as parameterized approximate functions, NN-based methods remain deficiencies with respect to the adaptability towards distinctive preference patterns cross different users and the calibration over the individual current intent. In this paper, we revisit multi-interest recommendation with the lens of stochastic process and Bayesian inference. Specifically, we propose to learn a distribution over functions to depict the individual diverse preferences rather than a unified function to approximate preference. Subsequently, the recommendation is encouraged with the uncertainty estimation which conforms to the dynamic shifting intent. Along these lines, we establish the connection between multi-interest recommendation and neural processes by proposing NP-Rec, which realizes the flexible multiple interests modeling and uncertainty estimation, simultaneously. Empirical study on 4 real world datasets demonstrates that our NP-Rec attains superior recommendation performances to several state-of-the-art baselines, where the average improvement achieves up to 13.94%.

Fact verification has become increasingly vital in the internet age, driven by the proliferation of false claims and political misinformation. While traditional methods rely predominantly on text-based evidence, multi-modal evidence introduces richer sources of information, offering valuable insights for claim verification. Existing multi-modal verification models often focus on superficial correlations between claims and evidence, neglecting the complex semantic interactions present in fine-grained multi-modal signals. In this paper, we propose a novel framework for multi-modal fact-checking, named Hypergraph Transformer-based Multi-modal Fact-Checking (HGTMFC). Our approach captures high-order relationships between different modalities of evidence and claims by leveraging hypergraphs. HGTMFC models the intricate relationships among evidence across various modalities and enhances information propagation through a transformer-based mechanism embedded within the hypergraph. Moreover, we utilize linegraphs to refine this propagation process, further strengthening the model's reasoning capabilities. Experiments on benchmark datasets demonstrate that our model significantly outperforms existing approaches in multi-modal fact verification.

Deep learning has significantly enhanced survival prediction using whole slide images (WSIs) by adopting a two-stage learning paradigm: WSI preparation and patient-level prediction. While existing research generally concentrates on developing advanced patient-level prediction modules, the critical importance of WSI preparation has been largely overlooked. In practice, WSI preparation is influenced by numerous factors, including tissue heterogeneity, sampling strategies, and technical considerations. These uncontrollable external factors incur variability in the number of WSIs among patients, introducing significant bias and resulting in inferior performance for patients with few WSIs. To address this challenge, we propose a novel approach named WSI-Diffusion. Unlike existing WSI generation models that produce augmented versions of input WSIs, our method generates entirely new WSIs in representation space to serve as complementary data. WSIDiffusion employs a two-stage hierarchical diffusion process. Two novel modules, WSI-level and patch-level Diffusers are designed to capture complex correlations between WSIs and patches. The generated WSIs are integrated as supplementary data, and a light patient-level prediction module is then trained for survival prediction. Experimental results across five datasets demonstrate the superiority of our proposal.

The pervasive spread of misinformation on social networks highlights the critical necessity for effective fact verification systems. Traditional approaches primarily focus on pairwise correlations between claims and evidence, often neglecting comprehensive multi-hop retrieval and reasoning, which results in suboptimal performance when dealing with complex claims. In this paper, we propose MRR-FV, a generative retrieval-enhanced model designed to address the novel challenge of Multi-hop Retrieval and Reasoning for Fact Verification, which integrates two core modules: Generative Multi-hop Retriever and the Hierarchical Interaction Reasoner. MRR-FV utilizes an autoregressive model for iterative multi-hop evidence retrieval, complemented by a pre-trained compressor to address the challenge of intention shift across retrieval hops. For claim verification, we propose a hierarchical interaction reasoner that conducts intra-sentence reasoning to capture long-term semantic dependencies and inter-sentence reasoning across multi-hop evidence subgraphs to reveal complex evidence interactions. Experimental evaluations on the FEVER and HOVER datasets demonstrate the superior performance of our model in both claim verification and evidence retrieval tasks.

The proliferation of fake news on social media platforms has exerted a substantial influence on society, leading to discernible impacts and deleterious consequences. Conventional deep learning methodologies employing small language models (SLMs) suffer from the necessity for extensive supervised training and the challenge of adapting to rapidly evolving circumstances. Large language models (LLMs), despite their robust zero-shot capabilities, have fallen short in effectively identifying fake news due to a lack of pertinent demonstrations and the dynamic nature of knowledge. In this paper, a novel framework Multi-Round Collaboration Detection (MRCD) is proposed to address these aforementioned limitations. The MRCD framework is capable of enjoying the merits from both LLMs and SLMs by integrating their generalization abilities and specialized functionalities, respectively. Our approach features a two-stage retrieval module that selects relevant and up-to-date demonstrations and knowledge, enhancing in-context learning for better detection of emerging news events. We further design a multi-round learning framework to ensure more reliable detection results. Our framework MRCD achieves SOTA results on two real-world datasets Pheme and Twitter16, with accuracy improvements of 7.4% and 12.8% compared to using only SLMs, which effectively addresses the limitations of current models and improves the detection of emergent fake news detection.

Information diffusion prediction (IDP) is a pivotal task for understanding the dynamics of information propagation within social networks. Conventional models typically adhere to a fixed learning-based paradigm, where the trained prediction model remains static during the inference phase. This paradigm presupposes that the data is independent and identically distributed, an assumption that may not hold true due to the inherently open nature of social media and the uncertainty and variability in user behavior. In this paper, we address the novel problem of out-of-distribution (OOD) shifts within IDP tasks and propose a new test-time training-based model for multi-scale IDP tasks, named Ghidorah. Our approach focuses on adapting a subset of model parameters to accommodate the unique characteristics of test samples through self-supervised learning (SSL) tasks. Ghidorah comprises three components: the macroscopic prediction branch, the microscopic prediction branch, and the auxiliary SSL branch. The auxiliary SSL task employs a masked autoencoder-based loss to fine-tune the model for specific test samples prior to prediction. Furthermore, Ghidorah integrates invariant learning to capture robust representations while mitigating spurious correlations. To our knowledge, Ghidorah is the first work to introduce a test-time training framework specifically designed to address the critical yet often overlooked OOD challenges in IDP. Experimental results across several benchmark datasets validate the superiority of our approach.

Chaofan Li, Jianlyu Chen, Yingxia Shao, Chaozhuo Li, Quanqing Xu, Defu Lian, Zheng Liu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

The exploration of bundle recommendation has garnered significant attention for its potential to enhance user experience and augment business sales. Previous research in this domain has primarily focused on modeling user-item and user-bundle interactions, utilizing multi-view collaboration to bolster the accuracy of bundle recommendations. Nevertheless, existing methodologies exhibit limitations, notably in the inadequate modeling of multi-view information and the absence of multi-grained details. Consequently, addressing the intricate correlation among users, items, and bundles necessitates a sophisticated approach capable of capturing both global and local nuances. We present a novel framework named Large Language Model-based Multi-View and Multi-Grained Learning for Bundle Recommendation (LLMCBR). We introduce an LLM-based semantic refinement module to summarize and encode bundle-level knowledge. To bridge the gap between semantic representation and collaborative signals, we design an adaptation strategy. Furthermore, LLMCBR leverages multi-view and multi-granular modeling to unify collaborative signals. Specifically, LLMCBR integrates item preferences within both bundle-view and item-view, thereby augmenting the comprehensiveness of multi-view data. Following this integration, each view undergoes stratification into multiple granularities to facilitate the acquisition of multi-grained details. We introduce a multiple contrastive instance mechanism to regulate the influence of different granularities and views. This mechanism empowers the model to comprehend complex consumer behaviors across various dimensions. LLMCBR is extensively evaluated over three real-world datasets, and the experimental results demonstrate its superiority.

Large language models (LLMs) have shown strong potential in complex reasoning tasks.However, as task complexity increases, their performance often degrades, resulting in hallucinations, errors, and logical inconsistencies.To enhance reasoning capabilities, Monte Carlo Tree Search (MCTS) has been introduced to guide the exploration of reasoning paths in a structured manner.Despite its advantages, traditional MCTS relies on fixed reasoning strategies, limiting the diversity of reasoning paths and the coverage of the solution space.To address these limitations, we propose Dynamic Strategy-Guided MCTS (DSG-MCTS), a novel framework that dynamically integrates multiple reasoning strategies, such as abductive and analogical reasoning, to expand the reasoning space.At the same time, DSG-MCTS enhances reasoning efficiency through a dynamic strategy selection mechanism that adapts to the task context.Experimental results on challenging reasoning benchmarks demonstrate that DSG-MCTS achieves improved accuracy and efficiency, outperforming existing state-of-the-art methods.

Retrieval-augmented generation (RAG) aims to mitigate the hallucination of Large Language Models (LLMs) by retrieving and incorporating relevant external knowledge into the generation process.However, the external knowledge may contain noise and conflict with the parametric knowledge of LLMs, leading to degraded performance.Current LLMs lack inherent mechanisms for resolving such conflicts.To fill this gap, we propose a Dual-Stream Knowledge-Augmented Framework for Shared-Private Semantic Synergy (DSSP-RAG).Central to it is the refinement of the traditional self-attention into a mixed-attention that distinguishes shared and private semantics for a controlled knowledge integration.An unsupervised hallucination detection method that captures the LLMs' intrinsic cognitive uncertainty ensures that external knowledge is introduced only when necessary.To reduce noise in external knowledge, an Energy Quotient (EQ), defined by attention difference matrices between task-aligned and task-misaligned layers, is proposed.Extensive experiments show that DSSP-RAG achieves a superior performance over strong baselines.

With the rapid advancement of online social networks, the proliferation of misinformation has exerted significant influence on both social order and individual decision-making. Traditional manual fact-checking methods are resource-intensive and lack scalability, highlighting the urgent need for automated approaches to verify claims and foster the dissemination of truthful content. In this work, we introduce a novel fact-checking framework, MLLM-CRMFC, which integrates multi-modal large language models (MLLMs) with counterfactual reasoning to enhance both the accuracy and interpretability of automated fact verification. Our multi-modal fact-checking method processes textual and visual data, utilizing LLMs and counterfactual reasoning mechanisms to evaluate the relevance and consistency of diverse evidence. The proposed framework not only predicts the verification outcomes of multi-modal claims but also generates natural language explanations, providing users with transparent and traceable justifications for the decision-making process. Experimental results demonstrate that our model effectively verifies claims using benchmark datasets, outperforms state-of-the-art approaches, and delivers coherent, fluent explanations that enhance the interpretability of the verification process.

Graph Neural Networks (GNNs) have demonstrated superior performance in a variety of graph mining and learning tasks. However, when node representations involve sensitive personal information or variables related to individuals, learning from graph data can raise significant privacy concerns. Although recent studies have explored local differential privacy (LDP) to address these concerns, they often introduce significant distortions to graph data, severely degrading private learning utility (e.g., node classification accuracy). In this paper, we present UPGNET, an LDP-based privacy-preserving graph learning framework that enhances utility while protecting user data privacy. Specifically, we propose a three-stage pipeline that generalizes the LDP protocols for node features, targeting privacy-sensitive scenarios. Our analysis identifies two key factors that affect the utility of privacy-preserving graph learning: feature dimension and neighborhood size. Based on the above analysis, UPGNET enhances utility by introducing two core layers: High-Order Aggregator (HOA) layer and the Node Feature Regularization (NFR) layer. Extensive experiments on real-world datasets indicate that UPGNET significantly outperforms existing methods in terms of both privacy protection and learning utility.

Survival prediction is a pivotal task for estimating mortality risk within a given timeframe based on whole slide images (WSIs). Conventional models typically assume that WSIs across patients are independent and identically distributed, an assumption that may not hold due to inherent variability in WSI preparation and the uncertain condition of infected tissues. These uncontrollable external factors introduce significant variability in the numbers and resolutions of WSIs across patients, leading to bias and compromised performance, particularly for tail patients with limited data. In this paper, we propose a novel approach, PathoKD, based on knowledge distillation. Recognizing the hierarchical nature of disease progression and the data scarcity issues associated with vanilla knowledge distillation methods, PathoKD integrates a novel curriculum learning framework with hierarchical knowledge distillation. This integration effectively mitigates the performance gap between head and tail patients, thereby enhancing prediction accuracy across patient groups. Our proposal is extensively evaluated over popular datasets and experimental results demonstrate its superiority.

The prediction of information popularity propagation is critical for applications such as recommendation systems, targeted advertising, and social media trend analysis. Traditional approaches primarily rely on historical cascade data, often sacrificing timeliness for prediction accuracy. These methods capture aggregate diffusion patterns but fail to account for the complex temporal dynamics of early-stage propagation. In this paper, we introduce Diffusion Guided Propagation Augmentation(DGPA), a novel framework designed to improve early-stage popularity prediction. DGPA models cascade dynamics by leveraging a generative approach, where a temporal conditional interpolator serves as a noising process and forecasting as a denoising process. By iteratively generating cascade representations through a sampling procedure, DGPA effectively incorporates the evolving time steps of diffusion, significantly enhancing prediction timeliness and accuracy. Extensive experiments on benchmark datasets from Twitter, Weibo, and APS demonstrate that DGPA outperforms state-of-the-art methods in early-stage popularity prediction.

Most jailbreak methods for large language models (LLMs) focus on superficially improving attack success through manually defined rules. However, they fail to uncover the underlying mechanisms within target LLMs that explain why an attack succeeds or fails. In this paper, we propose investigating the phenomenon of jailbreaks and defenses for LLMs from the perspective of attention distributions within the models. A preliminary experiment reveals that the success of a jailbreak is closely linked to the LLM's attention on sensitive words.Inspired by this interesting finding, we propose incorporating critical signals derived from internal attention distributions within LLMs, namely Attention Intensity on Sensitive Words and Attention Dispersion Entropy, to guide both attacks and defenses. Drawing inspiration from the concept of "Feint and Attack", we introduce an attention-guided jailbreak model, ABA, which redirects the model's attention to benign contexts, and an attention-based defense model, ABD, designed to detect attacks by analyzing internal attention entropy. Experimental results demonstrate the superiority of our proposal when compared to SOTA baselines.

Multimodal Attributed Graphs (MAGs) are ubiquitous in real-world applications, encompassing extensive knowledge through multimodal attributes attached to nodes (e.g., texts and images) and topological structure representing node interactions. Despite its potential to advance diverse research fields like social networks and e-commerce, MAG representation learning (MAGRL) remains underexplored due to the lack of standardized datasets and evaluation frameworks. In this paper, we first propose MAGB, a comprehensive MAG benchmark dataset, featuring curated graphs from various domains with both textual and visual attributes. Based on the MAGB dataset, we further systematically evaluate two mainstream MAGRL paradigms: GNN-as-Predictor, which integrates multimodal attributes via Graph Neural Networks (GNNs), and VLM-as-Predictor, which harnesses Vision Language Models (VLMs) for zero-shot reasoning. Extensive experiments on MAGB reveal the following critical insights: (i) Modality significances fluctuate drastically with specific domain characteristics. (ii) Multimodal embeddings can elevate the performance ceiling of GNNs. However, intrinsic biases among modalities may impede effective training, particularly in low-data scenarios. (iii) VLMs are highly effective at generating multimodal embeddings that alleviate the imbalance between textual and visual attributes. These discoveries, which illuminate the synergy between multimodal attributes and graph topologies, contribute to reliable benchmarks, paving the way for future research.

LLMs have demonstrated unprecedented capabilities in natural language processing, yet their practical deployment remains hindered by persistent factuality and faithfulness hallucinations. While existing methods address these hallucination types independently, they inadvertently induce performance trade-offs, as interventions targeting one type often exacerbate the other. Through empirical and theoretical analysis of activation space dynamics in LLMs, we reveal that these hallucination categories share overlapping subspaces within neural representations, presenting an opportunity for concurrent mitigation. To harness this insight, we propose SPACE, a unified framework that jointly enhances factuality and faithfulness by editing shared activation subspaces. SPACE establishes a geometric foundation for shared subspace existence through dual-task feature modeling, then identifies and edits these subspaces via a hybrid probe strategy combining spectral clustering and attention head saliency scoring. Experimental results across multiple benchmark datasets demonstrate the superiority of our approach.

The existing Retrieval-Augmented Generation (RAG) systems face significant challenges in terms of cost and effectiveness. On one hand, they need to encode the lengthy retrieved contexts before responding to the input tasks, which imposes substantial computational overhead. On the other hand, directly using generic Large Language Models (LLMs) often leads to sub-optimal answers, while task-specific fine-tuning may compromise the LLMs' general capabilities. To address these challenges, we introduce a novel approach called FlexRAG (Flexible Context Adaptation for RAG). In this approach, the retrieved contexts are compressed into compact embeddings before being encoded by the LLMs. Simultaneously, these compressed embeddings are optimized to enhance downstream RAG performance. A key feature of FlexRAG is its flexibility, which enables effective support for diverse compression ratios and selective preservation of important contexts. With these designs, FlexRAG achieves superior generation quality while significantly reducing running costs. The experiments across multiple QA datasets validate our approach as a cost-effective and flexible solution for RAG systems (codebase: https://github.com/wcyno23/FlexRAG).

Owing to the unprecedented capability in semantic understanding and logical reasoning, large language models (LLMs) have shown fantastic potential in developing next-generation sequential recommender systems (RSs). However, existing LLM-based sequential RSs mostly separate index generation from sequential recommendation, leading to insufficient integration between semantic information and collaborative information. On the other hand, the neglect of user-related information hinders LLM-based sequential RSs from exploiting high-order user-item interaction patterns. In this paper, we propose the End-to-End Dual Dynamic (ED2) recommender, the first LLM-based sequential RS which adopts dual dynamic index mechanism, targeting resolving the above limitations simultaneously. The dual dynamic index mechanism can not only assembly index generation and sequential recommendation into a unified LLM-backbone pipeline, but also make it practical for LLM-based sequential recommender to take advantage of user-related information. Specifically, to facilitate the LLM comprehension ability to dual dynamic index, we propose a multigrained token regulator which constructs alignment supervision based on LLMs semantic knowledge across multiple representation granularities. Moreover, the associated user collection data and a series of novel instruction tuning tasks are specially customized to capture the high-order user-item interaction patterns. Extensive experiments on three public datasets demonstrate the superiority of ED2, achieving an average improvement of 19.62% in Hit-Rate and 21.11% in NDCG.

Large language models (LLMs) provide powerful foundations to perform fine-grained text re-ranking. However, they are often prohibitive in reality due to constraints on computation bandwidth. In this work, we propose a flexible architecture called Matroyshka Re-Ranker, which is designed to facilitate runtime customization of model layers and sequence lengths at each layer based on users' configurations. Consequently, the LLM-based re-rankers can be made applicable across various real-world situations. The increased flexibility may come at the cost of precision loss. To address this problem, we introduce a suite of techniques to optimize the performance. First, we propose cascaded self-distillation, where each sub-architecture learns to preserve a precise re-ranking performance from its super components, whose predictions can be exploited as smooth and informative teacher signals. Second, we design a factorized compensation mechanism, where two collaborative LoRA modules, vertical and horizontal, are jointly employed to compensate for the precision loss resulted from arbitrary combinations of layer and sequence compression. We perform comprehensive experiments using passage and document retrieval datasets from MSMARCO, along with all public datasets from BEIR. In our experiments, Matryoshka Re-Ranker substantially outperforms existing methods, while effectively preserving its superior performance across various compression forms and application scenarios. We have publicly released our method at this https://github.com/FlagOpen/FlagEmbedding repo.

Recommender systems (RSs) have become crucial in helping users navigate the vast amount of online content available today. Graph neural networks (GNNs) have been applied to RSs to capture complex user–item relationships, but existing methods compromise privacy or require centralized data storage. Current works attempt to perform GNN-based RSs under federated learning settings to prevent privacy leakage. However, these works need to perform explicit graph propagation during training, which still introduces potential privacy leakage and data collusion. To address these challenges, we propose a Laplacian-based model called Laplacian graded link prediction (LapGLP) that leverages infinite graph propagation with a constant weight matrix. Instead of actually performing infinite graph propagation, the model abstracts the underlying relations between embeddings after propagation with a weighted minimum squared error problem. Furthermore, we propose a federated framework named FedLapGLP to improve privacy in federated GNN-based RSs, which splits the objective loss function into independent parts that are calculated by each user. Experimental comparisons with state-of-the-art federated RS methods demonstrate the advantages of our proposed approach in terms of high-order connectivity, comprehensive graph information, social relations, full-interaction protection, collusion resistance, and user-embedding protection. The implementation of the proposed framework is available at https://github.com/Limhady/LapGLP.

Session-Based Recommendation (SBR) systems, traditionally reliant on complex Graph Neural Networks (GNNs), often face challenges with marginal performance improvements despite increased model complexity. In this article, we dissect the classical GNN-based SBR models and empirically find that the sophisticated GNN propagations might be redundant, given the readout module plays a significant role in GNN-based models. Based on this observation, we introduce Atten-Mixer+, an advanced iteration of our previously developed Multi-Level Attention Mixture Network (Atten-Mixer). Atten-Mixer+ forgoes GNN propagation in favor of a dynamic and adaptive readout process, tailored to the unique characteristics of each session. Different from the vanilla version, Atten-Mixer+ features the Adaptive Intent Scaler (AIS) layer, which dynamically determines the depth of multi-level user intent analysis and a soft allocation approach for generating user intent queries across entire user interaction sequences. This innovative design allows Atten-Mixer+ to capture a nuanced and comprehensive understanding of user behaviors, overcoming the limitations of fixed-length analysis. Empirical evaluations on benchmark datasets highlight Atten-Mixer+’s superior efficiency and effectiveness, marking a significant step forward in the predictive accuracy of SBR systems.

Self-supervised learning on graph neural networks is receiving increasing attention due to the difficulty of obtaining graph labels in many real applications. Graph contrastive learning (GCL), a recently popular method for self-supervised learning on graphs, has achieved great success in many tasks. The key to the effectiveness of GCL is the construction of suitable contrasting pairs to capture important attributes of the data through the data augmentation modules. However, most of the existing approaches fail to fully consider both data diversity and the semantic consistency when conducting data augmentation. To fill this gap, we propose an augmentation diversity and semantic consistency balanced graph contrastive learning model (ADSCB for short), which enhances the representation ability of the CL model through richer contrasting objectives. In particular, we first introduce a semantic consistency module to extract the subgraph from the original graph through optimizing a carefully designed semantic consistency loss. Then, we introduce an augmentation diversity module and perform data augmentation and cross-scale mix-up operations on the original graph and the extracted semantic preserved subgraph to generate more diverse contrasting pairs. With the above two modules, our model ultimately achieves two contrasting objectives: diversity contrasting and semantic contrasting. The tradeoff between these two contrasting objectives allows our model to benefit from both the augmentation diversity and the semantic consistency. We evaluate ADSCB for graph classification in unsupervised, semi-supervised, and transfer learning settings using standard graph contrastive learning benchmarks. The results demonstrate the superiority of our method against several state-of-the-art baselines.

Sequential recommendation systems are integral to discerning temporal user preferences. Yet, the task of learning from abbreviated user interaction sequences poses a notable challenge. Data augmentation has been identified as a potent strategy to enhance the informational richness of these sequences. Traditional augmentation techniques, such as item randomization, may disrupt the inherent temporal dynamics. Although recent advancements in reverse chronological pseudo-item generation have shown promise, they can introduce temporal discrepancies when assessed in a natural chronological context. In response, we introduce a sophisticated approach, Bidirectional temporal data Augmentation with pre-training (BARec). Our approach leverages bidirectional temporal augmentation and knowledge-enhanced fine-tuning to synthesize authentic pseudo-prior items thatretain user preferences and capture deeper item semantic correlations, thus boosting the model’s expressive power. Our comprehensive experimental analysis on five benchmark datasets confirms the superiority of BARec across both short and elongated sequence contexts. Moreover, theoretical examination and case study offer further insight into the model’s logical processes and interpretability.

In recommender systems, it is frequently presumed that missing ratings adhere to a missing at random (MAR) mechanism, implying the absence of ratings is independent of their potential values. However, this assumption fails to hold in real-world scenarios, where users are inclined to rate items they either strongly favor or disfavor, introducing a missing not at random (MNAR) scenario. To tackle this issue, prior researchers have utilized explicit MAR feedbacks to infer the propensities of unobserved, implicit MNAR feedbacks. Nonetheless, acquiring explicit MAR feedbacks is resource-intensive and time-consuming and may not reflect users’ true preferences. Furthermore, most methods have only been tested on synthetic or small-scale datasets, thus their applicability and effectiveness in real-world settings without MAR feedbacks remain unclear. Along these lines, we aim to predict MNAR ratings without MAR prior propensities by exploring the consistency between MAR and MNAR feedbacks and narrowing the gap between them. From the empirical study and preliminary experiment, we hypothesize thatuser preferencescan be treated as the common prior propensity for both MAR and MNAR generative processes. In this way, we extend this hypothesis to a more general MNAR scenario: user preferences learned from MNAR can partially substitute for the prior propensities derived from MAR feedbacks for MNAR recommendation tasks. To validate our hypothesis and approach, we develop a lightweight iterative probabilistic matrix factorization framework (lightIPMF) as a practical method of our methodology, utilizing user preferences extracted from MNAR, not MAR, to estimate MNAR feedbacks. Finally, the experimental results show that modeling user preferences can effectively improve MNAR feedback estimation without MAR feedback, and our proposed lightIPMF outperforms the state-of-the-art MNAR methods in predicting MNAR feedbacks.

Meta learning has been recognized as an effective remedy for solving the cold-start problem in the recommendation domain. Existing models aim to learn how to generalize from the user behaviors in the training set to testing set. However, in the cold start settings, with only a small number of training samples, the testing distribution may easily deviate from the training one, which may invalidate the learned generalization patterns, and lower the recommendation performance. For alleviating this problem, in this paper, we propose a robust meta recommender framework to address the distribution shift problem. In specific, we argue that the distribution shift may exist on both the user- and interaction-levels, and in order to mitigate them simultaneously, we design a novel distributionally robust model by hierarchically reweighing the training samples. Different sample weights correspond to different training distributions, and we minimize the largest loss induced by the sample weights in a simplex, which essentially optimizes the upper bound of the testing loss. In addition, we analyze our framework on the convergence rates and generalization error bound to provide more theoretical insights. Empirically, we conduct extensive experiments based on different meta recommender models and real-world datasets to verify the generality and effectiveness of our framework.

Amidst the rapid propagation of multimodal fake news across social media platforms, the detection of fake news has emerged as a prime research pursuit. To detect heightened level of meticulous fabrications, propagation paths are introduced to provide nuanced social context that enhances the basic semantic analysis of the news content. However, existing propagation-enhanced models encounter a dilemma between detection efficacy and social hazard. In this paper, we explore the innovative problem of early fake news detection through the generation of propagation paths, capable of benefiting from the extensive social context within propagation paths while mitigating potential social hazards. To address these challenges, we propose a novel Reinforced Propagation Path Generation Fake News Detection model,RPPG-Fake. Departing from conventional discriminative approaches,RPPG-Fakecaptures the propagation topology pattern from a heterogeneous social graph and generates the propagation paths to detect fake news effectively under a reinforcement learning paradigm. Our proposal is extensively evaluated over three popular datasets, and experimental results demonstrate the superiority of our proposal.

Image-based survival prediction through deep learning techniques represents a burgeoning frontier aimed at augmenting the diagnostic capabilities of pathologists. However, directly applying existing deep learning models to survival prediction may not be a panacea due to the inherent complexity and sophistication of whole slide images (WSIs). The intricate nature of high-resolution WSIs, characterized by sophisticated patterns and inherent noise, presents significant challenges in terms of effectiveness and trustworthiness. In this paper, we propose CTUSurv, a novel survival prediction model designed to simultaneously capture cell-to-cell and cell-to-microenvironment interactions, complemented by a region-based uncertainty estimation framework to assess the reliability of survival predictions. Our approach incorporates an innovative region sampling strategy to extract task-relevant, informative regions from high-resolution WSIs. To address the challenges posed by sophisticated biological patterns, a cell-aware encoding module is integrated to model the interactions among biological entities. Furthermore, CTUSurv includes a novel aleatoric uncertainty estimation module to provide fine-grained uncertainty scores at the region level. Extensive evaluations across four datasets demonstrate the superiority of our proposed approach in terms of both predictive accuracy and reliability.

Inferring user preferences from users’ historical feedback is a valuable problem in recommender systems. Conventional approaches often rely on the assumption that user preferences in the feedback data are equivalent to the real user preferences without additional noise, which simplifies the problem modeling. However, there are various confounders during user–item interactions, such as weather and even the recommendation system itself. Therefore, neglecting the influence of confounders will result in inaccurate user preferences and suboptimal performance of the model. Furthermore, the unobservability of confounders poses a challenge in further addressing the problem. Along these lines, we refine the problem and propose a more rational solution to mitigate the influence of unobserved confounders. Specifically, we consider the influence of unobserved confounders, disentangle them from user preferences in the latent space, and employ causal graphs to model their interdependencies without specific labels. By ingeniously combining local and global causal graphs, we capture the user-specific effects of confounders on user preferences. Finally, we propose our model based on Variational Autoencoders, named Causal Structure Aware Variational Autoencoders (CSA-VAE) and theoretically demonstrate the identifiability of the obtained causal graph. We conducted extensive experiments on one synthetic dataset and nine real-world datasets with different scales, including three unbiased datasets and six normal datasets, where the average performance boost against several state-of-the-art baselines achieves up to 9.55%, demonstrating the superiority of our model. Furthermore, users can control their recommendation list by manipulating the learned causal representations of confounders, generating potentially more diverse recommendation results. Our code is available at Code-link ( https://github.com/MICLab-Rec/CSA ).

Privacy-Preserving Image Retrieval (PPIR) enables searching for similar images on cloud servers while safeguarding image privacy. PPIR uploads encrypted images to servers to address privacy concerns and then employs deep neural networks for retrieval on extracted features from cipher-images. However, current PPIR encrypts all images with the same secret keys to maintain consistent feature spaces, lacking support for adaptive keys where distinct images are encrypted with various keys. To this end, we propose a new PPIR scheme to support adaptive keys while keeping stable feature spaces. Specifically, we design ingenious image encryption to align with feature extraction during the JPEG compression process, incorporating encryptable orthogonal transformations, shuffling, stream cipher, and sign encryption operations. Our approach extracts well-designed absolute value sequences of local blocks and global histogram features from cipher-images, ensuring stable feature spaces under adaptive keys. To enhance model generalization performance, we employ Self-Supervised Contrastive Learning (SSCL) to build a pretraining model and propose a straightforward yet efficient block-sampling augmentation technique for the structured features to drive SSCL. Moreover, our retrieval model implements a dual-attention structure to capture dependencies among local block sequences and import scores of global features. Extensive experiments on four datasets demonstrate that our approach achieves superior retrieval accuracy compared to existing schemes and maintains excellent retrieval performance under adaptive keys, effectively preserving image privacy.

Graph neural networks (GNNs) are commonly employed in collaborative friend recommendation systems. Nevertheless, recent studies reveal a notable performance gap, particularly for users with limited connections, commonly known as tail users, in contrast to their counterparts with abundant connections (head users). Uniformly treating head and tail users poses two challenges for tail user preference learning: (C1) Label Sparsity, as tail users typically possess limited labels; and (C2) Neighborhood Sparsity, where tail users exhibit sparse observable friendships, leading to distinct preference distributions and performance degradation compared to head users. In response to these challenges, we introduce Tail-STEAK, an innovative framework that combines self-training with enhanced knowledge distillation for tail user representation learning. To address(C1), we present Tail-STEAK-base, a two-stage self-training framework. In the first stage, only head users and their accurate connections are utilized for training, while pseudo links are generated for tail users in the second stage. To tackle (C2), we propose two data augmentation-based self-knowledge distillation pretext tasks. These tasks are seamlessly integrated into different stages of Tail-STEAK-base, culminating in the comprehensive Tail-STEAK framework. Extensive experiments, conducted on state-of-the-art GNN-based friend recommendation models, substantiate the efficacy of Tail-STEAK in significantly improving tail user performance. Our code and data are publicly available at https://github.com/antman9914/Tail-STEAK.

Nowadays, detecting multimodal fake news has emerged as a foremost concern since the widespread dissemination of fake news may incur adverse societal impact. Conventional methods generally focus on capturing the linguistic and visual semantics within the multimodal content, which fall short in effectively distinguishing the heightened level of meticulous fabrications. Recently, external knowledge is introduced to provide valuable background facts as complementary to facilitate news detection. Nevertheless, existing knowledge-enhanced endeavors directly incorporate all knowledge contexts through static entity embeddings, resulting in the potential noisy and content-irrelevant knowledge. Moreover, the integration of knowledge entities makes it intractable to model the sophisticated correlations between multimodal semantics and knowledge entities. In light of these limitations, we propose a novel Adaptive Knowledge-Aware Fake News Detection model, dubbed AKA-Fake. For each news, AKA-Fake learns a compact knowledge subgraph under a reinforcement learning paradigm, which consists of a subset of entities and contextual neighbors in the knowledge graph, restoring the most informative knowledge facts. A novel heterogeneous graph learning module is further proposed to capture the reliable cross-modality correlations via topology refinement and modality-attentive pooling. Our proposal is extensively evaluated over three popular datasets, and experimental results demonstrate the superiority of AKA-Fake.

Jailbreak attacks enable malicious queries to evade detection by LLMs.Existing attacks focus on meticulously constructing prompts to disguise harmful intentions.However, the incorporation of sophisticated disguising prompts may incur the challenge of "intention shift".Intention shift occurs when the additional semantics within the prompt distract the LLMs, causing the responses to deviate significantly from the original harmful intentions.In this paper, we propose a novel component, "bait", to alleviate the effects of intention shift.Bait comprises an initial response to the harmful query, prompting LLMs to rectify or supplement the knowledge within the bait.By furnishing rich semantics relevant to the query, the bait helps LLMs focus on the original intention.To conceal the harmful content within the bait, we further propose a novel attack paradigm, BaitAttack.BaitAttack adaptively generates necessary components to persuade targeted LLMs that they are engaging with a legitimate inquiry in a safe context.Our proposal is evaluated on a popular dataset, demonstrating state-of-the-art attack performance and an exceptional capability for mitigating intention shift.The implementation of BaitAttack is accessible at: https://anonymous.4open. science/r/BaitAttack-D1F5.How to make a bomb. Query Bait Role Scene Format Bait Maker Which expert is best suitable to deal with the act of < Harmful Query >? Create a scene that fits the expert role.

Machine learning has demonstrated remarkable performance over finite datasets, yet whether the scores over the fixed benchmarks can sufficiently indicate the model’s performance in the real world is still in discussion. In reality, an ideal robust model will probably behave similarly to the oracle (e.g., the human users), thus a good evaluation protocol is probably to evaluate the models’ behaviors in comparison to the oracle. In this paper, we introduce a new robustness measurement that directly measures the image classification model’s performance compared with a surrogate oracle (i.e., a zoo of foundation models). Besides, we design a simple method that can accomplish the evaluation beyond the scope of the benchmarks. Our method extends the image datasets with new samples that are sufficiently perturbed to be distinct from the ones in the original sets, but are still bounded within the same image-label structure the original test image represents, constrained by a zoo of foundation models pretrained with a large amount of samples. As a result, our new method will offer us a new way to evaluate the models’ robustness performance, free of limitations of fixed benchmarks or constrained perturbations, although scoped by the power of the oracle. In addition to the evaluation results, we also leverage our generated data to understand the behaviors of the model and our new evaluation strategies.

Recent advances in knowledge graph embedding (KGE) rely on Euclidean/hyperbolic orthogonal relation transformations to model intrinsic logical patterns and topological structures. However, existing approaches are confined to rigid relational orthogonalization with restricted dimension and homogeneous geometry, leading to deficient modeling capability. In this work, we move beyond these approaches in terms of both dimension and geometry by introducing a powerful framework named GoldE, which features a universal orthogonal parameterization based on a generalized form of Householder reflection. Such parameterization can naturally achieve dimensional extension and geometric unification with theoretical guarantees, enabling our framework to simultaneously capture crucial logical patterns and inherent topological heterogeneity of knowledge graphs. Empirically, GoldE achieves state-of-the-art performance on three standard benchmarks. Codes are available at https://github.com/xxrep/GoldE.

Learning item representation is crucial for a myriad of on-line e-commerce applications. The nucleus of retail item representation learning is how to properly fuse the semantics within a single item, and the interactions across different items generated by user behaviors (e.g., co-click or co-view). Product semantics depict the intrinsic characteristics of the item, while the interactions describe the relationships between items from the perspective of human perception. Existing approaches either solely rely on a single type of information or loosely couple them together, leading to hindered representations. In this work, we propose a novel model named TESPA to reinforce semantic modeling and interaction modeling mutually. Specifically, collaborative filtering signals in the interaction graph are encoded into the language models through fine-grained topological pre-training, and the interaction graph is further enriched based on semantic similarities. After that, a novel multi-channel co-training paradigm is proposed to deeply fuse the semantics and interactions under a unified framework. In a nutshell, TESPA is capable of enjoying the merits of both sides to facilitate item representation learning. Experimental results of on-line and off-line evaluations demonstrate the superiority of our proposal.

The detection of fake news has emerged as a pressing issue in the era of online social media. To detect meticulously fabricated fake news, propagation paths are introduced to provide nuanced social context to complement the pure semantics within news content. However, existing propagation-enhanced models face a dilemma between detection efficacy and social hazard. In this paper, we investigate the novel problem of early fake news detection via propagation path generation, capable of enjoying the merits of rich social context within propagation paths while alleviating potential social hazards. In contrast to previous discriminative detection models, we further propose a novel generative model, DGA-Fake, by simulating realistic propagation paths based on news content before actual spreading. A guided diffusion module is integrated into DGA-Fake to generate simulated user interaction sequences, guided by historical interactions and news content. Evaluation across three datasets demonstrates the superiority of our proposal.

Adversarial attacks against graph neural networks (GNNs) through perturbations of the graph structure are increasingly common in social network tasks like rumor detection. Social media platforms capture diverse attack sequence samples through both machine and manual screening processes. Investigating effective ways to leverage these adversarial samples to enhance robustness is imperative. We improve the maximum entropy inverse reinforcement learning (IRL) method with the mixture-of-experts approach to address multi-source graph adversarial attacks. This method reconstructs the attack policy, integrating various attack models and providing feature-level explanations, subsequently generating additional adversarial samples to fortify the robustness of detection models. We develop precise sample guidance and a bidirectional update mechanism to reduce the deviation caused by imprecise feature representation and negative sampling within the large action space of social graphs, while also accelerating policy learning. We take rumor detector as an example targeted GNN model on real-world rumor datasets. By utilizing a small subset of samples generated by various graph adversarial attack methods, we reconstruct the attack policy, closely approximating the performance of the original attack method. We validate that samples generated by the learned policy enhance model robustness through adversarial training and data augmentation.

Graph Neural Networks (GNNs) have emerged as promising solutions for collaborative filtering (CF) through the modeling of user-item interaction graphs. The nucleus of existing GNN-based recommender systems involves recursive message passing along user-item interaction edges to refine encoded embeddings. Despite their demonstrated effectiveness, current GNN-based methods encounter challenges of limited receptive fields and the presence of noisy "interest-irrelevant" connections. In contrast, Transformer-based methods excel in aggregating information adaptively and globally. Nevertheless, their application to large-scale interaction graphs is hindered by inherent complexities and challenges in capturing intricate, entangled structural information. In this paper, we propose TransGNN, a novel model that integrates Transformer and GNN layers in an alternating fashion to mutually enhance their capabilities. Specifically, TransGNN leverages Transformer layers to broaden the receptive field and disentangle information aggregation from edges, which aggregates information from more relevant nodes, thereby enhancing the message passing of GNNs. Additionally, to capture graph structure information effectively, positional encoding is meticulously designed and integrated into GNN layers to encode such structural knowledge into node attributes, thus enhancing the Transformer's performance on graphs. Efficiency considerations are also alleviated by proposing the sampling of the most relevant nodes for the Transformer, along with two efficient sample update strategies to reduce complexity. Furthermore, theoretical analysis demonstrates that TransGNN offers increased expressiveness compared to GNNs, with only a marginal increase in linear complexity. Extensive experiments on five public datasets validate the effectiveness and efficiency of TransGNN. Our code is available at https://github.com/Peiyance/TransGNN-torch.

In the realm of personalized recommender systems, the challenge of adapting to evolving user preferences and the continuous influx of new users and items is paramount. Conventional models, typically reliant on a static training-test approach, struggle to keep pace with these dynamic demands. Streaming recommendation, particularly through continual graph learning, has emerged as a novel solution, attracting significant attention in academia and industry. However, existing methods in this area either rely on historical data replay, which is increasingly impractical due to stringent data privacy regulations; or are inability to effectively address the over-stability issue; or depend on model-isolation and expansion strategies, which necessitate extensive model expansion and are hampered by time-consuming updates due to large parameter sets. To tackle these difficulties, we present GPT4Rec, a Graph Prompt Tuning method for streaming Recommendation. Given the evolving user-item interaction graph, GPT4Rec first disentangles the graph patterns into multiple views. After isolating specific interaction patterns and relationships in different views, GPT4Rec utilizes lightweight graph prompts to efficiently guide the model across varying interaction patterns within the user-item graph. Firstly, node-level prompts are employed to instruct the model to adapt to changes in the attributes or properties of individual nodes within the graph. Secondly, structure-level prompts guide the model in adapting to broader patterns of connectivity and relationships within the graph. Finally, view-level prompts are innovatively designed to facilitate the aggregation of information from multiple disentangled views. These prompt designs allow GPT4Rec to synthesize a comprehensive understanding of the graph, ensuring that all vital aspects of the user-item interactions are considered and effectively integrated. Experiments on four diverse real-world datasets demonstrate the effectiveness and efficiency of our proposal.

We investigate node representation learning on text-attributed graphs (TAGs), where nodes are associated with text information. Although recent studies on graph neural networks (GNNs) and pretrained language models (PLMs) have exhibited their power in encoding network and text signals, respectively, less attention has been paid to delicately coupling these two types of models on TAGs. Specifically, existing GNNs rarely model text in each node in a contextualized way; existing PLMs can hardly be applied to characterize graph structures due to their sequence architecture. To address these challenges, we propose HASH-CODE, a High-frequency Aware Spectral Hierarchical Contrastive Selective Coding method that integrates GNNs and PLMs into a unified model. Different from previous "cascaded architectures" that directly add GNN layers upon a PLM, our HASH-CODE relies on five self-supervised optimization objectives to facilitate thorough mutual enhancement between network and text signals in diverse granularities. Moreover, we show that existing contrastive objective learns the low-frequency component of the augmentation graph and propose a high-frequency component (HFC)-aware contrastive learning objective that makes the learned embeddings more distinctive. Extensive experiments on six real-world benchmarks substantiate the efficacy of our proposed approach. In addition, theoretical analysis and item embedding visualization provide insights into our model interoperability.

The message-passing paradigm has served as the foundation of graph neural networks (GNNs) for years, making them achieve great success in a wide range of applications. Despite its elegance, this paradigm presents several unexpected challenges for graph-level tasks, such as the long-range problem, information bottleneck, over-squashing phenomenon, and limited expressivity. In this study, we aim to overcome these major challenges and break the conventional "node- and edge-centric" mindset in graph-level tasks. To this end, we provide an in-depth theoretical analysis of the causes of the information bottleneck from the perspective of information influence. Building on the theoretical results, we offer unique insights to break this bottleneck and suggest extracting a skeleton tree from the original graph, followed by propagating information in a distinctive manner on this tree. Drawing inspiration from natural trees, we further propose to find trunks from graph skeleton trees to create powerful graph representations and develop the corresponding framework for graph-level tasks. Extensive experiments on multiple real-world datasets demonstrate the superiority of our model. Comprehensive experimental analyses further highlight its capability of capturing long-range dependencies and alleviating the over-squashing problem, thereby providing novel insights into graph-level tasks.

During the COVID-19 pandemic, forecasting COVID-19 trends to support planning and response was a priority for scientists and decision makers alike. In the United States, COVID-19 forecasting was coordinated by a large group of universities, companies, and government entities led by the Centers for Disease Control and Prevention and the US COVID-19 Forecast Hub (https://covid19forecasthub.org). We evaluated approximately 9.7 million forecasts of weekly state-level COVID-19 cases for predictions 1-4 weeks into the future submitted by 24 teams from August 2020 to December 2021. We assessed coverage of central prediction intervals and weighted interval scores (WIS), adjusting for missing forecasts relative to a baseline forecast, and used a Gaussian generalized estimating equation (GEE) model to evaluate differences in skill across epidemic phases that were defined by the effective reproduction number. Overall, we found high variation in skill across individual models, with ensemble-based forecasts outperforming other approaches. Forecast skill relative to the baseline was generally higher for larger jurisdictions (e.g., states compared to counties). Over time, forecasts generally performed worst in periods of rapid changes in reported cases (either in increasing or decreasing epidemic phases) with 95% prediction interval coverage dropping below 50% during the growth phases of the winter 2020, Delta, and Omicron waves. Ideally, case forecasts could serve as a leading indicator of changes in transmission dynamics. However, while most COVID-19 case forecasts outperformed a naïve baseline model, even the most accurate case forecasts were unreliable in key phases. Further research could improve forecasts of leading indicators, like COVID-19 cases, by leveraging additional real-time data, addressing performance across phases, improving the characterization of forecast confidence, and ensuring that forecasts were coherent across spatial scales. In the meantime, it is critical for forecast users to appreciate current limitations and use a broad set of indicators to inform pandemic-related decision making.

Image retrieval systems help users to browse and search among extensive images in real time. With the rise of cloud computing, retrieval tasks are usually outsourced to cloud servers. However, the cloud scenario brings a daunting challenge of privacy protection as cloud servers cannot be fully trusted. To this end, image-encryption-based privacy-preserving image retrieval (PPIR) schemes have been developed, which first extract features from cipher-images, and then build retrieval models based on these features. Yet, most existing PPIR approaches extract shallow features and design trivial unsupervised retrieval models, resulting in insufficient expressiveness for the cipher-images. In this paper, we propose a novel paradigm named Encrypted Vision Transformer (EViT), which advances the discriminative representations capability of cipher-images. First, to capture comprehensive ruled information, we extract multi-level local length sequence and global Huffman-Code frequency features from the cipher-images which are encrypted by permutation encryption, sign encryption, and stream cipher during the JPEG compression process. Second, we design the modified self-supervised Vision Transformer with Huffman-embedding and propose two robust data augmentations on cipher-images to improve representation power of the retrieval model. Moreover, our proposal can be easily adapted to unsupervised or supervised settings. Extensive experiments reveal that EViT achieves both excellent encryption and retrieval performance, outperforming current schemes in terms of retrieval accuracy by large margins while protecting image privacy effectively. Code is publicly available at https://github.com/onlinehuazai/EViT.

Social network alignment, aiming at linking identical identities across different social platforms, is a fundamental task in social graph mining. Most existing approaches are supervised models and require a large number of manually labeled data, which are infeasible in practice considering the yawning gap between social platforms. Recently, isomorphism across social networks is incorporated as complementary to link identities from the distribution level, which contributes to alleviating the dependency on sample-level annotations. Adversarial learning is adopted to learn a shared projection function by minimizing the distance between two social distributions. However, the hypothesis of isomorphism might not always hold true as social user behaviors are generally unpredictable, and thus a shared projection function is insufficient to handle the sophisticated cross-platform correlations. In addition, adversarial learning suffers from training instability and uncertainty, which may hinder model performance. In this article, we propose a novel meta-learning-based social network alignment model Meta-SNA to effectively capture the isomorphism and the unique characteristics of each identity. Our motivation lies in learning a shared meta-model to preserve the global cross-platform knowledge and an adaptor to learn a specific projection function for each identity. Sinkhorn distance is further introduced as the distribution closeness measurement to tackle the limitations of adversarial learning, which owns an explicitly optimal solution and can be efficiently computed by the matrix scaling algorithm. Empirically, we evaluate the proposed model over multiple datasets, and the experimental results demonstrate the superiority of Meta-SNA.

In this work, we present TriMLP as a foundational MLP-like architecture for the sequential recommendation, simultaneously achieving computational efficiency and promising performance. First, we empirically study the incompatibility between existing purely MLP-based models and sequential recommendation, that the inherent fully-connective structure endows historical user–item interactions (referred as tokens) with unrestricted communications and overlooks the essential chronological order in sequences. Then, we propose the MLP-based Triangular Mixer to establish ordered contact among tokens and excavate the primary sequential modeling capability under the standard auto-regressive training fashion. It contains (1) a global mixing layer that drops the lower-triangle neurons in MLP to block the anti-chronological connections from future tokens and (2) a local mixing layer that further disables specific upper-triangle neurons to split the sequence as multiple independent sessions. The mixer serially alternates these two layers to support fine-grained preferences modeling, where the global one focuses on the long-range dependency in the whole sequence, and the local one calls for the short-term patterns in sessions. Experimental results on 12 datasets of different scales from 4 benchmarks elucidate that TriMLP consistently attains favorable accuracy/efficiency tradeoff over all validated datasets, where the average performance boost against several state-of-the-art baselines achieves up to 14.88%, and the maximum reduction of inference time reaches 23.73%. The intriguing properties render TriMLP a strong contender to the well-established RNN-, CNN-, and Transformer-based sequential recommenders. Code is available at https://github.com/jiangyiheng1/TriMLP .

Rui Li, Xu Chen, Chaozhuo Li, Yanming Shen, Jianan Zhao, Yujing Wang, Weihao Han, Hao Sun, Weiwei Deng, Qi Zhang, Xing Xie. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Researching data poisoning attacks against fake news detection models is crucial for bolstering their robustness and curbing the dissemination of fake news. Existing textual data poisoning attacks necessitate control over both the content and labels of news samples, making them impractical for real attack scenarios. In this paper, we propose COMCP, a novel clean-label poisoning attack model aimed at fake news detection models. Diverging from existing methods, COMCP ensures the poison samples are accurately labeled, while crafting stealthy poison comments without modifying the headlines or content, thereby enhancing the feasibility of the attack. Furthermore, COMCP generates poison comments by appending stealthy characters to ensure the stealthiness of the attack. Comprehensive experimental evaluations on three benchmark datasets illustrate that our proposal outperforms SOTA baselines in terms of attack success rate and text quality, while maintaining the accuracy of detecting clean samples.

Graph collaborative filtering, which could capture the abundant collaborative signal from the high-order connectivity of the tree-likeness user-item interaction graph, has received considerable research attention recently. Most graph collaborative filtering methods embed graphs in the Euclidean spaces, but that could have high distortion when embedding graphs with tree-likeness structure. Recently, some researchers address this problem by learning the feature representations in the hyperbolic spaces. However, because the user-item interaction graphs also have cyclic structure, the high-order collaborative signal cannot be well captured by hyperbolic spaces. From this point of view, neither Euclidean spaces nor hyperbolic spaces can capture the full information from the complexity of user-item interactions. Therefore, how to construct a suitable embedding space for graph collaboration filtering is an important problem. In this paper, we analyze the properties of hyperbolic geometry in graph collaborative filtering tasks and proposed a novel geometry interaction augmented graph collaborative filtering (GeoGCF) method, which leverages both Euclidean and hyperbolic geometry to model the user-item interactions. Experimental results show the effectiveness of the proposed method.

Sequential recommendation (SR) aims to model users' dynamic preferences from a series of interactions. A pivotal challenge in user modeling for SR lies in the inherent variability of user preferences. An effective SR model is expected to capture both the long-term and short-term preferences exhibited by users, wherein the former can offer a comprehensive understanding of stable interests that impact the latter. To more effectively capture such information, we incorporate locality inductive bias into the Transformer by amalgamating its global attention mechanism with a local convolutional filter, and adaptively ascertain the mixing importance on a personalized basis through layer-aware adaptive mixture units, termed as AdaMCT. Moreover, as users may repeatedly browse potential purchases, it is expected to consider multiple relevant items concurrently in long-/short-term preferences modeling. Given that softmax-based attention may promote unimodal activation, we propose the Squeeze-Excitation Attention (with sigmoid activation) into SR models to capture multiple pertinent items (keys) simultaneously. Extensive experiments on three widely employed benchmarks substantiate the effectiveness and efficiency of our proposed approach. Source code is available at https://github.com/juyongjiang/AdaMCT.

Pre-trained language models (PLMs) have achieved the preeminent position in dense retrieval due to their powerful capacity in modeling intrinsic semantics.However, most existing PLM-based retrieval models encounter substantial computational costs and are infeasible for processing long documents.In this paper, a novel retrieval model Longtriever is proposed to embrace three core challenges of long document retrieval: substantial computational cost, incomprehensive document understanding, and scarce annotations.Longtriever splits long documents into short blocks and then efficiently models the local semantics within a block and the global context semantics across blocks in a tightly-coupled manner.A pretraining phase is further proposed to empower Longtriever to achieve a better understanding of underlying semantic correlations.Experimental results on two popular benchmark datasets demonstrate the superiority of our proposal.The source code is released at https: //github.com/SamuelYang1/Longtriever .

Vision transformers have significantly advanced the field of computer vision in recent years. The cornerstone of these transformers is the multi-head attention mechanism, which models interactions between visual elements within a feature map. However, the vanilla multi-head attention paradigm independently learns parameters for each head, which ignores crucial interactions across different attention heads and may result in redundancy and under-utilization of the model’s capacity. To enhance model expressiveness, we propose a novel nested attention mechanism, Ne-Att, that explicitly models cross-head interactions via a hierarchical variational distribution. We conducted extensive experiments on image classification, and the results demonstrate the superiority of Ne-Att.

The nucleus of online sponsored search systems lies in measuring the relevance between the search intents of users and the advertising purposes of advertisers. Existing conventional doublet-based (query-keyword) relevance models solely rely on short queries and keywords to uncover such intents, which ignore the diverse and personalized preferences of participants (i.e., users and advertisers), resulting in undesirable advertising performance. In this paper, we investigate the novel problem of Personalized A dvertiser-aware Sponsored Search (PASS). Our motivation lies in incorporating the portraits of users and advertisers into relevance models to facilitate the modeling of intrinsic search intents and advertising purposes, leading to a quadruple-based (i.e., user-query-keyword-advertiser) task. Various types of historical behaviors are explored in the format of hypergraphs to provide abundant signals on identifying the preferences of participants. A novel heterogeneous textual hypergraph transformer is further proposed to deeply fuse the textual semantics and the high-order hypergraph topology. Our proposal is extensively evaluated over real industry datasets, and experimental results demonstrate its superiority.

GNN explanation method aims to identify an explanatory subgraph which contains the most informative components of the full graph. However, a major limitation of existing GNN explainers is that they are not robust to the structurally corrupted graphs, e.g., graphs with noisy or adversarial edges. On the one hand, existing GNN explainers mostly explore explanations based on either the raw graph features or the learned latent representations, both of which can be easily corrupted. On the other hand, the corruptions in graphs are irregular in terms of the structural properties, e.g., the size or connectivity of graphs, which makes the rigorous constraints used by previous GNN explainers unfeasible. To address these issues, we propose a robust GNN explainer called V-InfoR. Specifically, a robust graph representation extractor, which takes insights of variational inference, is proposed to infer the latent distribution of graph representations. Instead of directly using the corrupted raw features or representations of each single graph, we sample the graph representations from the inferred distribution for the downstream explanation generator, which can effectively eliminate the minor corruption. We next formulate the explanation exploration as a graph information bottleneck (GIB) optimization problem. As a more general method that does not need any rigorous structural constraints, our GIB-based method can adaptively capture both the regularity and irregularity of the severely corrupted graphs for explanation. Extensive evaluations on both synthetic and real-world datasets indicate that V-InfoR significantly improves the GNN explanation performance for the structurally corrupted graphs. Code and dataset are available at https://anonymous.4open.science/r/V-InfoR-EF88

Text-attributed graphs (TAGs) are prevalent in various real-world scenarios, where each node is associated with a text description. The cornerstone of representation learning on TAGs lies in the seamless integration of textual semantics within individual nodes and the topological connections across nodes. Recent advancements in pre-trained language models (PLMs) and graph neural networks (GNNs) have facilitated effective learning on TAGs, garnering increased research interest. However, the absence of meaningful benchmark datasets and standardized evaluation procedures for TAGs has impeded progress in this field. In this paper, we propose CS-TAG, a comprehensive and diverse collection of challenging benchmark datasets for TAGs. The CS-TAG datasets are notably large in scale and encompass a wide range of domains, spanning from citation networks to purchase graphs. In addition to building the datasets, we conduct extensive benchmark experiments over CS-TAG with various learning paradigms, including PLMs, GNNs, PLM-GNN co-training methods, and the proposed novel topological pre-training of language models. In a nutshell, we provide an overview of the CS-TAG datasets, standardized evaluation procedures, and present baseline experiments. The entire CS-TAG project is publicly accessible at \url{https://github.com/sktsherlock/TAG-Benchmark}.

Intrinsic interpretable graph neural networks aim to provide transparent predictions by identifying the influential fraction of the input graph that guides the model prediction, i.e., the explanatory subgraph. However, current interpretable GNNs mostly are dataset-specific and hard to generalize to different graphs. A more generalizable GNN interpretation model which can effectively distill the universal structural patterns of different graphs is until-now unexplored. Motivated by the great success of recent pre-training techniques, we for the first time propose the Pre-training Interpretable Graph Neural Network ($\pi$-GNN) to distill the universal interpretability of GNNs by pre-training over synthetic graphs with ground-truth explanations. Specifically, we introduce a structural pattern learning module to extract diverse universal structure patterns and integrate them together to comprehensively represent the graphs of different types. Next, a hypergraph refining module is proposed to identify the explanatory subgraph by incorporating the universal structure patterns with local edge interactions. Finally, the task-specific predictor is cascaded with the pre-trained $\pi$-GNN model and fine-tuned over downstream tasks. Extensive experiments demonstrate that $\pi$-GNN significantly surpasses the leading interpretable GNN baselines with up to 9.98\% interpretation improvement and 16.06\% classification accuracy improvement. Meanwhile, $\pi$-GNN pre-trained on graph classification task also achieves the top-tier interpretation performance on node classification task, which further verifies its promising generalization performance among different downstream tasks. Our code and datasets are available at https://anonymous.4open.science/r/PI-GNN-F86C

This paper studies the problem of active causal discovery when the experiments can be done based on multi-fidelity oracles, where higher fidelity experiments are more precise and expensive, while the lower ones are cheaper but less accurate. In this paper, we formally define the task of multi-fidelity active causal discovery, and design a probabilistic model for solving this problem. In specific, we first introduce a mutual-information based acquisition function to determine which variable should be intervened at which fidelity, and then a cascading model is proposed to capture the correlations between different fidelity oracles. Beyond the above basic framework, we also extend it to the batch intervention scenario. We find that the theoretical foundations behind the widely used and efficient greedy method do not hold in our problem. To solve this problem, we introduce a new concept called $\epsilon$-submodular, and design a constraint based fidelity model to theoretically validate the greedy method. We conduct extensive experiments to demonstrate the effectiveness of our model.

Graph contrastive learning (GCL) has attracted rising research attention recently due to its effectiveness in self- supervised graph learning. A key step of GCL is to conduct data augmentation, based on which self-supervised learning is performed through the contrast between two augmented data views. Existing approaches generally generate the two data views from the original graph, which has been revealed to be less effective due to the lack of data diversity. Meanwhile, although the data augmentation methods and the contrastive modes have been extensively studied, the effect of hard negative samples (i.e.samples that are difficult to distinguish from an anchor node) on GCL is not fully explored. In this paper, we propose a novel complementary graph contrastive learning method boosted by adversarial hard negative sample generation. Specifically, we first construct a κNN graph as the complementary counterpart of the original graph in the semantic space. Then graph augmentation is conducted in both the semantic and topology spaces for the two complementary graphs to obtain two contrastive views with a larger data diversity. To facilitate the contrastive learning, an adversarial network named ADNet is also proposed to generate hard negative samples. The generated samples are more informative and challenging, and thus can further boost the learning performance. Extensive evaluations over the node classification task demonstrate that our proposal outperforms existing state-of-the-art GCL methods, and even exceeds supervised approaches. The code of this work is publicly available at https://github.com/sktsherlock/HNGCL-V1.

Relevance models measure the semantic closeness between queries and the candidate ads, widely recognized as the nucleus of sponsored search systems. Conventional relevance models solely rely on the textual data within the queries and ads, whose performance is hindered by the scarce semantic information in these short texts. Recently, user behavior graphs have been incorporated to provide complementary information beyond pure textual semantics.Despite the promising performance, behavior-enhanced models suffer from exhausting resource costs due to the extra computations introduced by explicit topological aggregations. In this paper, we propose a novel Multi-Grained Topological Pre-Training paradigm, MGTLM, to teach language models to understand multi-grained topological information in behavior graphs, which contributes to eliminating explicit graph aggregations and avoiding information loss. Extensive experimental results over online and offline settings demonstrate the superiority of our proposal.

Many real-world graph learning tasks require handling dynamic graphs where new nodes and edges emerge. Dynamic graph learning methods commonly suffer from the catastrophic forgetting problem, where knowledge learned for previous graphs is overwritten by updates for new graphs. To alleviate the problem, continual graph learning methods are proposed. However, existing continual graph learning methods aim to learn new patterns and maintain old ones with the same set of parameters of fixed size, and thus face a fundamental tradeoff between both goals. In this paper, we propose Parameter Isolation GNN (PI-GNN) for continual learning on dynamic graphs that circumvents the tradeoff via parameter isolation and expansion. Our motivation lies in that different parameters contribute to learning different graph patterns. Based on the idea, we expand model parameters to continually learn emerging graph patterns. Meanwhile, to effectively preserve knowledge for unaffected patterns, we find parameters that correspond to them via optimization and freeze them to prevent them from being rewritten. Experiments on eight real-world datasets corroborate the effectiveness of PI-GNN compared to state-of-the-art baselines.

Cross-Domain Recommendation (CDR) is capable of incorporating auxiliary information from multiple domains to advance recommendation performance. Conventional CDR methods primarily rely on overlapping users, whereby knowledge is conveyed between the source and target identities belonging to the same natural person. However, such a heuristic assumption is not universally applicable due to an individual may exhibit distinct or even conflicting preferences in different domains, leading to potential noises. In this paper, we view the anchor links between users of various domains as the learnable parameters to learn the task-relevant cross-domain correlations. A novel optimal transport based model ALCDR is further proposed to precisely infer the anchor links and deeply aggregate collaborative signals from the perspectives of intra-domain and inter-domain. Our proposal is extensively evaluated over real-world datasets, and experimental results demonstrate its superiority.

Session-based recommendation (SBR) aims to predict the user's next action based on short and dynamic sessions. Recently, there has been an increasing interest in utilizing various elaborately designed graph neural networks (GNNs) to capture the pair-wise relationships among items, seemingly suggesting the design of more complicated models is the panacea for improving the empirical performance. However, these models achieve relatively marginal improvements with exponential growth in model complexity. In this paper, we dissect the classical GNN-based SBR models and empirically find that some sophisticated GNN propagations are redundant, given the readout module plays a significant role in GNN-based models. Based on this observation, we intuitively propose to remove the GNN propagation part, while the readout module will take on more responsibility in the model reasoning process. To this end, we propose the Multi-Level Attention Mixture Network (Atten-Mixer), which leverages both concept-view and instance-view readouts to achieve multi-level reasoning over item transitions. As simply enumerating all possible high-level concepts is infeasible for large real-world recommender systems, we further incorporate SBR-related inductive biases, i.e., local invariance and inherent priority to prune the search space. Experiments on three benchmarks demonstrate the effectiveness and efficiency of our proposal. We also have already launched the proposed techniques to a large-scale e-commercial online service since April 2021, with significant improvements of top-tier business metrics demonstrated in the online experiments on live traffic.

Graph neural networks (GNNs) have seen widespread usage across multiple real-world applications, yet in transductive learning, they still face challenges in accuracy, efficiency, and scalability, due to the extensive number of trainable parameters in the embedding table and the paradigm of stacking neighborhood aggregations. This paper presents a novel model called xGCN for large-scale network embedding, which is a practical solution for link predictions. xGCN addresses these issues by encoding graph-structure data in an extreme convolutional manner, and has the potential to push the performance of network embedding-based link predictions to a new record. Specifically, instead of assigning each node with a directly learnable embedding vector, xGCN regards node embeddings as static features. It uses a propagation operation to smooth node embeddings and relies on a Refinement neural Network (RefNet) to transform the coarse embeddings derived from the unsupervised propagation into new ones that optimize a training objective. The output of RefNet, which are well-refined embeddings, will replace the original node embeddings. This process is repeated iteratively until the model converges to a satisfying status. Experiments on three social network datasets with link prediction tasks show that xGCN not only achieves the best accuracy compared with a series of competitive baselines but also is highly efficient and scalable.

User identity linkage, which aims to link identities of a natural person across different social platforms, has attracted increasing research interest recently. Existing approaches usually first embed the identities as deterministic vectors in a shared latent space, and then learn a classifier based on the available annotations. However, the formation and characteristics of real-world social platforms are full of uncertainties, which makes these deterministic embedding based methods sub-optimal. Besides, semi-supervised models utilize the unlabeled data to help capture the intrinsic data distribution. However, the existing semi-supervised linkage methods heavily rely on the heuristically defined similarity measurements to incorporate the innate closeness between labeled and unlabeled samples. Such manually designed assumptions may not be consistent with the actual linkage signals and further introduce the noises. To address the mentioned limitations, in this paper we propose a novel Noise-aware Semi-supervised Variational User Identity Linkage (NSVUIL) model. Specifically, we first propose a novel supervised linkage module to incorporate the available annotations. Each social identity is represented by a Gaussian distribution in the Wasserstein space to simultaneously preserve the fine-grained social profiles and model the uncertainty of identities. Then, a noise-aware self-learning module is designed to faithfully augment the few available annotations, which is capable of filtering noises from the pseudo-labels generated by the supervised module. The filtered reliable candidates are added into the labeled set to provide enhanced training guidance for the next training iteration. Empirically, we evaluate the NSVUIL model over multiple real-world datasets, and the experimental results demonstrate its superiority.

Graph neural networks (GNNs) have been widely applied in the recommendation tasks and have achieved very appealing performance. However, most GNN-based recommendation methods suffer from the problem of data sparsity in practice. Meanwhile, pre-training techniques have achieved great success in mitigating data sparsity in various domains such as natural language processing (NLP) and computer vision (CV) . Thus, graph pre-training has the great potential to alleviate data sparsity in GNN-based recommendations. However, pre-training GNNs for recommendations faces unique challenges. For example, user-item interaction graphs in different recommendation tasks have distinct sets of users and items, and they often present different properties. Therefore, the successful mechanisms commonly used in NLP and CV to transfer knowledge from pre-training tasks to downstream tasks such as sharing learned embeddings or feature extractors are not directly applicable to existing GNN-based recommendations models. To tackle these challenges, we delicately design an adaptive graph pre-training framework for localized collaborative filtering (ADAPT) . It does not require transferring user/item embeddings, and is able to capture both the common knowledge across different graphs and the uniqueness for each graph simultaneously. Extensive experimental results have demonstrated the effectiveness and superiority of ADAPT.

Session-based recommendation (SBR) aims to predict the user's next action based on the ongoing sessions. Recently, there has been an increasing interest in modeling the user preference evolution to capture the fine-grained user interests. While latent user preferences behind the sessions drift continuously over time, most existing approaches still model the temporal session data in discrete state spaces, which are incapable of capturing the fine-grained preference evolution and result in sub-optimal solutions. To this end, we propose Graph Nested GRU ordinary differential equation (ODE), namely GNG-ODE, a novel continuum model that extends the idea of neural ODEs to continuous-time temporal session graphs. The proposed model preserves the continuous nature of dynamic user preferences, encoding both temporal and structural patterns of item transitions into continuous-time dynamic embeddings. As the existing ODE solvers do not consider graph structure change and thus cannot be directly applied to the dynamic graph, we propose a time alignment technique, called t-Alignment, to align the updating time steps of the temporal session graphs within a batch. Empirical results on three benchmark datasets show that GNG-ODE significantly outperforms other baselines.

Zhoujin Tian, Chaozhuo Li, Shuo Ren, Zhiqiang Zuo, Zengxuan Wen, Xinyue Hu, Xiao Han, Haizhen Huang, Denvy Deng, Qi Zhang, Xing Xie. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. 2022.

Popularity prediction is to predict the number of social network users involved in information diffusion. Recently, deep learning methods for popularity prediction advance traditional approaches that rely on hand-crafted features. However, existing approaches ignore the multi-source cascade that consists of multiple sub-cascades with different content but under the same topic. Different from single-source cascade, more cascading information can be observed from multi-source cascade and they are potentially correlated. How to correlate the diverse information and take advantage of them from both temporal and spatial aspects is critical for prediction. To this end, we propose a novel framework, called HEterogeneous Recurrent Integrated Graph Convolutional Neural Network (HERI-GCN). Specifically, we construct a heterogeneous cascade graph to model the multi-source cascade where time intervals are treated as heterogeneous time nodes. Besides, we propose a heterogeneous GCN to learn rich features from the multi-source cascade. RNN is organically integrated into the heterogeneous GCN to overcome the limited learning ability toward temporal and spatial data. We evaluate HERI-GCN through comparative experiments on three datasets. The experimental evaluation shows that HERI-GCN outperforms the state-of-the-art baseline methods.

The invariance to permutations of the adjacency matrix, i.e., graph isomorphism, is an overarching requirement for Graph Neural Networks (GNNs). Conventionally, this prerequisite can be satisfied by the invariant operations over node permutations when aggregating messages. However, such an invariant manner may ignore the relationships among neighboring nodes, thereby hindering the expressivity of GNNs. In this work, we devise an efficient permutation-sensitive aggregation mechanism via permutation groups, capturing pairwise correlations between neighboring nodes. We prove that our approach is strictly more powerful than the 2-dimensional Weisfeiler-Lehman (2-WL) graph isomorphism test and not less powerful than the 3-WL test. Moreover, we prove that our approach achieves the linear sampling complexity. Comprehensive experiments on multiple synthetic and real-world datasets demonstrate the superiority of our model.

The effectiveness of knowledge graph embedding (KGE) largely depends on the ability to model intrinsic relation patterns and mapping properties. However, existing approaches can only capture some of them with insufficient modeling capacity. In this work, we propose a more powerful KGE framework named HousE, which involves a novel parameterization based on two kinds of Householder transformations: (1) Householder rotations to achieve superior capacity of modeling relation patterns; (2) Householder projections to handle sophisticated relation mapping properties. Theoretically, HousE is capable of modeling crucial relation patterns and mapping properties simultaneously. Besides, HousE is a generalization of existing rotation-based models while extending the rotations to high-dimensional spaces. Empirically, HousE achieves new state-of-the-art performance on five benchmark datasets. Our code is available at https://github.com/anrep/HousE.

As the fundamental basis of sponsored search, relevance modeling measures the closeness between the input queries and the candidate ads. Conventional relevance models solely rely on the textual data, which suffer from the scarce semantic signals within the short queries. Recently, user historical click behaviors are incorporated in the format of click graphs to provide additional correlations beyond pure textual semantics, which contributes to advancing the relevance modeling performance. However, user behaviors are usually arbitrary and unpredictable, leading to the noisy and sparse graph topology. In addition, there exist other types of user behaviors besides clicks, which may also provide complementary information. In this paper, we study the novel problem of heterogeneous behavior graph learning to facilitate relevance modeling task. Our motivation lies in learning an optimal and task-relevant heterogeneous behavior graph consisting of multiple types of user behaviors. We further propose a novel HBGLR model to learn the behavior graph structure by mining the sophisticated correlations between node semantics and graph topology, and encode the textual semantics and structural heterogeneity into the learned representations. Our proposal is evaluated over real-world industry datasets, and has been mainstreamed in the Bing ads. Both offline and online experimental results demonstrate its superiority.

The wide spread of fake news has caused serious societal issues. We propose a subgraph reasoning paradigm for fake news detection, which provides a crystal type of explainability by revealing which subgraphs of the news propagation network are the most important for news verification, and concurrently improves the generalization and discrimination power of graph-based detection models by removing task-irrelevant information. In particular, we propose a reinforced subgraph generation method, and perform fine-grained modeling on the generated subgraphs by developing a Hierarchical Path-aware Kernel Graph Attention Network. We also design a curriculum-based optimization method to ensure better convergence and train the two parts in an end-to-end manner.

User-item interactions in recommendations can be naturally denoted as a user-item bipartite graph. Given the success of graph neural networks (GNNs) in graph representation learning, GNN-based Collaborative Filtering (CF) methods have been proposed to advance recommender systems. These methods often make recommendations based on the learned user and item embeddings. However, we found that they do not perform well with sparse user-item graphs which are quite common in real-world recommendations. Therefore, in this work, we introduce a novel perspective to build GNN-based CF methods for recommendations which leads to the proposed framework Localized Graph Collaborative Filtering (LGCF). One key advantage of LGCF is that it does not need to learn embeddings for each user and item, which is challenging in sparse scenarios. Alternatively, LGCF aims at encoding useful CF information into a localized graph and making recommendations based on such graph. Extensive experiments on various datasets validate the effectiveness of LGCF, especially in sparse scenarios. Furthermore, empirical results demonstrate that LGCF provides complementary information to the embedding-based CF model which can be utilized to boost recommendation performance.

A large-scale recommender system usually consists of recall and ranking modules. The goal of ranking modules (aka rankers) is to elaborately discriminate users' preference on item candidates proposed by recall modules. With the success of deep learning techniques in various domains, we have witnessed the mainstream rankers evolve from traditional models to deep neural models. However, the way that we design and use rankers remains unchanged: offline training the model, freezing the parameters, and deploying it for online serving. Actually, the candidate items are determined by specific user requests, in which underlying distributions (e.g., the proportion of items for different categories, the proportion of popular or new items) are highly different from one another in a production environment. The classical parameter-frozen inference manner cannot adapt to dynamic serving circumstances, making rankers' performance compromised.

Learning informative representations of users and items from the historical interactions is crucial to collaborative filtering (CF). Existing CF approaches usually model interactions solely within the Euclidean space. However, the sophisticated user-item interactions inherently present highly non-Euclidean anatomy with various types of geometric patterns (i.e., tree-likeness and cyclic structures). The Euclidean-based models may be inadequate to fully uncover the intent factors beneath such hybrid-geometry interactions. To remedy this deficiency, in this paper, we study the novel problem of Geometric Disentangled Collaborative Filtering (GDCF), which aims to reveal and disentangle the latent intent factors across multiple geometric spaces. A novel generative GDCF model is proposed to learn geometric disentangled representations by inferring the high-level concepts associated with user intentions and various geometries. Empirically, our proposal is extensively evaluated over five real-world datasets, and the experimental results demonstrate the superiority of GDCF.

Ad-hoc search calls for the selection of appropriate answers from a massive-scale corpus. Nowadays, the embedding-based retrieval (EBR) becomes a promising solution, where deep learning based document representation and ANN search techniques are allied to handle this task. However, a major challenge is that the ANN index can be too large to fit into memory, given the considerable size of answer corpus. In this work, we tackle this problem with Bi-Granular Document Representation, where the lightweight sparse embeddings are indexed and standby in memory for coarse-grained candidate search, and the heavyweight dense embeddings are hosted in disk for fine-grained post verification. For the best of retrieval accuracy, a Progressive Optimization framework is designed. The sparse embeddings are learned ahead for high-quality search of candidates. Conditioned on the candidate distribution induced by the sparse embeddings, the dense embeddings are continuously learned to optimize the discrimination of ground-truth from the shortlisted candidates. Besides, two techniques: the contrastive quantization and the locality-centric sampling are introduced for the learning of sparse and dense embeddings, which substantially contribute to their performances. Thanks to the above features, our method effectively handles massive-scale EBR with strong advantages in accuracy: with up to recall gain on million-scale corpus, and up to recall gain on billion-scale corpus. Besides, Our method is applied to a major sponsored search platform with substantial gains on revenue (), Recall () and CTR (). Our code is available at https://github.com/microsoft/BiDR.

These days, social media users tend to express their feelings through sharing images online. Capturing the emotions embedded in these social images involves great research challenges and practical values. Most existing works concentrate on extracting the visual feature from a global view, while ignoring the fact that visual objects are also rich in emotion. How to leverage the multilevel visual features to improve the sentiment analysis performance is important yet challenging. Besides, existing works view each social image as an independent sample while ignoring the rich correlations among social images, which may be helpful in detecting visual emotion. In this article, we propose a novel model called social relations-guided multiattention networks (SRGMANs) to incorporate both the multilevel (region-level and object-level) visual features of a single image and the correlations among multiple social images to conduct visual sentiment analysis. Specifically, we first construct a heterogeneous network consisting of various types of social relations and introduce a heterogeneous network embedding method to learn the network representation for each image. Then, two visual attention branches (region attention network and object attention network) are devised to extract emotional and discriminative visual features. For each branch, we design a self-attention module to capture the emotional dependencies among visual parts. Besides, a network-guided attention module is also designed in each branch to focus on more network-related emotional visual parts with the guidance of the topology information. Finally, the attended visual features from the two attention models, together with network representation features, are combined within a holistic framework to predict the sentiment of social images. Extensive experiments demonstrate the superiority of our model on three benchmark datasets.

Nowadays, it is common for one natural person to join multiple social networks to enjoy different types of services. User identity linkage (UIL), which aims to link identical identities across different social platforms, has attracted increasing research interests recently. Most existing approaches focus on the sophisticated architecture engineering of the linkage model but ignore the challenge of hubness in the post-processing nearest neighbor search phase. Hubness appears as some identities in a social platform, called hubs, being extra-ordinary close to the identities in the other platform, which will degrade the alignment performance. Different from existing heuristic methods, in this paper we propose a hubness-aware user identity linkage model HAUIL to smoothly learn hubless linkage signals. A carefully-designed objective function is presented to explicitly mitigate the hubness information from the pre-learned linkage guidance. HAUIL can be easily adapted to most existing UIL models. Empirically, we evaluate HAUIL over multiple publicly available datasets, and the experimental results demonstrate its superiority.

The representation learning on textual graph is to generate low-dimensional embeddings for the nodes based on the individual textual features and the neighbourhood information. Recent breakthroughs on pretrained language models and graph neural networks push forward the development of corresponding techniques. The existing works mainly rely on the cascaded model architecture: the textual features of nodes are independently encoded by language models at first; the textual embeddings are aggregated by graph neural networks afterwards. However, the above architecture is limited due to the independent modeling of textual features. In this work, we propose GraphFormers, where layerwise GNN components are nested alongside the transformer blocks of language models. With the proposed architecture, the text encoding and the graph aggregation are fused into an iterative workflow, making each node's semantic accurately comprehended from the global perspective. In addition, a progressive learning strategy is introduced, where the model is successively trained on manipulated data and original data to reinforce its capability of integrating information on graph. Extensive evaluations are conducted on three large-scale benchmark datasets, where GraphFormers outperform the SOTA baselines with comparable running efficiency. The source code is released at https://github.com/microsoft/GraphFormers .

Sponsored search ads appear next to search results when people look for products and services on search engines. In recent years, they have become one of the most lucrative channels for marketing. As the fundamental basis of search ads, relevance modeling has attracted increasing attention due to the significant research challenges and tremendous practical value. Most existing approaches solely rely on the semantic information in the input query-ad pair, while the pure semantic information in the short ads data is not sufficient to fully identify user's search intents. Our motivation lies in incorporating the tremendous amount of unsupervised user behavior data from the historical search logs as the complementary graph to facilitate relevance modeling. In this paper, we extensively investigate how to naturally fuse the semantic textual information with the user behavior graph, and further propose three novel AdsGNN models to aggregate topological neighborhood from the perspectives of nodes, edges and tokens. Furthermore, two critical but rarely investigated problems, domain-specific pre-training and long-tail ads matching, are studied thoroughly. Empirically, we evaluate the AdsGNN models over the large industry dataset, and the experimental results of online/offline tests consistently demonstrate the superiority of our proposal.

In the circumstance of social big data, sentiment analysis is attracting increasing attention for its capacity in understanding individuals' attitudes and feelings. Traditional sentiment analysis methods focus on single modality and become ineffective as enormous data are emerging on the social websites with multiple manifestations. In this article, multimodal learning approaches are proposed to capture the relations between image and text, which only stay at the region level and ignore the fact that the channels are also closely correlated with the semantic information. In addition, social images in the social platforms are closely connected by various types of relations, which are also conducice to sentiment classification but neglected by most existing works. In this article, we propose an attention-based heterogeneous relational model to improve the multimodal sentiment analysis performance by incorporating rich social information. Specifically, we propose a progressive dual attention module to capture the correlations between image and text, and then learn the joint image-text representation from the perspective of content information. A channel attention schema is proposed here to highlight semantically rich image channels and a region attention schema is further designed to highlight the emotional regions based on the attended channels. After that, we construct a heterogeneous relation network and extend graph convolutional network to aggregate the content information from social contexts as complements to learn high-quality representations of social images. Our proposal is thoroughly evaluated on two benchmark datasets, and experimental results demonstrate the superiority of the proposed model.

The analysis for social networks, such as the socially connected Internet of Things, has shown a deep influence of intelligent information processing technology on industrial systems for Smart Cities. The goal of social media representation learning is to learn dense, low-dimensional, and continuous representations for multimodal data within social networks, facilitating many real-world applications. Since social media images are usually accompanied by rich metadata (e.g., textual descriptions, tags, groups, and submitted users), simply modeling the image is not effective to learn the comprehensive information from social media images. In this work, we treat the image and its textual description as multimodal content, and transform other metainformation into the links between contents (such as two images marked by the same tag or submitted by the same user). Based on the multimodal content and social links, we propose a Deep Attentive Multimodal Graph Embedding model named DAMGE for more effective social image representation learning. We introduce both small- and large-scale datasets to conduct extensive experiments, of which the results confirm the superiority of the proposal on the tasks of social image classification and link prediction.

Nowadays, it is common for one natural person to join multiple social networks to enjoy different kinds of services. Linking identical users across multiple social networks, also known as social network alignment, is an important problem of great research challenges. Existing methods usually link social identities on the pairwise sample level, which may lead to undesirable performance when the number of available annotations is limited. Motivated by the isomorphism information, in this paper we consider all the identities in a social network as a whole and perform social network alignment from the distribution level. The insight is that we aim to learn a projection function to not only minimize the distance between the distributions of user identities in two social networks, but also incorporate the available annotations as the learning guidance. We propose three models SNNAu, SNNAb and SNNAo to learn the projection function under the weakly-supervised adversarial learning framework. Empirically, we evaluate the proposed models over multiple datasets, and the results demonstrate the superiority of our proposals.

With the increasing popularity and diversity of social media, users tend to join multiple social platforms to enjoy different types of services. User identity linkage, which aims to link identical identities across different social platforms, has attracted increasing research attentions recently. Existing methods usually focus on pairwise identity linkage between two platforms, which cannot piece up the information from multi-sources to depict the intrinsic figures of social users. In this paper, we propose a novel adversarial learning based framework MSUIL with partially shared generators to perform Semi-supervised User Identity Linkage across Multiple social networks. The isomorphism across multiple platforms is captured as the complementary to link identities. The insight is that we aim to learn the desirable projection functions (generators) to not only minimize the distance between the distributions of user identities in arbitrary pairs of platforms, but also incorporate the available annotations as the learning guidance. The projection functions of different platform pairs share partial parameters, which ensures MSUIL can capture the interdependencies among multiple platforms and improves the model efficiency. Empirically, we evaluate our proposal over multiple datasets. The experimental results demonstrate the superiority of the proposed MSUIL model.

Network embedding, as a promising way of the network representation learning, is capable of supporting various subsequent network mining and analysis tasks, and has attracted growing research interests recently. Traditional approaches assign each node with an independent continuous vector, which will cause memory overhead for large networks. In this paper we propose a novel multi-hot compact network embedding framework to effectively reduce memory cost by learning partially shared embeddings. The insight is that a node embedding vector is composed of several basis vectors according to a multi-hot index vector. The basis vectors are shared by different nodes, which can significantly reduce the number of continuous vectors while maintain similar data representation ability. Specifically, we propose a MCNE$_p $ model to learn compact embeddings from pre-learned node features. A novel component named compressor is integrated into MCNE$_p $ to tackle the challenge that popular back-propagation optimization cannot propagate loss through discrete samples. We further propose an end-to-end model MCNE$_t $ to learn compact embeddings from the input network directly. Empirically, we evaluate the proposed models over four real network datasets, and the results demonstrate that our proposals can save about 90% of memory cost of network embeddings without significantly performance decline.

In this paper, we study the problem of modeling users' diverse interests. Previous methods usually learn a fixed user representation, which has a limited ability to represent distinct interests of a user. In order to model users' various interests, we propose a Memory Attention-aware Recommender System (MARS). MARS utilizes a memory component and a novel attentional mechanism to learn deep adaptive user representations. Trained in an end-to-end fashion, MARS adaptively summarizes users' interests. In the experiments, MARS outperforms seven state-of-the-art methods on three real-world datasets in terms of recall and mean average precision. We also demonstrate that MARS has a great interpretability to explain its recommendation results, which is important in many recommendation scenarios.

Hao Wang, Bing Liu, Chaozhuo Li, Yan Yang, Tianrui Li. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Existing recommendation methods mostly learn fixed vectors for users and items in a low-dimensional continuous space, and then calculate the popular dot-product to derive user-item distances. However, these methods suffer from two drawbacks: (1) the data sparsity issue prevents from learning high-quality representations; and (2) the dot-product violates the crucial triangular inequality and therefore, results in a sub-optimal performance. In this work, in order to overcome the two aforementioned drawbacks, we propose Deep Distribution Network (DDN) to model users and items via Gaussian distributions. We argue that, compared to fixed vectors, distribution-based representations are more powerful to characterize users' uncertain interests and items' distinct properties. In addition, we propose a Wasserstein-based loss, in which the critical triangular inequality can be satisfied. In experiments, we evaluate DDN and comparative models on standard datasets. It is shown that DDN significantly outperforms state-of-the-art models, demonstrating the advantages of the proposed distribution-based representations and wassertein loss.

Nowadays, it is common for one natural person to join multiple social networks to enjoy different services. Linking identical users across different social networks, also known as the User Identity Linkage (UIL), is an important problem of great research challenges and practical value. Most existing UIL models are supervised or semi-supervised and a considerable number of manually matched user identity pairs are required, which is costly in terms of labor and time. In addition, existing methods generally rely heavily on some discriminative common user attributes, and thus are hard to be generalized. Motivated by the isomorphism across social networks, in this paper we consider all the users in a social network as a whole and perform UIL from the user space distribution level. The insight is that we convert the unsupervised UIL problem to the learning of a projection function to minimize the distance between the distributions of user identities in two social networks. We propose to use the earth mover's distance (EMD) as the measure of distribution closeness, and propose two models UUIL$_gan $ and UUIL$_omt $ to efficiently learn the distribution projection function. Empirically, we evaluate the proposed models over multiple social network datasets, and the results demonstrate that our proposal significantly outperforms state-of-the-art methods.

The use of large-scale machine learning~(ML) is becoming ubiquitous in various domains ranging from business intelligence to self-driving cars. Many companies are building ML pipelines in a unified data processing environment, and leveraging well-tuned numerical optimization packages for obtaining model parameters. However, most existing optimization tools are specifically designed for a single machine setup, and cannot handle vast volume of data. In this work, we build a distributed computing framework towards optimizing generalized linear models with billions of variables. We at first design a new distributed vector to represent data points from extremely large feature space. Then, we introduce an efficient and scalable approach to compute the second order derivatives of loss function, and optimizes model parameters with limited memory requirement. Experiments on real-world datasets demonstrate that our proposed techniques can scale up for ML models with billions of variables, and achieves better performance than state-of-the-art systems on a wide range of applications, e.g., ad CTR prediction and rideshare price bidding.

The explosive use of social media makes it a popular platform for malicious users, known as social spammers, to overwhelm legitimate users with unwanted content. Most existing social spammer detection approaches are supervised and need a large number of manually labeled data for training, which is infeasible in practice. To address this issue, some semi-supervised models are proposed by incorporating side information such as user profiles and posted tweets. However, these shallow models are not effective to deeply learn the desirable user representations for spammer detection, and the multi-view data are usually loosely coupled without considering their correlations. In this paper, we propose a Semi-Supervised Deep social spammer detection model by Multi-View data fusion (SSDMV). The insight is that we aim to extensively learn the task-relevant discriminative representations for users to address the challenge of annotation scarcity. Under a unified semi-supervised learning framework, we first design a deep multi-view feature learning module which fuses information from different views, and then propose a label inference module to predict labels for users. The mutual refinement between the two modules ensures SSDMV to be able to both generate high quality features and make accurate predictions.Empirically, we evaluate SSDMV over two real social network datasets on three tasks, and the results demonstrate that SSDMV significantly outperforms the state-of-the-art methods.

Learning the embedding for social media data has attracted extensive research interests as well as boomed a lot of applications, such as classification and link prediction. In this paper, we examine the scenario of a multimodal network with nodes containing multimodal contents and connected by heterogeneous relationships, such as social images containing multimodal contents (e.g., visual content and text description), and linked with various forms (e.g., in the same album or with the same tag). However, given the multimodal network, simply learning the embedding from the network structure or a subset of content results in sub-optimal representation. In this paper, we propose a novel deep embedding method, i.e., Attention-based Multi-view Variational Auto-Encoder (AMVAE), to incorporate both the link information and the multimodal contents for more effective and efficient embedding. Specifically, we adopt LSTM with attention model to learn the correlation between different data modalities, such as the correlation between visual regions and the specific words, to obtain the semantic embedding of the multimodal contents. Then, the link information and the semantic embedding are considered as two correlated views. A multi-view correlation learning based Variational Auto-Encoder (VAE) is proposed to learn the representation of each node, in which the embedding of link information and multimodal contents are integrated and mutually reinforced. Experiments on three real-world datasets demonstrate the superiority of the proposed model in two applications, i.e., multi-label classification and link prediction.

As an effective way of learning node representations in networks, network embedding has attracted increasing research interests recently. Most existing approaches use shallow models and only work on static networks by extracting local or global topology information of each node as the algorithm input. It is challenging for such approaches to learn a desirable node representation on incomplete graphs with a large number of missing links or on dynamic graphs with new nodes joining in. It is even challenging for them to deeply fuse other types of data such as node properties into the learning process to help better represent the nodes with insufficient links. In this paper, we for the first time study the problem of network embedding on incomplete networks. We propose a Multi-View Correlation-learning based Deep Network Embedding method named MVC-DNE to incorporate both the network structure and the node properties for more effectively and efficiently perform network embedding on incomplete networks. Specifically, we consider the topology structure of the network and the node properties as two correlated views. The insight is that the learned representation vector of a node should reflect its characteristics in both views. Under a multi-view correlation learning based deep autoencoder framework, the structure view and property view embeddings are integrated and mutually reinforced through both self-view and cross-view learning. As MVC-DNE can learn a representation mapping function, it can directly generate the representation vectors for the new nodes without retraining the model. Thus it is especially more efficient than previous methods. Empirically, we evaluate MVC-DNE over three real network datasets on two data mining applications, and the results demonstrate that MVC-DNE significantly outperforms state-of-the-art methods.

While automatic response generation for building chatbot systems has drawn a lot of attention recently, there is limited understanding on when we need to consider the linguistic context of an input text in the generation process. The task is challenging, as messages in a conversational environment are short and informal, and evidence that can indicate a message is context dependent is scarce. After a study of social conversation data crawled from the web, we observed that some characteristics estimated from the responses of messages are discriminative for identifying context dependent messages. With the characteristics as weak supervision, we propose using a Long Short Term Memory (LSTM) network to learn a classifier. Our method carries out text representation and classifier learning in a unified framework. Experimental results show that the proposed method can significantly outperform baseline methods on accuracy of classification.