While Reinforcement Learning (RL) has advanced LLM reasoning, applying it to longcontext scenarios is hindered by sparsity of outcome rewards.This limitation fails to penalize ungrounded "lucky guesses," leaving the critical process of needle-in-a-haystack evidence retrieval largely unsupervised.To address this, we propose EAPO (Evidence-Augmented Policy Optimization).We first establish the Evidence-Augmented Reasoning paradigm, validating via Tree-Structured Evidence Sampling that precise evidence extraction is the decisive bottleneck for long-context reasoning.Guided by this insight, EAPO introduces a specialized RL algorithm where a reward model computes a Group-Relative Evidence Reward, providing dense process supervision to explicitly improve evidence quality.To sustain accurate supervision throughout training, we further incorporate an Adaptive Reward-Policy Co-Evolution mechanism.This mechanism iteratively refines the reward model using outcome-consistent rollouts, sharpening its discriminative capability to ensure precise process guidance.Comprehensive evaluations across eight benchmarks demonstrate that EAPO significantly enhances long-context reasoning performance compared to SOTA baselines.

Ruwen Zhang, Bo Liu, Zhang Sheng Xiang, Yida Chen, Hantao Zhao, Ding Ding, Jiahui Jin, Jiuxin Cao. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

People frequently face preference-based planning tasks requiring balancing goals with nuanced constraints, yet even advanced LLMs demand considerable effort to produce and adjust plans reflecting complex user preferences. We present MAVIS (Multi-Agent Virtual Interactive Synergy), a multi-agent system within a virtual workspace that introduces an incremental collaboration mechanism. This mechanism automatically decomposes tasks into guidelines and sequentially introduces expert agents. Each agent proactively engages users in focused dialog to uncover implicit preferences, while successive agents add perspectives and transparently negotiate trade-offs. To mitigate textual overload, MAVIS employs spatial visualizations that externalize agents’ reasoning through step-linked summaries and context-aware boards, with embodied avatars supporting natural interaction. Across studies, Study 1 showed our collaboration mechanism doubled expressed preferences and improved planning quality by 60.3% over a conventional LLM baseline. Study 2 affirmed visualization’s benefits over a non-spatial baseline, while Study 3 confirmed its versatility across VR and desktop modalities and diverse tasks.

The rapid emergence of smart homes offers convenience but also raises significant privacy concerns, such as challenges in comprehending privacy-related statuses, information, and settings. This study aims to mitigate these concerns by leveraging augmented reality technology to enhance privacy protection user experiences. Through interviews with experienced users in the Chinese culture context, we systematically identify and categorize three primary aspects of concerns encountered during smart home usage: sensor range and status, data transmission directions, and transparency of mode settings. Drawing on these insights, we design and implement a streamlined AR system for interacting with smart home devices. Our user studies demonstrate that our AR system offers significant advantages over traditional 2D smart home applications in terms of information representation, workload reduction (i.e., number of clicks), and enhanced privacy awareness. This research presents a notable step forward in privacy protection while suggesting pathways for AR-enhanced smart home systems.

Multi-label image classification is an essential task in computer vision that aims to identify multiple objects in images. Recently, there has been growing research interest in modeling the relationships between labels to enhance label representation learning. An intuitive approach is to train a network to estimate label co-occurrence probabilities in a supervised manner, which are then leveraged to guide the interactions between label representations. However, the extreme sparsity of label co-occurrence signals poses substantial challenges. To address this issue, we commence by examining the potential interaction behaviors between label representations under the guidance of ground-truth label co-occurrence signals. Inspired by our findings, a novel contrastive learning mechanism is crafted to mimic and enhance such behaviors, facilitating effective label representation interactions without relying on explicit supervision from label co-occurrence signals. Based on this, we develop a pioneering contrastive co-occurrence learning framework, which operates on the instance-level label co-occurrence graph for multi-label image classification. This framework involves sequential processes of label representation learning followed by co-occurrence perception learning. Cross-entropy loss for label classification learning and contrastive loss for co-occurrence perception learning are used to jointly optimize the entire framework end-to-end. In this way, label representations can interact effectively, fully perceiving their co-occurrence relationships at the instance level, thereby significantly improving the performance in label recognition. Extensive experiments on public benchmarks demonstrate the superiority of the proposed framework in multi-label image classification. Codes are available on https://github.com/jasonseu/CoCo.

With the rapid development of the Internet, the information dissemination paradigm has changed and the efficiency has been improved greatly. While this also brings the quick spread of fake news and leads to negative impacts on cyberspace. Currently, the information presentation formats have evolved gradually, with the news formats shifting from texts to multimodal contents. As a result, detecting multimodal fake news has become one of the research hotspots. However, multimodal fake news detection research field still faces two main challenges: the inability to fully and effectively utilize multimodal information for detection, and the low credibility or static nature of the introduced external information, which limits dynamic updates. To bridge the gaps, we propose ERIC-FND, an external reliable information-enhanced multimodal contrastive learning framework for fake news detection. ERIC-FND strengthens the representation of news contents by entity-enriched external information enhancement method. It also enriches the multimodal news information via multimodal semantic interaction method where the multimodal constrative learning is employed to make different modality representations learn from each other. Moreover, an adaptive fusion method is taken to integrate the news representations from different dimensions for the eventual classification. Experiments are done on two commonly used datasets in different languages, X (Twitter) and Weibo. Experiment results demonstrate that our proposed model ERIC-FND outperforms existing state-of-the-art fake news detection methods under the same settings.

Proactive questioning is essential in psychological conversations as it helps uncover deeper issues and unspoken concerns.Current psychological LLMs are constrained by passive response mechanisms, limiting their capacity to deploy proactive strategies for psychological counseling.To bridge this gap, we first develop the ProPsyC (Proactive Psychological Conversation) dataset, a multi-turn conversation dataset with interpretive labels including strategy decision logic and reaction attribution.Based on ProPsyC, we propose PsyAdvisor by supervised fine-tuning, a plug-and-play proactive questioning strategy planner that empowers psychological LLMs to initiate welltimed questioning through strategic prompting.Experimental results demonstrate that psychological LLMs integrated with PsyAdvisor substantially improve proactive questioning capacity, conversation depth, and response quality.Furthermore, PsyAdvisor shows promising potential in assisting novice counselors by providing strategy recommendations.This study provides new optimization directions for psychological conversation systems and offers valuable insights for future research on proactive questioning mechanisms in psychological LLMs.Our code are available at https: //github.com/EthanHu777/PsyAdvisor.

With the quick advancement in text generation ability of Large Language Mode(LLM), concerns about the misuse of machine-generated content have grown, raising potential violations of legal and ethical standards. Some existing studies concentrate on detecting machine-generated text in open-source models using in-model features, but their performance on closed-source large models is limited. This limitation occurs because, in the closed-source model detection, the only reference that can be obtained is the texts, which may differ significantly due to random sampling. In this paper, we demonstrate that texts generated by the same model can align both semantically and statistically under similar prompts, facilitating effective detection and traceability. Specifically, we fine-tune a BERT encoder through contrastive learning to achieve semantic alignment in randomly generated texts from the same model. Then, we propose a method called Machine-Generated Text Detection with Rewritten Texts, which designed several prompt refactoring methods and used them to request rewritten text from LLMs. Semantic and statistical relationships between rewritten and original texts provide a basis for detection and traceability. Finally, we expanded the text dataset with multi-parameter random sampling and verified the performance of MAGRET on three text-generated datasets. Experimental results show that previous methods struggle with closed-source model detection, while our approach significantly outperforms baseline methods in this regard. It also shows MagRet’s stable performance in detection and tracing tasks across various randomly sampled texts.

Differing from sentiment transfer, positive reframing seeks to substitute negative perspectives with positive expressions while preserving the original meaning. With the emergence of pre-trained language models (PLMs), it is possible to achieve acceptable results by fine-tuning PLMs. Nevertheless, generating fluent, diverse and task-constrained reframing text remains a significant challenge. To tackle this issue, a multi-strategy optimization framework (MSOF) is proposed in this paper. Starting from the objective of positive reframing, we first design positive sentiment reward and content preservation reward to encourage the model to transform the negative expressions of the original text while ensuring the integrity and consistency of the semantics. Then, different decoding optimization approaches are introduced to improve the quality of text generation. Finally, based on the modeling formula of positive reframing, we propose a multi-dimensional re-ranking method that further selects candidate sentences from three dimensions: strategy consistency, text similarity and fluency. Extensive experiments on two Seq2Seq PLMs, BART and T5, demonstrate our framework achieves significant improvements on unconstrained and controlled positive reframing tasks.

CLIP, widely used in multimodal learning, excels due to its large-scale image-text pretraining. However, applying CLIP-like architectures to skeleton-based action representation learning presents challenges due to the incompatible non-visual data structure and the limited scale of skeleton datasets, which hinder robust generalization. To address these issues, we propose SKL-CLIP, a framework that incorporates Supervised Self-Contrastive Learning to mitigate overfitting and enhance transferable representation learning, Knowledge Distillation from the textual encoder of pretrained CLIP models to preserve generalization while adapting to skeleton-text scenarios, and Multi-Domain Parallel Training to leverage diverse support datasets, improving cross-dataset and zero-shot recognition. Extensive experiments on NTU and PKU datasets demonstrate that SKL-CLIP significantly advances skeleton-based action representation, achieving state-of-the-art performance across fully-supervised, cross-data unsupervised domain adaptation, and zero-shot tasks.

Current text-driven Video Moment Retrieval (VMR) methods encode all video clips, including irrelevant ones, disrupting multimodal alignment and hindering optimization. To this end, we propose a denoise-then-retrieve paradigm that explicitly filters text-irrelevant clips from videos and then retrieves the target moment using purified multimodal representations. Following this paradigm, we introduce the Denoise-then-Retrieve Network (DRNet), comprising Text-Conditioned Denoising (TCD) and Text-Reconstruction Feedback (TRF) modules. TCD integrates cross-attention and structured state space blocks to dynamically identify noisy clips and produce a noise mask to purify multimodal video representations. TRF further distills a single query embedding from purified video representations and aligns it with the text embedding, serving as auxiliary supervision for denoising during training. Finally, we perform conditional retrieval using text embeddings on purified video representations for accurate VMR. Experiments on Charades-STA and QVHighlights demonstrate that our approach surpasses state-of-the-art methods on all metrics. Furthermore, our denoise-then-retrieve paradigm is adaptable and can be seamlessly integrated into advanced VMR models to boost performance.

Hate Speech Detection (HSD) plays a crucial role in ensuring respectful online communication and preventing the spread of harmful content. However, existing studies on HSD often focus on the overall intent of the target message while overlooking the fine-grained details in the message. We propose a RatIonale-guided multi-taSk lEarning framework (RISE), which frames HSD as the main task and Human Rationales Tagging (HRT) as the auxiliary task. This enables simultaneous message-level understanding and the identification of expressions that trigger human judgments of hate speech, mimicking human cognitive processing to capture fine-grained information and enhance HSD performance. Furthermore, we extend rationale annotations from binary labels to BIO tagging, capturing the positional roles of tokens within rationales. Additionally, we integrate an emoji semantics interpretation module that interprets emoji meanings, enriching contextual information. Extensive experiments demonstrate that RISE outperforms state-of-the-art models, with each component contributing significantly to improved performance.

The performance of current Vision-Language Tracking (VLT) models is constrained by the limited diversity and quantity of labeled data. Compared to constructing large-scale datasets, data augmentation offers a more cost-saving strategy for VLT by synthesizing new samples from existing data, rather than generating them from scratch. However, conventional techniques like rotation and flipping may disrupt scene composition, causing conflicts between visual layouts and textual annotations. Recent advances in generative models have inspired the use of synthetic videos for data augmentation. Yet, existing approaches fail to address the core concerns of data augmentation in VLT (shown in Fig. 1)-target location accuracy, text-video consistency, and video content coherency. To bridge the gap, we propose Gen4Track, a tuning-free data augmentation framework that leverages the self-correcting mechanism to dynamically generate high-quality video data with annotations. Our approach involves (1) optimizing the attention calculations in a frozen text-to-image diffusion model to synthesize coherent videos that satisfy specific conditions (e.g., spatial location, category, color, and style), and (2) implementing a self-correcting mechanism based on a Large Language Model (LLM) to improve text-video consistency. During video augmentation, we propose content-coherent self-attention and location-enhanced cross-attention mechanisms, ensuring that image-level editings are accurately and coherently propagated throughout the video. Then, with the goal of maximizing text-video consistency, we iteratively refine the augmentation instruction with our designed self-correcting mechanism for a more aligned video. Extensive experiments validate that Gen4Track significantly boosts the performance of SOTA VLT models (achieving improvements of up to 3.2% in SUC and 3.5% in PRE), opening a new chapter of training Vision-Language trackers with synthetic videos rather than manually annotated data.

Secrets, whether structured like API keys or un-structured like passwords, are essential for securing applications and services. However, the growing use of open -source projects and rapid development cycles has amplified the risk of secret leakage. Current detection tools suffer from high false positive rates and low recall due to simplistic methods like regular expressions and entropy checks, often missing unstructured secrets or mislabeling non-sensitive data. In this paper, we introduce Keysentinel, an advanced automated secret detection tool that addresses these limitations through machine learning, semantic analysis, and prefix matching. To evaluate KEYSENTINEL, we created the first cross-platform benchmark with 11,826 labeled secrets in 1,806,530 files across GitHub, PyPI, and WeChat. We compare Key-sentinelwith six currently available tools. The results show KEYSENTINEL achieves state-of-the-art performance, with precision (91.18%), recall (81.71%), and an F1 score (0.86), surpassing industry-standard tools and significantly reducing false positives. It also outperforms large language models like GPT-4 and o1 in accuracy and cost-effectiveness. Besides, we conduct a large-scale measurement study, analyzing 80,330,098 files from GitHub, PyPI, and WeChat. We found that up to 30% of projects are at risk of secret leaks. Furthermore, we also scan the code base of an IT company to assess real-world secret leakage risks. Our findings underscore the pervasive nature of secret leaks and highlight the urgent need for enhanced secret management practices across platforms.

Online social network has emerged as a prominent place for the propagation of fake news due to its low cost of information dissemination. Although the existing methods have made many attempts in news content and propagation structure, the detection of fake news is still facing two challenges: one is how to mine the unique key features and evolution patterns, and the other is how to tackle the problem of small samples to build the high-performance model. Different from popular methods, which take full advantage of the propagation topology structure, in this article, we propose a novel framework for fake news detection from perspectives of semantics, emotion and data enhancement. The semantic and emotional features of news and comments, the inconsistent emotion between news and news participants as well as the emotion evolution features in comments are fused by the designed dual deep interaction channel network to obtain a more comprehensive and fine-grained news representation. Meanwhile, with the construction of large language model (LLM) prompt, a LLM-based data enhancement module is used to obtain more diverse labeled data of high quality filtered by confidence, further improving the performance of the classification model. Experiments show that the proposed approach outperforms the state-of-the-art methods.

In the context of rapid urbanization, traditional manual guidance and static evacuation signs are increasingly inadequate for addressing complex and dynamic emergencies. This study proposes an innovative emergency evacuation framework that optimizes the crowd evacuation by integrating multiagent reinforcement learning (MARL) with adversarial reinforcement learning (ARL). The developed simulation environment models realistic human behavior in complex buildings and incorporates robotic navigation and intelligent path planning. A novel simulated human behavior model was integrated, capable of complex human–robot interaction, independent escape route searching, and exhibiting herd mentality and memory mechanisms. We also proposed a multiagent framework that combines MARL and ARL to enhance overall evacuation efficiency and robustness. Additionally, we developed a new ARL evaluation framework that provides a novel method for quantifying agents’ performance. Various experiments of differing difficulty levels were conducted, and the results demonstrate that the proposed framework exhibits advantages in emergency evacuation scenarios. Specifically, our ARLR approach increased survival rates by 1.8% points in low-difficulty evacuation tasks compared to the RLR approach using only MARL algorithms. In high-difficulty evacuation tasks, the ARLR approach raised survival rates from 46.7% without robots to 64.4%, exceeding the RLR approach by 1.7% points. This study aims to enhance the efficiency and safety of human–robot collaborative fire evacuations and provides theoretical support for evaluating and improving the performance and robustness of ARL agents.

Gradually infiltrating through latent malicious connections to form cyberattacks has become a significant threat in cyberspace. Advanced persistent threats (APTs) are such cyberattacks, where intruders maintain a presence on networks to steal secrets. Detecting APTs by comparing user behavior with normal activity is challenging because attackers often mimic legitimate behavior sequences to evade detection. This article detects APTs by modeling the graph with interconnection intervals embedding to weaken attackers’ mimicry, namedCyberProber. We first construct an attack intent graph (AIG) to scale down the original network, while preserving key suspicious paths. Then, we embed the AIG by learning connection representations with interconnection interval features, capturing both structural and temporal dependencies. In this way, we can detect APTs in the small AIG, and further uncover more anomalies in the original network based on the identified compromised nodes in the AIG. Our experiments demonstrate that CyberProber outperforms baselines in APTs detection.

Industrial anomaly detection aims to identify and localize defective regions in images. Among various architectures, reconstruction-based methods have demonstrated exceptional performance. These methods reconstruct anomalous samples into normal ones and identify anomalies through a comparison between them. However, reconstruction process within these methods often focuses on PL similarity, overlooking the high-frequency consistency between the input and output, which constrains the model’s accuracy. This article proposes tGARD, a novel text-guided adversarial reconstruction method for anomaly detection. Specifically, we introduce feature aggregation module, using nonlocal block and dilated convolution to handle complex anomaly patterns. Subsequently, text-guided reconstruction module is meticulously designed to harness CLIP’s multimodal alignment capabilities, allowing for a semantically controllable reconstruction process. This control is achieved by incorporating dynamic text embeddings derived from the CLIP encoder within discriminator. Meanwhile, during reconstruction process, high-frequency details are preserved through a convolutional adversarial discriminator. Finally, category-aware loss weighting strategy is conceived to balance similarity and adversarial loss. Experiments demonstrate that our model achieves significant improvements in anomaly localization, surpassing all reconstruction-based models on MVTec-AD. It also establishes a new state-of-the-art on VisA dataset, outperforming all existing architectures.

Current methods for Video Moment Retrieval (VMR) struggle to align complex situations involving specific environmental details, character descriptions, and action narratives. To tackle this issue, we propose a Large Language Model-guided Moment Retrieval (LMR) approach that employs the extensive knowledge of Large Language Models (LLMs) to improve video context representation as well as cross-modal alignment, facilitating accurate localization of target moments. Specifically, LMR introduces a context enhancement technique with LLMs to generate crucial target-related context semantics. These semantics are integrated with visual features for producing discriminative video representations. Finally, a language-conditioned transformer is designed to decode free-form language queries, on the fly, using aligned video representations for moment retrieval. Extensive experiments demonstrate that LMR achieves state-of-the-art results, outperforming the nearest competitor by up to 3.28% and 4.06% on the challenging QVHighlights and Charades-STA benchmarks, respectively. More importantly, the performance gains are significantly higher for localization of complex queries.

Single object tracking aims to locate one specific target in video sequences, given its initial state. Classical trackers rely solely on visual cues, restricting their ability to handle challenges such as appearance variations, ambiguity, and distractions. Hence, Vision-Language Tracking (VLT) has emerged as a promising approach, incorporating language descriptions to directly provide high-level semantics and enhance tracking performance. However, current Vision-Language (VL) trackers have not fully exploited the power of multi-modal learning, as they suffer from limitations such as heavily relying on off-the-shelf backbones for feature extraction, ineffective asynchronous fusion designs, and the absence of VL-related loss functions for optimizing multi-modal representation. Consequently, we present a novel tracker that progressively explores target-centric semantics for VLT. Specifically, we propose the first Synchronous Learning Backbone (SLB) for VLT, which consists of two novel modules: the Target Enhance Module (TEM) and the Semantic-Aware Module (SAM). These modules together ensure the multi-modal feature extraction and interaction at the same pace, facilitating the tracker to synchronously perceive target-related semantics from both visual and textual modalities. Moreover, we devise the dense matching loss to further strengthen multi-modal representation learning. Extensive experiments on VLT datasets demonstrate the superiority and effectiveness of our methods.

Skeleton-based Action Recognition (SAR) is widely recognized for its robustness and efficiency in human action analysis, but its performance in cross-dataset tasks has been limited due to domain shifts between different datasets. To address this challenge, current methods typically approach cross-dataset SAR as an Unsupervised Domain Adaptation (UDA) task, which is tackled using domain adaptation or self-supervised learning strategies. In this article, we propose a Dual-Domain Triple Contrast (D2TC) framework for cross-dataset SAR under the UDA setting. Unlike existing UDA methods that either focus on a single strategy or superficially combine strategies, our D2TC leverages contrastive learning to integrate both strategies into a unified framework. It performs three types of contrastive learning: Self-Supervised Contrastive Learning, Supervised Contrastive Learning, and UDA with Contrastive Learning, across both source and target domains. The triple contrasts go beyond mere summation, effectively bridging the domain gap and enhancing the model’s representational capacity. Additionally, we introduce multi-modal ensemble contrast and extreme skeleton augmentation methods to further enhance the skeleton-based representation learning. Extensive experiments on six cross-dataset settings validate the superiority of our D2TC framework over state-of-the-art methods, demonstrating its effectiveness in reducing domain discrepancies and improving cross-dataset SAR performance. The codes are available on https://github.com/KennCoder7/DualDomainTripleContrast .

Identifying labels that are unseen during training, known as multi-label zero-shot learning, is a non-trivial task in computer vision. Recent studies have increasingly focused on utilizing vision-language pre-training (VLP) models to recognize unseen labels in an open-vocabulary manner. However, these approaches like knowledge distillation have offered only modest performance gains. The challenge of fully harnessing the potential of VLP models for effective multi-label zero-shot learning remains open. In this work, an advanced query-based knowledge sharing framework is proposed to explore the multi-modal knowledge from VLP models for open-vocabulary multi-label classification. Specifically, we introduce a set of label-agnostic query tokens that are designed to capture essential and informative visual knowledge from input images. These tokens are subsequently shared across all labels, allowing them to select pertinent one as visual clues for accurate recognition. Then, by integrating the pre-trained knowledge of VLP models, these query tokens, trained on seen labels, can be efficiently generalized to the recognition of unseen labels. Additionally, we reformulate ranking learning into a form of classification to enable the magnitude of feature vectors for prediction, which significantly benefits label recognition. Experiment results show that our framework outperforms state-of-the-art methods in multi-label zero-shot learning task by a significant margin, reaching 4.2% and 2.4% in mAP on the NUS-WIDE and Open Images datasets, respectively. Code and models are available at https://github.com/jasonseu/QKS .

Emotion Recognition in Conversation (ERC) has attracted increasing attention due to its wide applications in public opinion analysis, empathetic conversation generation, and so on. However, ERC research suffers from the problems of data imbalance and the presence of similar linguistic expressions for different emotions. These issues can result in limited learning for minority emotions, biased predictions for common emotions, and the misclassification of different emotions with similar linguistic expressions. To alleviate these problems, we propose a Contrast-Enhanced Prompt-Tuning (CEPT) framework for ERC. We transform the ERC task into a Masked Language Modeling (MLM) generation task and generate the emotion for each utterance in the conversation based on the prompt-tuning of the Pre-trained Language Model (PLM), where a novel mixed prompt template and a label mapping strategy are introduced for better context and emotion feature modeling. Moreover, Supervised Contrastive Learning (SCL) is employed to help the PLM mine more information from the labels and learn a more discriminative representation space for utterances with different emotions. We conduct extensive experiments and the results demonstrate that CEPT outperforms the state-of-the-art methods on all three benchmark datasets and excels in recognizing minority emotions.

The excellent text-to-image synthesis capability of diffusion models has driven progress in synthesizing coherent visual stories. The current state-of-the-art method combines the features of historical captions, historical frames, and the current captions as conditions for generating the current frame. However, this method treats each historical frame and caption as the same contribution. It connects them in order with equal weights, ignoring that not all historical conditions are associated with the generation of the current frame. To address this issue, we propose Causal-Story. This model incorporates a local causal attention mechanism that considers the causal relationship between previous captions, frames, and current captions. By assigning weights based on this relationship, Causal-Story generates the current frame, thereby improving the global consistency of story generation. We evaluated our model on the PororoSV and FlintstonesSV datasets and obtained state-of-the-art FID scores, and the generated frames also demonstrate better storytelling in visuals.

Vision-Language Tracking (VLT) requires locating a specific target in video sequences, given a natural language prompt and an initial object box. Despite recent advancements, existing approaches heavily rely on expensive and time-consuming human annotations. To mitigate this limitation, directly generating pseudo labels from raw videos seems to be a straightforward solution; however, it inevitably introduces undesirable noise during the training process. Moreover, we insist that an efficient tracker should excel in tracking the target, regardless of the temporal direction. Building upon these insights, we propose the pioneering semi-supervised learning scheme for VLT task, representing a crucial step towards reducing the dependency on high-quality yet costly labeled data. Specifically, drawing inspiration from the natural attributes of a video (i.e., space, time, and semantics), our approach progressively leverages inherent consistencies from these aspects: (1) Spatially, each frame and any object cropped from it naturally form an image-bbox (bounding box) pair for self-training; (2) Temporally, bidirectional tracking trajectories should exhibit minimal differences; (3) Semantically, the correlation between visual and textual features is expected to remain consistent. Furthermore, the framework is validated with a simple yet effective tracker we devised, named ATTracker (Asymmetrical Transformer Tracker). It modifies the self-attention operation in an asymmetrical way, striving to enhance target-related features while suppressing noise. Extensive experiments confirm that our ATTracker serves as a robust baseline, outperforming fully supervised base trackers. By unveiling the potential of learning with limited annotations, this study aims to attract attention and pave the way for Semi-supervised Vision-Language Tracking (SS-VLT).

Response generation is the task of automatically generating human-like content based on the provided context. One of its prominent applications is to simulate realistic response content for social network posts. In the digital age, social network platforms play a vital role in information exchange and social interaction. This study focuses on response generation techniques for the platform of public opinion evolution simulation that simulate realistic response content, enabling a deeper understanding of the emotional expressions of network users. Recent advancements in deep learning techniques, particularly the sequence-to-sequence (Seq2Seq) model, have shown promise in the response generation field. However, we still face two challenges: content variety, topic and emotion relevancy. To this end, we propose the EmoTG-ETRS model which comprises three parts. The first is a response generation module based on Transformer architecture. Then, an auxiliary emotion improvement module is incorporated to enhance the emotional expressiveness of the response candidates. Finally, a reverse selection module, which combines maximum mutual information (MMI) evaluation, emotional expression evaluation, and topic consistency evaluation, is devised to select the highest-scoring response. Extensive experiments have been conducted to evaluate the effectiveness of the proposed model and the results demonstrate that the EmoTG-ETRS model improves the quality of produced replies in terms of topic consistency and emotional accuracy rate when compared with the SOTA research works.

The influence and dissemination of users’ opinions are the essence of the opinion dynamics in online social networks (OSNs). Understanding the process of users’ opinion formation and propagation can provide better service for public opinion monitoring and product advertising. However, most opinion dynamics research typically do not distinguish between user opinion formation and propagation, instead focusing on the process of mass opinion polarization. This article presents a multiagent system (MAS) to analyze fine-grained opinion dynamics (FOD) by agent-based modeling (ABM). At the macro level, the MAS-FOD system we designed can produce a statistical analysis of public opinion evolution based on different parameter constellations and analyze the changes in personalized agent opinions. At the micro level, the agent-based model is presented to describe the single user in OSNs as an individualized agent in MAS-FOD. Specifically, we propose two mechanisms, agent-based opinion formation (AOF) mechanism and agent-based opinion propagation (AOP) mechanism, for agents to form and disseminate opinions, respectively. Meanwhile, multidimensional social influence features mining from self, local neighbors, and global topic community are defined and used in two mechanisms to train personalized agents to self-adapt to the MAS-FOD system. We demonstrate the rationality and effectiveness of the MAS-FOD system from two types of experiments: empirical analysis and simulation analysis. The former is driven by empirical social network data to analyze the predictive performance of AOF; the latter simulates the public opinion propagation through the AOP mechanism in the MAS-FOD system. The results demonstrate that: 1) the AOF outperforms SOTA methods in the accuracy of agent opinion prediction; 2) the dissemination effect based on the AOP is more closer to the actual evolution trend of public opinion than the baseline; and 3) the MAS-FOD system can perform fine-grained analysis of opinion dynamics by adjusting different parameters.

The community plays a crucial role in understanding user behavior and network characteristics in social networks. Some users can use multiple social networks at once for a variety of objectives. These users are called overlapping users who bridge different social networks. Detecting communities across multiple social networks is vital for interaction mining, information diffusion, and behavior migration analysis among networks. This article presents a community detection method based on nonnegative matrix trifactorization for multiple heterogeneous social networks, which formulates a common consensus matrix to represent the global fused community. Specifically, the proposed method involves creating adjacency matrices based on network structure and content similarity, followed by alignment matrices that distinguish overlapping users in different social networks. With the generated alignment matrices, the method could enhance the fusion degree of the global community by detecting overlapping user communities across networks. The effectiveness of the proposed method is evaluated with new metrics on Twitter, Instagram, and Tumblr datasets. The results of the experiments demonstrate its superior performance in terms of community quality and community fusion.

Micro-video venue recognition aims to predict the venue category where a micro-video was filmed. Different from traditional long videos which contain rich temporal context, venue prediction for micro-videos is difficult due to its limited duration (generally within 6s). The existing works usually extract features of each modality from a global perspective for prediction, neglecting the semantics carried by local objects. To this end, we propose Multi-Modal and Multi-Granularity Object Relations (M2ORE) to address the above issues, which learns multi-granularity interactive semantics between venues and multimodal semantic objects to help understand venues. Specifically, M2ORE comprises of two modules: it first extract semantic objects of different modalities, i.e. visual objects in keyframes and keywords in texts, and models the affiliation relationship between semantic objects and venues and the co-occurrence relationship among semantic objects, forming a heterogeneous venue-object relation graph. Then, to achieve the interactive semantics between venues and objects from the relation graph, a novel Parallel-Graph Inference Model (Parallel-GIM) is proposed, which updates the representation of nodes through graph propagation and fuse multi-level features (local-global-multimodal) through the devised hierarchical attention mechanism. Finally, the probability distribution of venues can be obtained through a multi-layer perceptron with the comprehensive features of the venue nodes. Extensive experiments on real-world micro-video dataset demonstrate the superiority of the proposed M2ORE.

Multi-label image classification is an essential yet challenging task that requires to recognize multiple objects of images. To this end, recent studies have sought to acquire visual representations for each label by attention models, and then train binary classifiers for prediction. However, these methods have two major drawbacks: 1) They rely heavily on the precise alignments between two modalities, which is still challenging for current attention models; 2) They ignore patch-level representations rich in local object features, which are also of great importance for label recognition. In this paper, we propose a semantic-guided representation enhancement framework, which augments patch-level representations with object-level representations for robust label recognition. Concretely, the proposed framework consists of two significant components: 1) an inter-modal attention module that accounts for coarsely locating object regions and producing object-level representations for each label; 2) an intra-modal attention module that aggregates object representations to enhance patch representations based on their correlations. In this way, both local clues and global glances of objects are fully exploited simultaneously, rather than relying solely on object-level representations obtained by the inter-modal attention, thus improving the performance of label recognition. Experimental results show that our framework outperforms the state-of-the-art methods by 0.5%, 0.6%, 0.7% and 0.8% in mAP on Pascal VOC 2007, Microsoft COCO, NUS-WIDE and Visual Genome datasets, respectively. Codes and models are available on https://github.com/jasonseu/SGRE.

Online rumor detection is crucial for a healthier online environment. Traditional methods mainly rely on content understanding. However, these contents can be easily adjusted to avoid such supervision and are insufficient to improve the detection result. Compared with the content, information propagation patterns are more informative to support further performance promotion. Unfortunately, learning the propagation patterns is difficult, since the retweeting tree is more topologically complicated than linear sequences or binary trees. In light of this, we propose a novel rumor detection framework based on structure-aware retweeting graph neural networks. To capture the propagation patterns, we first design a novel conversion method to transform the complex retweeting tree as more tractable binary tree without losing the reconstruction information. Then, we serialize the retweeting tree as a corpus of meta-tree paths, where each meta-tree can preserve a basic substructure. A deep neural network is then designed to integrate all meta-trees and to generate the global structural embeddings. Furthermore, we propose to integrate content, users, and propagation patterns to enhance more reliable performance. To this end, we propose a novel self-attention-based retweeting neural network to learn individual features from both content and users. We then fuse the node-level features with our global structural embeddings via a mutual attention unit. In this way, we can generate more comprehensive representations for rumor detection. Extensive evaluations on two real-world datasets show remarkable superiorities of our model compared with existing methods.

Event-based Social Network (EBSN) is a special online social network that not only provides users with a convenient virtual platform to communicate, but also helps them to participate in offline face-to-face social activities on a group basis. This paper studies the group popularity prediction problem in EBSN which is important for online advertising and event recommendation. Firstly, we construct the EBSN temporal network model with a timeline involved and propose a novel definition of group popularity in EBSN. Then on the basis of the above, we extract corresponding key features from four aspects: inherent group characteristics (i.e., founder, created time), historical popularity, the sentiment of users towards events as an internal factor, and the initiative of newly added users as an external factor. Combining these four features, we propose a group popularity prediction algorithm in EBSN based on an improved self-excited Hawkes Process: EBSN-Hawkes Process Algorithm (EHP). Experiments conducted by real EBSN datasets demonstrate that the proposed EHP algorithm has better prediction results than other comparative algorithms. We also conclude that the external dynamic feature plays a more important role than the internal sentiment feature by the ablation study.

Multi-label image classification refers to assigning a set of labels for an image. One of the main challenges of this task is how to effectively capture the correlation among labels. Existing studies on this issue mostly rely on the statistical label co-occurrence or semantic similarity of labels. However, an important fact is ignored that the co-occurrence of labels is closely related with image scenes (indoor, outdoor, etc.), which is a vital characteristic in multi-label image classification. In this paper, a novel scene-aware label graph learning framework is proposed, which is capable of learning visual representations for labels while fully perceiving their co-occurrence relationships under variable scenes. Specifically, our framework is able to detect scene categories of images without relying on manual annotations, and keeps track of the co-occurring labels by maintaining a global co-occurrence matrix for each scene category throughout the whole training phase. These scene-independent co-occurrence matrices are further employed to guide the interactions among label representations in a graph propagation manner towards accurate label prediction. Extensive experiments on public benchmarks demonstrate the superiority of our framework.

When user-centric location privacy protections are used by the public (e.g., published as a mobile application to Mac app store), they are found to be vulnerable to a novel kind of potential inference attacks. These attacks can utilize the open-use protection’s input–output pairs to closely study the protection’s structure and mechanism. We name this kind of attacks as knowing-and-learning (KL) attacks. Targeted at such potential but threatening deep learning-based KL attacks in local search services, we propose a novel location privacy protection framework LPP2KL. LPP2KL aims to generate an obfuscated location, which is robust to the aforementioned attacks, so that the user’s current location (real) can be replaced by the generated location (pseudo) when submitted to the untrusted search server. To achieve this, we first depict a feasible net structure to implement the deep KL attack and verify its power with extensive experiments. Second, to simulate the attack-and-defense process, we develop a novel network structure with an adversary net and a protection net. These nets play a mini–max game toward privacy inference in an adversarial manner until an equilibrium is reached. Meanwhile, the protection net also achieves the balance between preserved privacy and quality of service through utility-constrained privacy optimization. We introduce a novel penalty function to handle the utility-constrained privacy optimization problem. Extensive experiments are conducted on Yelp and FoursSwarm datasets, which validate the generality of our proposed LPP2KL in protection handling and illustrate its effectiveness in utility control under the acceptable quality of service.

Clickbait is a sort of news that contains low-quality content but intriguing or overstated titles. Most of the existing studies usually utilize titles to detect clickbait and suffer from poor performance. However, from the perspective of user cognition, the essence of clickbait news is due to the fact that the body contents fall short of the titles’ expectations. So, in addition to the explicit features of titles, many extra informative signals could help to distinguish clickbait. In this article, we proposed a novel Multiview learning approach-based ClickBait Detection (MCBD) model that takes into account useful signals from both titles and body contents. From the view of titles, a multilayer gated convolutional network is employed to learn the local- and long-distance dependence relationships between words. From the view of body contents, we first extract the key topic sentences from the body content and then we developed a novel attention-fused deep relevance matching network (ARMN) to thoroughly mine the similarity between titles and body contents. Combined with the information from the views of titles and body contents, our model can distinguish well-hidden clickbait news. Finally, extensive experiments were conducted on two real-world datasets to demonstrate that the MCBD model outperforms the state-of-the-art models, improving the final performance by 3% in terms of the F1-score. The ablation studies demonstrate that the relevancy between titles and body contents can improve the model’s performance by 1.7% on the F1-score.

With the dramatic growth of various short video platforms, users are more likely to share their social stream online and make their social connections stronger. To better understand their preferences, personality analysis has attracted more attention. Unlike single modal data such as text or images, which is hard to comprehensively uncover one’s personal traits, personality analysis on short video is verified to be much more accurate but also more challenging because of the huge gap between incompatible data modalities. We have noticed that the key problem is how to disentangle the complexity from multimodal data to find their consistency and uniqueness. In this article, we propose a novel video analysis framework for personality detection with visual, acoustic, and textual neural networks. Specifically, to enhance our model’s sensitivity to personality detection, we first propose three deep learning channels to learn modal features. The framework can not only extract each modal feature but also learn time-varying pattern via a temporal alignment network. To identify the consistency and uniqueness across multiple modalities, we creatively propose to maximize the similarity of common information learned by a shared neural network across multiple modalities and extend the distance of exclusive information learned by private networks of different modalities. Extensive experiments on the real-world dataset demonstrate that our model can outperform existing baselines.

Online social networks have greatly facilitated our lives but have also propagated the spreading of rumours. Traditional works mostly find rumors from content, but content can be strategically manipulated to evade such detection, making these methods brittle. To improve the accuracy and robustness of rumor detection, we propose to integrate and exploit the content, propagation structure, and temporal relations because information in the networks always spreads dynamically with significant structures. In this paper, we propose a novel rumor detection framework in online temporal networks via structure learning. Specifically, to exploit the propagation structure, we propose a novel hyperedge walking strategy on a meta-hyperedge graph to learn the representations of sub-structures in the networks. Then a hyperedge expansion method is proposed to generate more global structural features. The expanded hyperedges are more hierarchical, making the learned structural embeddings more expressive. To make full use of content, we design a hypergraph learning model using hyperedge expansion to fuse node content with structural features and generate comprehensive representations for the entire graph. To exploit temporal relations, we design a masked temporal attention unit for learning the evolving patterns of the network. Extensive evaluations with six state-of-the-art baselines on two real-world datasets demonstrate the superiority of our solution.

Visual entailment (VE) is to recognize whether the semantics of a hypothesis text can be inferred from the given premise image, which is one special task among recent emerged vision and language understanding tasks. Currently, most of the existing VE approaches are derived from the methods of visual question answering. They recognize visual entailment by quantifying the similarity between the hypothesis and premise in the content semantic features from multi modalities. Such approaches, however, ignore the VE's unique nature of relation inference between the premise and hypothesis. Therefore, in this paper, a new architecture called AlignVE is proposed to solve the visual entailment problem with a relation interaction method. It models the relation between the premise and hypothesis as an alignment matrix. Then it introduces a pooling operation to get feature vectors with a fixed size. Finally, it goes through the fully-connected layer and normalization layer to complete the classification. Experiments show that our alignment-based architecture reaches 72.45% accuracy on SNLI-VE dataset, outperforming previous content-based models under the same settings.

In the news recommendation, users are overwhelmed by thousands of news daily, which makes the users’ behavior data have high sparsity. Therefore, only considering a single user’s personalized preferences cannot support the news recommendation. How to improve the relatedness of news and users and reduce data sparsity has become a hot issue. Recent studies have attempted to use graph models to enrich the relationship between users and news, but they are still limited to modeling the historical behaviors of a single user. To fill the gap, we integrate user-news relationships and the overall user historical clicked news sequences to construct a global heterogeneous transition graph. And a refinement approach is proposed to recognize the news transition patterns in the graph. Based on the global heterogeneous transition graph, we propose a heterogeneous transition graph attention network to capture the common behavior patterns of most users to enhance the representation of user interest. Fusing the users’ personalized and common interest, we propose the GAINRec model to recommend news effectively. Extensive experiments are conducted on two public news recommendation datasets, and the results show the superiority of the proposed GAINRec model compared with the state-of-the-art news recommendation models. The implementation of our model is available at https://github.com/newsrec/GAINRec .

Commonsense question answering (QA) requires machines to utilize the QA content and external commonsense knowledge graph (KG) for reasoning when answering questions. Existing work uses two independent modules to model the QA contextual text representation and relationships between QA entities in KG, which prevents information sharing between modules for co-reasoning. In this paper, we propose a novel model, Co-Reasoning Network (CORN), which adopts a bidirectional multi-level connection structure based on Co-Attention Transformer. The structure builds bridges to connect each layer of the text encoder and graph encoder, which can introduce the QA entity relationship from KG to the text encoder and bring contextual text information to the graph encoder, so that these features can be deeply interactively fused to form comprehensive text and graph node representations. Meanwhile, we propose a QA-aware node based KG subgraph construction method. The QA-aware nodes aggregate the question entity nodes and the answer entity nodes, and further guide the expansion and construction process of the subgraph to enhance the connectivity and reduce the introduction of noise. We evaluate our model on QA benchmarks in the CommonsenseQA and OpenBookQA datasets, and CORN achieves state-of-the-art performance.

Multi-label image classification is a fundamental yet challenging task in computer vision that aims to identify multiple objects from a given image. Recent studies on this task mainly focus on learning cross-modal interactions between label semantics and high-level visual representations via an attention operation. However, these one-shot attention based approaches generally perform poorly in establishing accurate and robust alignments between vision and text due to the acknowledged semantic gap. In this paper, we propose a two-stream transformer (TSFormer) learning framework, in which the spatial stream focuses on extracting patch features with a global perception, while the semantic stream aims to learn vision-aware label semantics as well as their correlations via a multi-shot attention mechanism. Specifically, in each layer of TSFormer, a cross-modal attention module is developed to aggregate visual features from spatial stream into semantic stream and update label semantics via a residual connection. In this way, the semantic gap between two streams gradually narrows as the procedure progresses layer by layer, allowing the semantic stream to produce sophisticated visual representations for each label towards accurate label recognition. Extensive experiments on three visual benchmarks, including Pascal VOC 2007, Microsoft COCO and NUS-WIDE, consistently demonstrate that our proposed TSFormer achieves state-of-the-art performance on the multi-label image classification task.

In recent years, social networks have shown diversity in function and applications. People begin to use multiple online social networks simultaneously for different demands. The ability to uncover a user’s latent topic and social network preference is critical for community detection, recommendation, and personalized service across social networks. Unfortunately, most current works focus on the single network, necessitating new technology and models to address this issue. This paper proposes a user preference discovery model on multiple social networks. Firstly, the global and local topic concepts are defined, then a latent semantic topic discovery method is used to obtain global and local topic word distributions, along with user topic and social network preferences. After that, the topic distribution characteristics of different social networks are examined, as well as the reasons why users choose one network over another to create a post. Next, a Gibbs sampling algorithm is adopted to obtain the model parameters. In the experiment, we collect data from Twitter, Instagram, and Tumblr websites to build a dataset of multiple social networks. Finally, we compare our research to previous works, and both qualitative and quantitative evaluation results have demonstrated the effectiveness.

Recently, graph neural networks have been widely used for network embedding because of their prominent performance in pairwise relationship learning. In the real world, a more natural and common situation is the coexistence of pairwise relationships and complex non-pairwise relationships, which is, however, rarely studied. In light of this, we propose a graph neural network-based representation learning framework for heterogeneous hypergraphs, an extension of conventional graphs, which can well characterize multiple non-pairwise relations. Our framework first projects the heterogeneous hypergraph into a series of snapshots and then we take the Wavelet basis to perform localized hypergraph convolution. Since the Wavelet basis is usually much sparser than the Fourier basis, we develop an efficient polynomial approximation to the basis to replace the time-consuming Laplacian decomposition. Extensive evaluations have been conducted and the experimental results show the superiority of our method. In addition to the standard tasks of network embedding evaluation such as node classification, we also apply our method to the task of spammers detection and the superior performance of our framework shows that relationships beyond pairwise are also advantageous in the spammer detection. To make our experiment repeatable, source codes and related datasets are available at https://xiangguosun.mystrikingly.com

Social linking prediction is one of the most fundamental problems in online social networks and has attracted researchers’ persistent attention. Most of the existing works predict unobserved links using graph neural networks (GNNs) to learn node embeddings upon pair-wise relations. Despite promising results given enough observed links, these models are still challenging to achieve heart-stirring performance when observed links are extremely limited. The main reason is that they only focus on the smoothness of node representations on pair-wise relations. Unfortunately, this assumption may fall when the networks do not have enough observed links to support it. To this end, we go beyond pair-wise relations and propose a new and novel framework using hypergraph neural networks with multi-level hyperedge distillation strategies. To break through the limitations of sparsely observed links, we introduce the hypergraph to uncover higher-level relations, which is exceptionally crucial to deduce unobserved links. A hypergraph allows one edge to connect multiple nodes, making it easier to learn better higher-level relations for link prediction. To overcome the restrictions of manually designed hypergraphs, which is constant in most hypergraph researches, we propose a new method to learn high-quality hyperedges using three novel hyperedges distillation strategies automatically. The generated hyperedges are hierarchical and follow the power-law distribution, which can significantly improve the link prediction performance. To predict unobserved links, we present a novel hypergraph neural networks named HNN. HNN takes the multi-level hypergraphs as input and makes the node embeddings smooth on hyperedges instead of pair-wise links only. Extensive evaluations on four real-world datasets demonstrate our model’s superior performance over state-of-the-art baselines, especially when the observed links are extremely reduced.

According to social psychology studies, social networks are strongly subject to group influence; that is, users' behaviors or sentiment are primarily influenced by their group environments. But only a few studies to date have examined group influence in online social networks (OSNs). In this paper, a Factor Graph-based Group Influence model (F2GI) is proposed to depict the influence of perceived groups on a user. First, we analyze individuals' group environments and propose definitions and detecting methods of two types of groups: following groups and interacting groups. Secondly, based on these groups, we quantity the groups' features and propose the F2GI model able to describe group influence. Finally, using the retweet behaviors as the primarily manifestations of group influence, we apply our model to predict individuals' retweet behaviors and to validate how much influence the group puts on individuals. The results show that our group influence model can depict group influence more effectively and predict the actions of users who are affected by perceived groups accurately with social psychology ideas than classical machine learning methods. We believe that, moreover, our model can also be applied to marketing strategies, advertisement strategies and the prediction of public opinion.

Internet of Things (IoT) are increasingly common in our society, and can be found in civilian settings as well as sensitive applications, such as battlefields and national security. Given the potential of these devices to be targeted by attackers, they are a valuable source in digital forensic investigations. In addition, incriminating evidence may be stored on an IoT device (e.g., Amazon Echo in a home environment and Fitbit worn by the victim or an accused person). In comparison to IoT security and privacy literature, IoT forensics is relatively under-studied. IoT forensics is also challenging in practice, particularly due to the complexity, diversity, and heterogeneity of IoT devices and ecosystems. In this paper, we present an IoT-based forensic model that supports the identification, acquisition, analysis, and presentation of potential artifacts of forensic interest from IoT devices and the underpinning infrastructure. Specifically, we use the popular Amazon Echo as a use case to demonstrate how our proposed model can be used to guide forensics analysis of IoT devices.

With the development of the tourism, vast amount of tourist attraction information makes it complex and time-consuming for users to obtain satisfactory travel destination. Rich topological, temporal and spatial information in Location-Based Social Network (LBSN) helps to mine user preference deeply and evokes effectiveness of tourist attraction recommendation. In this paper, a personalized tourist attraction recommendation mechanism is designed in terms of different factors, including user preference, social relationship, location distance and location popularity. Firstly, a large number of check-ins in LBSN are filtered for further geographical space clustering to get real tourist attraction check-ins by DBSCAN (Density Based Spatial Clustering of Applications with Noise) algorithm. Then, a location-aware LDA model is used to excavate user's latent preference topic distribution and region's latent topic distribution on tourist attractions. After that, these two attributes together with social relationship are used to make user interest scoring function. Furthermore, another two scoring functions are established respectively based on tourist attraction location distance and popularity. At last, a personalized tourist attraction recommendation algorithm is designed to recommend tourist attractions to users, which ranks tourist attractions based on weighted average by combining the above three scores. Experiments are carried out on Foursquare data sets and results show that our method can perform well on recommendation in home city and new city.

Recently, personality detection based on texts from online social networks has attracted more and more attentions. However, most related models are based on letter, word or phrase, which is not sufficient to get good results. In this paper, we present our preliminary but interesting and useful research results to show that the structure of texts can be also an important feature in the study of personality detection from texts. We propose a model named 2CLSTM, which is a bidirectional LSTMs (Long Short Term Memory networks) concatenated with CNN (Convolutional Neural Network), to detect user's personality using structures of texts. Besides, a concept, Latent Sentence Group (LSG), is put forward to express the abstract feature combination based on closely connected sentences and we use our model to capture it. To the best of our knowledge, most related works only conducted their experiments on one data set, which may not well explain the versatility of their models. We implement our evaluations on two different kinds of datasets, containing long texts and short texts. Evaluations on both datasets have achieved better results, which demonstrate that our model can efficiently learn valid text structure features to accomplish the task.

In Location-Based Services (LBSs) platforms, such as Foursquare and Swarm, the submitted position for a share or search leads to the exposure of users’ activities. Additionally, the cross-platform account linkage could aggravate this exposure, as the fusion of users’ information can enhance inference attacks on users’ next submitted location. Hence, in this paper, we propose GLPP, a personalized and continuous location privacy-preserving framework in account linked platforms with different LBSs (i.e., search-based LBSs and share-based LBSs). The key point of GLPP is to obfuscate every location submitted in search-based LBSs so as to defend dynamic inference attacks. Specifically, first, possible inference attacks are listed through user behavioral analysis. Second, for each specific attack, an obfuscation model is proposed to minimize location privacy leakage under a given location distortion, which ensures submitted locations’ utility for search-based LBSs. Third, for dynamic attacks, a framework based on zero-sum game is adopted to joint specific obfuscation above and minimize the location privacy leakage to a balanced point. Experiments on real dataset prove the effectiveness of our proposed attacks in Accuracy, Certainty, and Correctness and, meanwhile, also show the performance of our preserving solution in defense of attacks and guarantee of location utility.

Crowdsourcing is often used to solicit contributions from an online community for ideas, evaluation and opinions. However, spamming can pollute such a system and manipulate the results of crowdsourcing. For detection of those spammers, the training data used in previous studies is often derived by experts labeling collected data and manually identifying spammers. The reliability of such training data is questionable. In this paper, we utilize two web based service providers Zhubajie (ZBJ) and Sandaha (SDH) and obtain reliable data about the spammers. We use such data to investigate the crowd-retweeting spam in Sina Weibo. We analyze profile features, social relationship and retweeting behavior of such spammers. We find that although these spammers are likely to connect more closely than legitimate users, the underlying social tie is different from the social relationship in other spam campaigns because of the unique retweeting features with the information cascade effect. Based on these findings, we propose retweeting-aware link based ranking algorithms to detect suspect spam accounts using seeds of identified spammers. Our evaluation shows that our algorithm is more effective than other link-based methods.

In this paper, we aim at the product promotion in O2O model and carry out the research of location-based influence maximization on the platform of LBSN. As offline consuming behavior exists under the O2O environment, the traditional online influence diffusion model could not describe the product acceptance accurately. Moreover, the existing researches of influence maximization tend to only concern on the online network of relationships but rarely take the offline part into consideration. This paper introduces the location property into the influence maximization to accord with the characteristic of O2O model. Firstly, we propose an improved influence diffusion model called TP Model which could accurately describe the process of accepting products under the O2O environment. Meanwhile, the definition of location-based influence maximization is presented. Then the user mobility pattern is analyzed and the calculation method of offline probability is designed. Considering the influence ability, a location-based influence maximization algorithm named TPH is proposed. Experiments prove TPH algorithm has general advantage. Finally, focusing on the performance of TPH algorithm under special circumstances, MR algorithm is designed as complement and experiments also verify its high effectiveness.

Summary Cloud data centers today usually lack network resource isolation. Meanwhile, it is easy to deploy and terminate large number of malicious virtual machines in a few seconds, while the administrator is probably difficult to identify these malicious virtual machines immediately. These features open doors for attackers to launch denial‐of‐service (DoS) attacks that target at degrading the quality of cloud service. This paper studies an attack scenario that malicious tenants use cloud resources to launch DoS attack targeting at data center subnets. Unlike traditional data flow‐based detections, which heavily depend on the pattern of data flows, we propose an approach that takes advantage of virtual machine status including CPU usage and network usage to identify the attack. We notice that malicious virtual machines exhibit similar status patterns when attack is launched. Based on this observation, information entropy is applied in monitoring the status of virtual machines to identify the attack behaviors. We conduct our experiments in the campus‐wide data center, and the results show our detection system can promptly and accurately response to DoS attacks. Copyright © 2015 John Wiley & Sons, Ltd.

In the composite service which runs for a long time under the heterogeneous and loose-coupled circumstance, the failure of service tends to occur. The transaction and recovery mechanism is urgently needed in order to guarantee the end-to-end QoS of the workflow and satisfy the user requirement. In this paper we address the composite service recovery issue in the way of substitution with the consideration of QoS constraint, based on our previous research work of the transactional construction and processing rules and the global optimization service selection algorithm, TSSA. Firstly, the service execution graph (SEG) is introduced and a service execution solution selection algorithm is proposed to choose one from the solution set of TSSA which has the highest success rate of recovery. Then, the concepts of execution backup path and switch cost are introduced and a search algorithm is presented to search for the optimal backup path when current service failure occurs. Meanwhile, a local induction algorithm based on positive feedback is described which could rapidly construct an transactional execution path when no backup execution path can be found in SEG and also guarantees the transactional constraint of composite service. Finally, experimental results show the recovery algorithm proposed in this paper is efficient and has high reliability.

As there are various risks of failure when Web Services are deployed in unreliable environment, the execution of a composite service requires the assurance of the transaction mechanism. However, existing QoS-aware composition approaches do not consider the transactional constraints during service selection. We address this issue by considering the combination of transactional and QoS requirements. Firstly, the novel construction and processing rules are proposed to guarantee the atomic consistency of the composite service and the correctness of these rules is proved subsequently. Then, on basis of these rules, an Ant Colony System based service selection algorithm is presented to guarantee the end-to-end QoS constraints on the premise of ensuring the atomic consistency during service selection. Therein, an optimization strategy is suggested to shrink the searching space of the algorithm tremendously. Finally, experimental results show the efficiency and effectiveness of the algorithm and demonstrate further the correctness of the construction and processing rules through simulations.

In order to integrate the massive distributed resources to accomplish the complex engineering applications cooperatively, workflow scheduling is an important aspect. However, as the available computing power of Grid resources is changing dynamically, static scheduling scheme will lead to low performance in real Grid environment. Therefore, the rescheduling mechanism should be taken into consideration. Although a few relevant mechanisms have been proposed in recent year, as they do not consider the essence of dynamic feature and the relevant algorithms are too simple, they can not obtain a good enough result yet. To address these problems, a load-aware based adaptive rescheduling mechanism for DAG application called LAR is proposed. Therein, during application running, the load exception will be detected and the execution state of application will be judged to decide whether the rescheduling process should be triggered. And in rescheduling stage, an effective rescheduling algorithm which utilizes the latest prediction information is present. The simulation results show that our mechanism can outperform the relevant algorithms in NRSL, and can effectively solve the performance decreasing problem in real Grid environment.

With the development of Web Service, it has become a key issue to select appropriate services from a large number of candidates for creating complex composite services according to users' different QoS levels requirements. However, the existing service selection algorithms have many defects such as high time complexity, non-global optimal solutions, and poor quality solutions. To solve these defects, an efficient multi-objective services selection algorithm, EMOSS, is proposed in this paper based on particle swarm optimization. The essence of EMOSS is to model the service selection problem as a constrained multi-objective optimization problem. First the services in each sub-service set are sorted by their concept of domination, then the new sub-service set nSi, whose size is far less than the original one, is constructed and finally output pare to optimal set. The theoretical analysis and experimental results show that EMOSS can effectively obtain high quality solutions.

Service-oriented computing is the new paradigm for distributed applications. A key issue of utilizing web services is to design a scalable service discovery mechanism. Current discovery method based on centralized registries can easily suffer from performance bottleneck in the service network of large scale. To address such problem, this paper presents a dynamic and scalable mechanism for discovery and registry of semantic web services. Registry proxies are introduced to achieve the data distribution of web service. Behavior of registry proxies are carefully designed to maintain the overall structure in a self-organization way. A distributed balance-aware algorithm is designed to improve the system performance. The result of simulation experiment shows that the proposed mechanism can serve as a scalable solution for semantic web service publication and discovery.

In recent years, as a popular technique to support CSCW, Grid computing is becoming more and more attractive. Hereinto, as the CPU load information can guide task scheduling process greatly, the long-term CPU load prediction becomes a very hot research field and has been widely studied. However, as the prediction errors will be accumulated gradually and meanwhile the relevant parameters' optimal values may change dynamically with the variance of load series, the previous prediction algorithms usually can not obtain good prediction accuracy when the length of prediction interval is quite large. To address these feature, a Data Aggregation based Adaptive Long term load Prediction mechanism called DA2LP is proposed in this paper. Therein, in order to reduce the number of prediction step and increase the amount of useful input load information, the data aggregation concept is introduced to integrate with AR model. Meanwhile, with the observation and analysis of the relevant parameters' impact on prediction accuracy in our prediction model, an adaptive parameter selection mechanism is proposed, where the optimal relevant parameters can be adapted automatically to enhance prediction accuracy during the prediction process. The experiments show that our proposed mechanism can outperform significantly the previous prediction methods in mean square error (MSE) for long term load prediction.

Service composition with Quality of Service (QoS) is widely studied nowadays and there are some approaches effective for the composition problem in ideal conditions but no feasible solution for dynamic and uncertain QoS constraints environment. Therefore, a novel service negotiation mechanism is proposed in this paper. The mechanism is composed by three parts for negotiation: model, protocol and strategy. The model creates a new hierarchy architecture based on negotiation agent to improve the efficiency of tasks execution of each agent. The protocol dynamically updates solution search line, on which proposals are made by trading partners, and approximate Pareto-optimal solution is achieved through the neutral mediator. The strategy enhances the negotiation sensibility by considering three factors: time, opponent actions and global negotiation states. The negotiation process is coordinated by Manger Agent (MA), which provides suggestions for the next negotiation round based on the overall negotiation context. Experimental results show that the proposed approach is effective to find a feasible solution in the dynamically changeable composition environment.

Service composition is the current research focus in the field of Service-Oriented Computing. However, the service discovery and selection mechanisms are static and not flexible in existing approaches on service composition, and the end-to-end QoS of a composite service can not also be ensured. In this paper, Multi-agent based QoS-aware Service Composition solution (MQSC) is presented, in which the concept of user satisfaction degree is introduced to depict the QoS of the service and the composite service execution mechanism based on Directed Acyclic Graph (DAG) is presented in detail. Through the collaboration between agents, MQSC not only can provide a mechanism for the dynamic service composition but also can ensure the end-to-end QoS of the composite service.

Abstract It is an effective approach to adopt software agent in pervasive computing (Per‐Com) environment. Meanwhile, task execution is the key problem which is affected directly by task representation method in Per‐Com environment. The task of agent in Per‐Com environment is dynamic and changeful, so the task representation should be flexible and agent‐oriented. However, most of the existing task representation methods cannot present task relationship clearly and lack unified format, which leads to a drawback that the corresponding task processing efficiency is not satisfying. In this paper, firstly, an agent‐based Per‐Com architecture is presented in which the agent is endowed with certain role. Secondly, through analyzing task relationship, an agent role based task relationship tree (TRT) model and an XML‐based agent task representation method are proposed. Thirdly, based on TRT, an agent task decomposing and online multi‐agent scheduling algorithm are presented to solve time‐optimal problems. Meanwhile, a load balancing strategy based on probability is proposed to increase agent utilization. Finally, a prototype of an application on traffic monitoring is presented. The results of evaluation indicate that the architecture is feasible and the simulation tests of time performance indicate that the proposed TRT‐based task processing methods have a better performance than others. Copyright © 2009 John Wiley & Sons, Ltd.

Abstract In order to effectively utilize massive heterogeneous resources and provide transparent computing capability to upper applications, task scheduling as the key issue of pervasive computing system becomes significantly important. Previous proposed priority and duplication based task scheduling algorithms, which can be applied in pervasive computing environment, usually have following limitations: critical path cannot be calculated accurately while neglecting the effect of resource availability in scheduling; in duplication based resource allocation stage, duplications without restriction would lead to some negative effects on final schedule length (SL). For the purpose of solving these problems, a novel task scheduling algorithm based on dynamic critical path (DCP) and effective duplication, called DCPED, is presented in this paper. In DCPED, a more accurate DCP calculation method which takes resource availability into account is introduced. Meanwhile an effective task duplication strategy is proposed to eliminate ineffective duplications and make an optimized schedule result by using space compression technique and dynamic critical path length (DCPL) based evaluation technique respectively. Finally, simulation results show that DCPED can outperform previous algorithms significantly in NSL and speedup rate metrics. Especially, it is very effective for utilizing computing resources and scheduling the fine‐grain and large‐scale workflow applications in pervasive computing system. Copyright © 2008 John Wiley & Sons, Ltd.

A fundamental problem that confronts QoS-aware service selection and composition is the efficient and timely QoS information obtainment. Current research on this problem usually involves query-based or monitoring-based methods. However, in a dynamic and volatile service oriented Computing (SOC) environment, these solutions suffer some or all of the limitations, such as cost, timeliness guarantee and salability. This paper presents Pat, a P2P based publish/subscribe system to disseminate new revised QoS information. Pat aims at reliable and efficient QoS information dissemination in large-scale SOC environments. It exploits specialized rendezvous points (RP) and a replicas mechanism to reduce the risk of subscriptions loss and consequently improve reliability. A reverse RP ring is designed to quicken subscription delivery and QoS information publication. In addition, an optimization mechanism for composite services is built into Pat, which helps to reduce notification traffic. Simulation results show that Pat is reliable, efficient and scalable.

A fundamental problem that confronts SOA-based CSCW applications is the efficient and timely QoS information obtainment of component services. However, this issue has largely been overlooked. This paper presents a P2P based publish/subscribe service to disseminate new revised QoS information reliably and efficiently. Specialized rendezvous points and a replica mechanism are introduced to reduce the risk of subscriptions loss and consequently improve reliability. A message buffering and packaged delivery mechanism helps to reduce notification traffic. A reverse Chord ring, called RP ring, is designed to quicken subscription delivery and QoS information publication. Node partitioning technique is suggested to tackle the load balancing issue. Simulation results show that the service is reliable, efficient and scalable.

In this paper, a super-media courseware system infrastructure based on the knowledge description is propose which features in separation of knowledge description from knowledge entity. Using agent technology, a personalized learning environment is developed with dynamical, real-time user information matching. The courseware system infrastructure is not only capable of balancing network load, but also proves self-adaptive to the users' circumstances. The simulation tests are run to demonstrate the advantage in performance.