Existing backdoor attacks on multivariate time series (MTS) forecasting enforce strict temporal and dimensional coupling between triggers and target patterns, requiring synchronous activation at fixed positions across variables. However, realistic scenarios often demand delayed and variable-specific activation. We identify this critical unmet need and propose TDBA, a temporally decoupled backdoor attack framework for MTS forecasting. By injecting triggers that encode the expected location of the target pattern, TDBA enables the activation of the target pattern at any positions within the forecasted data, with the activation position flexibly varying across different variable dimensions. TDBA introduces two core modules: (1) a position-guided trigger generation mechanism that leverages smoothed Gaussian priors to generate triggers that are position-related to the predefined target pattern; and (2) a position-aware optimization module that assigns soft weights based on trigger completeness, pattern coverage, and temporal offset, facilitating targeted and stealthy attack optimization. Extensive experiments on real-world datasets show that TDBA consistently outperforms existing baselines in effectiveness while maintaining good stealthiness. Ablation studies confirm the controllability and robustness of its design.

Continual instruction tuning aims to incrementally adapt large language models to new tasks without forgetting previously acquired knowledge. Existing approaches often struggle to balance plasticity and stability. Replay-based methods retrain on historical data, which raises privacy concerns. Architecture-based methods allocate task-specific components, resulting in significant parameter growth. To address this, we consider a structure-sharing strategy that enables parameter reuse across similar tasks and expands only when necessary, avoiding any data replay. Specifically, we introduce Grow-on-Demand (GoD-MoE), a parameter-efficient framework that is based on sparse and adaptive expert module expansion for continual instruction tuning. GoD-MoE inserts multiple LoRA-based experts into attention layers and dynamically activates a small subset of experts for each task. To avoid redundant parameter growth, we develop an Expert Demand Detector that determines whether new experts are added, facilitating adaptive structural sharing and minimizing parameter overhead. We conduct comprehensive experiments on the TRACE benchmark, demonstrating that GoD-MoE achieves state-of-the-art performance. Furthermore, it effectively mitigates catastrophic forgetting and even outperforms several advanced replay-based baselines.

Temporal reasoning remains a critical challenge for large language models (LLMs), particularly when it requires encompassing relational dependencies and numerical constraints.Yet, existing benchmarks largely overlook the joint consideration of these two dimensions and primarily rely on single-task evaluation paradigms, making it difficult to assess whether correct answers reflect grounded reasoning or arise from superficial statistical recall.To address these gaps, we introduce TNR, a benchmark designed to evaluate both Temporal Numerical and Relational reasoning.We propose a bi-directional evaluation framework consisting of forward generation via Question Answering (QA) and backward verification via Fact Verification (FV).By measuring the alignment between QA and FV, we introduce a Consistency Rate to quantify the robustness of reasoning across these two directions.Experiments on a range of LLMs reveal notable discrepancies between QA and FV performance, particularly in numerical and interval-based tasks.Moreover, our bi-directional error analysis demonstrates that these inconsistencies often stem from heuristic shortcuts and statistical co-occurrences rather than grounded logical deduction, flaws that are frequently masked in standard single-task evaluations.

Temporal Knowledge Graph Question Answering (TKGQA) aims to answer natural language questions based on quadruple facts stored in Temporal Knowledge Graphs (TKGs). Recent studies have integrated Large Language Models (LLMs) to handle the complex semantic reasoning required by temporal questions. They typically linearize TKG quadruples into plain text, reducing TKGQA to a top-n text retrieval and context-based question answering problem. However, this textualization process destroys the original quadruple structure and entangles structural and temporal information with text semantics. Moreover, selecting top-n facts based on semantic similarity inevitably introduces a large amount of irrelevant noise into the LLM input, which degrades reasoning performance. To address these limitations, we propose SSR, a Structured Subgraph Retrieval framework for TKGQA with LLMs. In SSR, a Temporal Question Parser is first introduced to extract structured subgraph patterns and temporal constraints from the input questions, leveraging background context obtained via text retrieval. The Subgraph Retrieval module is then applied to directly filter a relevant subgraph from the TKG that contains the facts required to answer the question. The retrieved subgraph is further temporally compressed and subsequently incorporated into the LLM context for final question answering. Experimental results on two benchmark datasets demonstrate that SSR consistently outperforms strong baselines by a clear margin, achieving state-of-the-art performance. Our code is available at https://github.com/zhangli-coding/SSR.

Multimodal information extraction (MIE) constitutes a set of essential tasks aimed at extracting structural information from Web texts with integrating images, to facilitate the structural construction of Web-based semantic knowledge. To address the expanding category set including newly emerging entity types or relations on websites, prior research proposed the zero-shot MIE (ZS-MIE) task which aims to extract unseen structural knowledge with textual and visual modalities. However, the ZS-MIE models are limited to recognizing the samples that fall within the unseen category set, and they struggle to deal with real-world scenarios that encompass both seen and unseen categories. The shortcomings of existing methods can be ascribed to two main aspects. On one hand, these methods construct representations of samples and categories within Euclidean space, failing to capture the hierarchical semantic relationships between the two modalities within a sample and their corresponding category prototypes. On the other hand, there is a notable gap in the distribution of semantic similarity between seen and unseen category sets, which impacts the generative capability of the ZS-MIE models. To overcome the above disadvantages, we delve into the generalized zero-shot MIE (GZS-MIE) task and propose the hyperbolic multimodal generative representation learning framework (HMGRL). The variational information bottleneck and autoencoder networks are reconstructed with hyperbolic space for modeling the multi-level hierarchical semantic correlations among samples and prototypes. Furthermore, the proposed model is trained with the unseen samples generated by the decoder, and we introduce the semantic similarity distribution alignment loss to enhance the model's generalization performance. Experimental evaluations on two benchmark datasets underscore the superiority of HMGRL compared to existing baseline methods.

Spatial-temporal data collected across different geographic locations often suffer from missing values, posing challenges to data analysis. Existing methods primarily leverage fixed spatial graphs to impute missing values, which implicitly assume that the spatial relationship is roughly the same for all features across different locations. However, they may overlook the different spatial relationships of diverse features recorded by sensors in different locations. To address this, we introduce the multi-scale Graph Structure Learning framework for spatial-temporal Imputation (GSLI) that dynamically adapts to the heterogeneous spatial correlations. Our framework encompasses node-scale graph structure learning to cater to the distinct global spatial correlations of different features, and feature-scale graph structure learning to unveil common spatial correlation across features within all stations. Integrated with prominence modeling, our framework emphasizes nodes and features with greater significance in the imputation process. Furthermore, GSLI incorporates cross-feature and cross-temporal representation learning to capture spatial-temporal dependencies. Evaluated on six real incomplete spatial-temporal datasets, GSLI showcases the improvement in data imputation and downstream applications.

Missing values often occur in multivariate time series, affecting data analysis and applications. Existing studies typically use complete data to train imputation models, which are then used to fill missing values. However, in practice, missing values could appear in various cells. Such varieties unfortunately prevent imputation models performing, even making fillings unavailable without the convergence guarantee, i.e., lacking the ensurance of obtaining the optimal solution when the iteration tends to infinite. The reasons are that (1) the imputed values of multiple cells could affect each other towards the conformance to models, and (2) dependencies obtained from complete data may not be accurate enough to impute many unobserved values, which poses a tougher challenge of the convergence. In this work, we study the collaborative imputation with the convergence guarantee. By “collaborative”, we mean (1) all the missing cells can be collaboratively imputed with the guaranteed conformance to models, and (2) the imputation models are collaboratively optimized according to fillings as well. Our major technical highlights include 1) introducing the statistically explainable collaborative imputation via likelihood maximization, 2) designing a collaborative imputation algorithm for multiple missing cells and extending it into a parallel version equivalently, 3) improving the algorithm by both imputation values and models collaboratively optimized with the convergence guarantee in parallel, 4) designing the streaming imputation and adaptive parameter determination strategies. Experiments on real incomplete datasets demonstrate the superiority of our methods against twelve baselines, in both imputation accuracy and downstream applications.

The multimodal knowledge graph reasoning (MKGR) task aims to predict the missing facts in the incomplete MKGs by leveraging auxiliary images and descriptions of entities. Existing approaches are trained with single-target objectives, which neglect the probabilistic correlations of entity labels, especially in non-target entities. Moreover, previous studies incorporate all modalities statically or adaptively, overlooking the negative impacts of irrelevant or misleading information in the incompetent modalities. To address these issues, we introduce a novel Reinforced Multimodal Distillation framework, exploiting the Dark Side of Modalities (DSoM) from two perspectives: (1) Dark knowledge from non-target entities: We propose to train a unimodal KGR model through logit distillation to mimic the multimodal soft labels provided by pre-trained multimodal teacher models. The multimodal soft labels could provide rich supervision signals with subtle correlations among both target and non-target entities from multiple perspectives. We further decouple logits into neighbor entities and non-neighbor entities to divide into two types of correlations. (2) Dark side in unhelpful modalities: To exclude the adverse effects of unhelpful modalities, we introduce a reinforced teacher combination mechanism that dynamically selects the optimal set of multimodal teachers for each triple. The agent is trained to maximize the rewards, which are only assigned to the beneficial multimodal combination strategies for the student model. Comprehensive experiments demonstrate the effectiveness of DSoM framework on 5 MKGR datasets. Codes are available at github.com/OreOZhao/DSoM.

End-to-end automated fact-checking (AFC) aims to assess the truthfulness of claims using retrieved evidence. Some researchers use crawlers or search APIs to retrieve evidence from the web for veracity classification. However, existing methods indiscriminately rely on the retrieved evidence and overlook that the retrieved results are not always reliable. This unilateral reliance on evidence significantly hampers the performance of fact-checking. In this paper, we account for the diverse reliability levels of retrieved evidence and eliminate the negative impact from the causal perspective. To achieve our goal, we propose a novel Causal intervention and Counterfactual reasoning based Multi-Checker framework (CCMC), which introduces two additional counterfactual fact-checkers to verify claims from the counterfactual perspective. Specifically, we construct two distinct types of counterfactual instances via causal intervention to imitate the situation where the evidence is partially reliable or totally unreliable. Correspondingly, two counterfactual fact-checkers are trained with tailored counterfactual instances by counterfactual reasoning. During inference, the two counterfactual fact-checkers are employed to estimate and eliminate the potential impact of unreliable evidence. Extensive experiments on two real-world datasets demonstrate the superiority of our approach for improving end-to-end AFC. Especially, we surpass existing methods by 3.70% and 5.55% under gold and system evidence on the MOCHEG benchmark, respectively. Our code is available at https://github.com/BeiyuXuboL/CCMC.

Multimodal knowledge graphs inevitably contain numerous errors due to the lack of human supervision in their automated construction and updating processes. These errors can significantly degrade the performance of downstream applications that rely on them. Existing researches on knowledge graph error detection primarily focus on leveraging graph structural and textual information to identify triplet errors in unimodal knowledge graphs. However, unlike unimodal knowledge graphs, multimodal knowledge graphs also suffer from mismatches between images and their corresponding entities, referred to as modality errors. These modality errors not only hinder the performance of downstream applications but also impede our effective utilization of the abundant complementary information provided by the visual modality for detecting triplet errors. To this end, we introduce a novel task of multimodal knowledge graph error detection (MKGED) in this paper, aiming at simultaneously identifying both modality errors and triplet errors. Given the lack of datasets for evaluating this task, we first establish two comprehensive MKGED datasets. Furthermore, we propose a novel framework, KGDMC, to address the MKGED task. Within KGDMC, we devise a disentanglement modality reconstruction (DMR) module for modality error detection. This module disentangles each original modality representation into two disjoint components: modality-specific representations and modality-invariant representations, leveraging the cross-modality reconstruction process to detect mismatched visual modalities. Additionally, for the triplet error detection, we propose a multi-grained triplet confidence (MTC) module, incorporating local triplet confidence, global structure confidence, and global path confidence, to collaboratively detect mismatched triplets. Extensive experiments on our constructed two datasets demonstrate the superiority of our proposed framework.

Multimodal information extraction on social media is a series of fundamental tasks to construct the multimodal knowledge graph. The tasks aim to extract the structural information in free texts with the incorporate images, including: multimodal named entity typing and multimodal relation extraction. However, the growing number of multimodal data implies a growing category set and the newly emerged entity types or relations should be recognized without additional training. To address the aforementioned challenges, we focus on the zero-shot multimodal information extraction tasks which require using textual and visual modalities for recognizing unseen categories. Compared with text-based zero-shot information extraction models, the existing multimodal ones make the textual and visual modalities aligned directly and exploit various fusion strategies to improve their performances. But the existing methods ignore the fine-grained semantic correlation of text-image pairs and samples. Therefore, we propose the multimodal graph-based variational mixture of experts network (MG-VMoE) which takes the MoE network as the backbone and exploits it for aligning multimodal representations in a fine-grained way. Considering to learn informative representations of multimodal data, we design each expert network as a variational information bottleneck to process two modalities in a uni-backbone. Moreover, we also propose the multimodal graph-based virtual adversarial training to learn the semantic correlation between the samples. The experimental results on the two benchmark datasets demonstrate the superiority of MG-VMoE over the baselines.

Next Point-of-Interests (POIs) recommendation task aims to provide a dynamic ranking of POIs based on users' current check-in trajectories. The recommendation performance of this task is contingent upon a comprehensive understanding of users' personalized behavioral patterns through Location-based Social Networks (LBSNs) data. While prior studies have adeptly captured sequential patterns and transitional relationships within users' check-in trajectories, a noticeable gap persists in devising a mechanism for discerning specialized behavioral patterns during distinct time slots, such as noon, afternoon, or evening. In this paper, we introduce an innovative data structure termed the ``Mobility Tree'', tailored for hierarchically describing users' check-in records. The Mobility Tree encompasses multi-granularity time slot nodes to learn user preferences across varying temporal periods. Meanwhile, we propose the Mobility Tree Network (MTNet), a multitask framework for personalized preference learning based on Mobility Trees. We develop a four-step node interaction operation to propagate feature information from the leaf nodes to the root node. Additionally, we adopt a multitask training strategy to push the model towards learning a robust representation. The comprehensive experimental results demonstrate the superiority of MTNet over eleven state-of-the-art next POI recommendation models across three real-world LBSN datasets, substantiating the efficacy of time slot preference learning facilitated by Mobility Tree.

Knowledge-based Visual Question Generation aims to generate visual questions with outside knowledge other than the image. Existing approaches are answer-aware, which incorporate answers into the question-generation process. However, these methods just focus on leveraging the semantics of inputs to propose questions, ignoring the logical coherence among generated questions (Q), images (V), answers (A), and corresponding acquired outside knowledge (K). It results in generating many non-expected questions with low quality, lacking insight and diversity, and some of them are even without any corresponding answer. To address this issue, we inject logical verification into the processes of knowledge acquisition and question generation, which is defined as LVˆ2-Net. Through checking the logical structure among V, A, K, ground-truth and generated Q twice in the whole KB-VQG procedure, LVˆ2-Net can propose diverse and insightful knowledge-based visual questions. And experimental results on two commonly used datasets demonstrate the superiority of LVˆ2-Net. Our code will be released to the public soon.

Temporal knowledge graph forecasting aims to reason over known facts to complete the missing links in the future. Existing methods are highly dependent on the structures of temporal knowledge graphs and commonly utilize recurrent or graph neural networks for forecasting. However, entities that are infrequently observed or have not been seen recently face challenges in learning effective knowledge representations due to insufficient structural contexts. To address the above disadvantages, in this paper, we propose a Contrastive Prompt-based framework with Entity background information for TKG forecasting, which we named CoPET. Specifically, to bring the time-invariant entity background information to time-variant structural information, we employ a dual encoder architecture consisting of a candidate encoder and a query encoder. A contrastive learning framework is used to encourage the query representation to be closer to the candidate representation. We further propose three kinds of trainable time-variant prompts aimed at capturing temporal structural information. Experiments on two datasets demonstrate that our method is effective and stays competitive in inference with limited structural information. Our code is available at https://github.com/qianxinying/CoPET.

Multimodal information extraction (MIE) is a challenging task which aims to extract the structural information in free text coupled with the image for constructing the multimodal knowledge graph. The entity-based MIE tasks are based on the entity information to complete the specific tasks. However, the existing methods only investigated the entity-based MIE tasks under supervised learning with adequate labeled data. In the real-world scenario, collecting enough data and annotating the entity-based samples are time-consuming, and impractical. Therefore, we propose to investigate the entity-based MIE tasks under the low-resource settings. The conventional models are prone to overfitting on limited labeled data, which can result in poor performance. This is because the models tend to learn the bias existing in the limited samples, which can lead them to model the spurious correlations between multimodal features and task labels. To provide a more comprehensive understanding of the bias inherent in multimodal features of MIE samples, we decompose the features into image, entity, and context factors. Furthermore, we investigate the causal relationships between these factors and model performance, leveraging the structural causal model to delve into the correlations between the input features and output labels. Based on this, we propose the multimodal counterfactual instance learning framework to generate the counterfactual instances by the interventions on the limited observational samples. In the framework, we analyze the causal effect of the counterfactual instances and exploit it as a supervisory signal to maximize the effect for reducing the bias and improving the generalization of the model. Empirically, we evaluate the proposed method on the two public MIE benchmark datasets and the experimental results verify the effectiveness of it.

Temporal Knowledge Graph Question Answering (TKGQA) aims to answer temporal questions using knowledge in Temporal Knowledge Graphs (TKGs).Previous works employ pre-trained TKG embeddings or graph neural networks to incorporate the knowledge of TKGs.However, these methods fail to fully understand the complex semantic information of time constraints.In contrast, Large Language Models (LLMs) have shown exceptional performance in knowledge graph reasoning, unifying both semantic understanding and structural reasoning.To further enhance LLMs' temporal reasoning ability, this paper aims to integrate temporal knowledge from TKGs into LLMs through a Time-aware Retrieve-Rewrite-Retrieve-Rerank framework, which we named TimeR 4 .Specifically, to reduce temporal hallucination in LLMs, we propose a retrieve-rewrite module to rewrite questions using background knowledge stored in the TKGs, thereby acquiring explicit time constraints.Then, we implement a retrievererank module aimed at retrieving semantically and temporally relevant facts from the TKGs and reranking according to the temporal constraints.To achieve this, we fine-tune a retriever using the contrastive time-aware learning framework.Our approach achieves great improvements, with relative gains of 47.8% and 22.5% on two datasets, underscoring its effectiveness in boosting the temporal reasoning abilities of LLMs.Our code is available at https://github.com/qianxinying/TimeR4 .

Referring video object segmentation (RVOS) is a cross-modal task that aims to segment the target object described by language expressions. A video typically consists of multiple frames and existing works conduct segmentation at either the clip-level or the frame-level. Clip-level methods process a clip at once and segment in parallel, lacking explicit inter-frame interactions. In contrast, frame-level methods facilitate direct interactions between frames by processing videos frame by frame, but they are prone to error accumulation. In this paper, we propose a novel tracking-forced framework, introducing high-quality tracking information and forcing the model to achieve accurate segmentation. Concretely, we utilize the ground-truth segmentation of previous frames as accurate inter-frame interactions, providing high-quality tracking references for segmentation in the next frame. This decouples the current input from the previous output, which enables our model to concentrate on accurately segmenting just based on given tracking information, improving training efficiency and preventing error accumulation. For the inference stage without ground-truth masks, we carefully select the beginning frame to construct tracking information, aiming to ensure accurate tracking-based frame-by-frame object segmentation. With these designs, our tracking-forced method significantly outperforms existing methods on 4 widely used benchmarks by at least 3%. Especially, our method achieves 88.3% [email protected] accuracy and 87.6 overall IoU score on the JHMDB-Sentences dataset, surpassing previous best methods by 5.0% and 8.0, respectively.

A large number of studies have emerged for Multimodal Knowledge Graph Completion (MKGC) to predict the missing links in MKGs. However, fewer studies have been proposed to study the inductive MKGC (IMKGC) involving emerging entities unseen during training. Existing inductive approaches focus on learning textual entity representations, which neglect rich semantic information in visual modality. Moreover, they focus on aggregating structural neighbors from existing KGs, which of emerging entities are usually limited. However, the semantic neighbors are decoupled from the topology linkage and usually imply the true target entity. In this paper, we propose the IMKGC task and a semantic neighbor retrieval-enhanced IMKGC framework CMR, where the contrast brings the helpful semantic neighbors close, and then the memorize supports semantic neighbor retrieval to enhance inference. Specifically, we first propose a unified cross-modal contrastive learning to simultaneously capture the textual-visual and textual-textual correlations of query-entity pairs in a unified representation space. The contrastive learning increases the similarity of positive query-entity pairs, therefore making the representations of helpful semantic neighbors close. Then, we explicitly memorize the knowledge representations to support the semantic neighbor retrieval. At test time, we retrieve the nearest semantic neighbors and interpolate them to the query-entity similarity distribution to augment the final prediction. Extensive experiments validate the effectiveness of CMR on three inductive MKGC datasets. Codes are available at https://github.com/OreOZhao/CMR.

Enterprises and organizations are faced with potential threats from insider employees that may lead to serious consequences. Previous studies on insider threat detection (ITD) mainly focus on detecting abnormal users or abnormal time periods (e.g., a week or a day). However, a user may have hundreds of thousands of activities in the log, and even within a day there may exist thousands of activities for a user, requiring a high investigation budget to verify abnormal users or activities given the detection results. On the other hand, existing works are mainly post-hoc methods rather than real-time detection, which can not report insider threats in time before they cause loss. In this paper, we conduct the first study towards real-time ITD at activity level, and present a fine-grained and efficient framework LAN. Specifically, LAN simultaneously learns the temporal dependencies within an activity sequence and the relationships between activities across sequences with graph structure learning. Moreover, to mitigate the data imbalance problem in ITD, we propose a novel hybrid prediction loss, which integrates self-supervision signals from normal activities and supervision signals from abnormal activities into a unified loss for anomaly detection. We evaluate the performance of LAN on two widely used datasets, i.e., CERT r4.2 and CERT r5.2. Extensive and comparative experiments demonstrate the superiority of LAN, outperforming 9 state-of-the-art baselines by at least 8.43% and 6.35% in AUC for real-time ITD on CERT r4.2 and r5.2, respectively. Moreover, LAN can be also applied to post-hoc ITD, surpassing 8 competitive baselines by at least 7.70% and 4.03% in AUC on two datasets. Finally, the ablation study, parameter analysis, and compatibility analysis evaluate the impact of each module and hyper-parameter in LAN. The source code can be obtained fromhttps://github.com/Li1Neo/LAN.

User mobility behaviors frequently exhibit a spatial clustering phenomenon, wherein points of interest (POIs) visited by the same user tend to be in close proximity. Consequently, leveraging geographical influences for user preference modeling remains a prevalent approach in POI recommendation tasks. However, existing studies often overlook users’ hidden geographical habits for the following reasons: (1) Geographical features are commonly approximated by manually partitioned regions or fixed distributions, inadequately capturing the nuanced spatial proximity among POIs. (2) POIs with high geographical correlations are not explicitly incorporated as feedback signals during the training process, resulting in a lack of spatial clustering pattern learning within users’ preference representations. This paper introduces GeoCo, aGeographicalCorrelation enhanced network for POI recommendation. First, we model POIs’ geographical features using fine-grained hierarchical sequences to capture multilevel spatial relations. Subsequently, we propose a pre-training network that employs the sentence similarity assessment technique to comprehend the semantics of geographical correlations. Second, we introduce a novel multi-objective training process that intuitively learns spatial clustering patterns through user mobility behaviors. Extensive experiments conducted on two location-based social network (LBSN) datasets, Gowalla and Foursquare, demonstrate the superiority of our proposed model over fourteen state-of-the-art baseline models in POI recommendation tasks. Compared with the baselines, GeoCo has achieved a performance improvement of at least 5$\%$in Rec@5 and HR@5 on both datasets. Furthermore, we verify the effectiveness of pre-trained location vectors and the multi-objective training process in enhancing the model's understanding of geographical correlations for user preference construction.

Automatically recording surgical procedures and generating surgical reports are crucial for alleviating surgeons' workload and enabling them to concentrate more on the operations. Despite some achievements, there still exist several issues for the previous works: 1) failure to model the interactive relationship between surgical instruments and tissue; and 2) neglect of fine-grained differences within different surgical images in the same surgery. To address these two issues, we propose an improved scene graph-guided Transformer, also named by SGT++, to generate more accurate surgical report, in which the complex interactions between surgical instruments and tissue are learnt from both explicit and implicit perspectives. Specifically, to facilitate the understanding of the surgical scene graph under a graph learning framework, a simple yet effective approach is proposed for homogenizing the input heterogeneous scene graph. For the homogeneous scene graph that contains explicit structured and fine-grained semantic relationships, we design an attention-induced graph transformer for node aggregation via an explicit relation-aware encoder. In addition, to characterize the implicit relationships about the instrument, tissue, and the interaction between them, the implicit relational attention is proposed to take full advantage of the prior knowledge from the interactional prototype memory. With the learnt explicit and implicit relation-aware representations, they are then coalesced to obtain the fused relation-aware representations contributing to generating reports. Some comprehensive experiments on two surgical datasets show that the proposed STG++ model achieves state-of-the-art results.

Biomedical entity linking is an essential task in biomedical text processing, which aims to map entity mentions in biomedical text to standard terms in a given knowledge base.However, this task is challenging due to the rarity of many biomedical entities in real-world scenarios, which leads to a lack of annotated data for them.Limited by understanding these unseen entities, traditional biomedical entity linking models suffer from multiple types of linking errors.In this paper, we propose a novel latent feature generation framework BioFEG to address these challenges.Specifically, our BioFEG leverages domain knowledge to train a generative adversarial network, which generates latent semantic features of corresponding mentions for unseen entities.Utilizing these features, we fine-tune our entity encoder to capture fine-grained coherence information of unseen entities and better understand them.This allows models to make linking decisions more accurately, particularly for ambiguous mentions involving rare entities.Extensive experiments on the two benchmark datasets demonstrate the superiority of our proposed method.

Federated learning allows a large number of resource-constrained clients to train a globally-shared model together without sharing local data. These clients usually have only a few classes (categories) of data for training, where the data distribution is non-iid (not independent identically distributed). In this article, we put forward the concept ofcategory privacyfor the first time to indicatewhich classes of data a client has, which is an important but ignored privacy goal in the federated learning with non-iid data. Although secure aggregation protocols are designed for federated learning to protect the input privacy of clients, we perform the first systematic study oncategory inference attackand demonstrate that these protocols cannot fully protect category privacy. We design a differential selection strategy and two de-noising approaches to achieve the attack goal successfully. In our evaluation, we apply the attack to non-iid federated learning settings with various datasets. On MNIST, CIFAR-10, AG_news, and DBPedia dataset, our attack achieves$>90\%$accuracy measured in F1-score in most cases. We further consider a possible detection method and propose two strategies to make the attack more inconspicuous.

Multimodal cross-domain sentiment analysis aims at transferring domain-invariant sentiment information across datasets to address the insufficiency of labeled data. Existing adaptation methods achieve well performance by remitting the discrepancies in characteristics of multiple modalities. However, the expressive styles of different datasets also contain domain-specific information, which hinders the adaptation performance. In this article, we propose a disentangled sentiment representation adversarial network (DiSRAN) to reduce the domain shift of expressive styles for multimodal cross-domain sentiment analysis. Specifically, we first align the multiple modalities and obtain the joint representation through a cross-modality attention layer. Then, we disentangle sentiment information from the multimodal joint representation that contains domain-specific expressive style by adversarial training. The obtained sentiment representation is domain-invariant, which can better facilitate the sentiment information transfer between different domains. Experimental results on two multimodal cross-domain sentiment analysis tasks demonstrate that the proposed method performs favorably against state-of-the-art approaches.

Entity linking, which aims at aligning ambiguous entity mentions to their referent entities in a knowledge base, plays a key role in multiple natural language processing tasks. Recently, zero-shot entity linking task has become a research hotspot, which links mentions to unseen entities to challenge the generalization ability. For this task, the training set and test set are from different domains, and thus entity linking models tend to be overfitting due to the tendency of memorizing the properties of entities that appear frequently in the training set. We argue that general ultra-fine-grained type information can help the linking models to learn contextual commonality and improve their generalization ability to tackle the overfitting problem. However, in the zero-shot entity linking setting, any type information is not available and entities are only identified by textual descriptions. Thus, we first extract the ultra-fine entity type information from the entity textual descriptions. Then, we propose a hierarchical multi-task model to improve the high-level zero-shot entity linking candidate generation task by utilizing the entity typing task as an auxiliary low-level task, which introduces extracted ultra-fine type information into the candidate generation task. Experimental results demonstrate the effectiveness of utilizing the ultra-fine entity type information and our proposed method achieves state-of-the-art performance.

Despite the great progress of Visual Question Answering (VQA), current VQA models heavily rely on the superficial correlation between the question type and its corresponding frequent answers (i.e., language priors) to make predictions, without really understanding the input. In this work, we define the training instances with the same question type but different answers as superficially similar instances, and attribute the language priors to the confusion of VQA model on such instances. To solve this problem, we propose a novel training framework that explicitly encourages the VQA model to distinguish between the superficially similar instances. Specifically, for each training instance, we first construct a set that contains its superficially similar counterparts. Then we exploit the proposed distinguishing module to increase the distance between the instance and its counterparts in the answer space. In this way, the VQA model is forced to further focus on the other parts of the input beyond the question type, which helps to overcome the language priors. Experimental results show that our method achieves the state-of-the-art performance on VQA-CP v2. Codes are available at Distinguishing-VQA.

Multimodal knowledge graph completion (MKGC) aims to predict missing entities in MKGs.Previous works usually share relation representation across modalities.This results in mutual interference between modalities during training, since for a pair of entities, the relation from one modality probably contradicts that from another modality.Furthermore, making a unified prediction based on the shared relation representation treats the input in different modalities equally, while their importance to the MKGC task should be different.In this paper, we propose MoSE, a Modality Split representation learning and Ensemble inference framework for MKGC.Specifically, in the training phase, we learn modality-split relation embeddings for each modality instead of a single modality-shared one, which alleviates the modality interference.Based on these embeddings, in the inference phase, we first make modality-split predictions and then exploit various ensemble methods to combine the predictions with different weights, which models the modality importance dynamically.Experimental results on three KG datasets show that MoSE outperforms state-ofthe-art MKGC methods.Codes are available at https://github.com/OreOZhao/MoSE4MKGC.

Multimodal named entity recognition (MNER) on social media is a challenging task which aims to extract named entities in free text and incorporate images to classify them into userdefined types.The existing semi-supervised named entity recognition methods focus on the text modal and are utilized to reduce labeling costs in traditional NER.However, the previous methods are not efficient for semisupervised MNER.Because the MNER task is defined to combine the text information with image one and needs to consider the mismatch between the posted text and image.To fuse the text and image features for MNER effectively under semi-supervised setting, we propose a novel span-based multimodal variational autoencoder (SMVAE) model for semisupervised MNER.The proposed method exploits modal-specific VAEs to model text and image latent features, and utilizes product-ofexperts to acquire multimodal features.In our approach, the implicit relations between labels and multimodal features are modeled by multimodal VAE.Thus, the useful information of unlabeled data can be exploited in our method under semi-supervised setting.Experimental results on two benchmark datasets demonstrate that our approach not only outperforms baselines under supervised setting, but also improves MNER performance with less labeled data than existing semi-supervised methods.

Medical entity normalization is a fundamental task in medical natural language processing and clinical applications. The task aims to map medical mentions to standard entities in a given knowledge base. In this paper, we focus on Chinese medical procedure entity normalization. This task brings an extra multi-implication challenge that a mention may link to multiple standard entities. To perform the task, we propose a novel deep neural multi-task learning framework to jointly model implication number prediction and entity normalization. Our model utilizes the multi-head attention mechanism to provide mutual benefits between the two tasks. Experimental results show that our method achieves comparable performance compared with the baseline methods.

The prompt-based learning paradigm has gained much research attention recently. It has achieved state-of-the-art performance on several NLP tasks, especially in the few-shot scenarios. While steering the downstream tasks, few works have been reported to investigate the security problems of the prompt-based models. In this paper, we conduct the first study on the vulnerability of the continuous prompt learning algorithm to backdoor attacks. We observe that the few-shot scenarios have posed a great challenge to backdoor attacks on the prompt-based models, limiting the usability of existing NLP backdoor methods. To address this challenge, we propose BadPrompt, a lightweight and task-adaptive algorithm, to backdoor attack continuous prompts. Specially, BadPrompt first generates candidate triggers which are indicative for predicting the targeted label and dissimilar to the samples of the non-targeted labels. Then, it automatically selects the most effective and invisible trigger for each sample with an adaptive trigger optimization algorithm. We evaluate the performance of BadPrompt on five datasets and two continuous prompt models. The results exhibit the abilities of BadPrompt to effectively attack continuous prompts while maintaining high performance on the clean test sets, outperforming the baseline models by a large margin. The source code of BadPrompt is publicly available.

The federated learning provides an effective solution to train collaborative models over a large scale of participated Industrial Internet of Things (IIoT) applications with the help of a global server, building an intelligent life. However, the federated learning is vulnerable to the backdoor attack from strong malicious participants. The backdoor attack is inconspicuous and may result in devastating consequences. To resist the attack on IIoT applications, we propose the federated backdoor filter defense that can identify backdoor inputs and restore the data to availability by theblur-label-flippingstrategy. We build multiple filters with eXplainable AI models on the server and send them to clients randomly, preventing advanced attackers from evading the defense. Our backdoor filters show significant backdoor recognition with the accuracy up to 99%. After the implementation of the blur-label-flipping strategy, victim's local model on suspicious backdoor samples can achieve the accuracy up to 88%.

Image captioning for low-resource languages has attracted much attention recently. Researchers propose to augment the low-resource caption dataset into (image, rich-resource language, and low-resource language) triplets and develop the dual attention mechanism to exploit the existence of triplets in training to improve the performance. However, datasets in triplet form are usually small due to their high collecting cost. On the other hand, there are already many large-scale datasets, which contain one pair from the triplet, such as caption datasets in the rich-resource language and translation datasets from the rich-resource language to the low-resource language. In this article, we revisit the caption-translation pipeline of the translation-based approach to utilize not only the triplet dataset but also large-scale paired datasets in training. The caption-translation pipeline is composed of two models, one caption model of the rich-resource language and one translation model from the rich-resource language to the low-resource language. Unfortunately, it is not trivial to fully benefit from incorporating both the triplet dataset and paired datasets into the pipeline, due to the gap between the training and testing phases and the instability in the training process. We propose to jointly optimize the two models of the pipeline in an end-to-end manner to bridge the training and testing gap, and introduce two auxiliary training objectives to stabilize the training process. Experimental results show that the proposed method improves significantly over the state-of-the-art methods.

Automated medical named entity recognition and normalization are fundamental for constructing knowledge graphs and building QA systems. When it comes to medical text, the annotation demands a foundation of expertise and professionalism. Existing methods utilize active learning to reduce costs in corpus annotation, as well as the multi-task learning strategy to model the correlations between different tasks. However, existing models do not take task-specific features for different tasks and diversity of query samples into account. To address these limitations, this paper proposes a multi-task adversarial active learning model for medical named entity recognition and normalization. In our model, the adversarial learning keeps the effectiveness of multi-task learning module and active learning module. The task discriminator eliminates the influence of irregular task-specific features. And the diversity discriminator exploits the heterogeneity between samples to meet the diversity constraint. The empirical results on two medical benchmarks demonstrate the effectiveness of our model against the existing methods.

Baohang Zhou, Xiangrui Cai, Ying Zhang, Xiaojie Yuan. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

A simultaneous understanding of questions and images is crucial in Visual Question Answering (VQA). While the existing models have achieved satisfactory performance by associating questions with key objects in images, the answers also contain rich information that can be used to describe the visual contents in images. In this paper, we propose a re-attention framework to utilize the information in answers for the VQA task. The framework first learns the initial attention weights for the objects by calculating the similarity of each word-object pair in the feature space. Then, the visual attention map is reconstructed by re-attending the objects in images based on the answer. Through keeping the initial visual attention map and the reconstructed one to be consistent, the learned visual attention map can be corrected by the answer information. Besides, we introduce a gate mechanism to automatically control the contribution of re-attention to model training based on the entropy of the learned initial visual attention maps. We conduct experiments on three benchmark datasets, and the results demonstrate the proposed model performs favorably against state-of-the-art methods.

The prevailing use of both images and text to express opinions on the web leads to the need for multimodal sentiment recognition. Some commonly used social media data containing short text and few images, such as tweets and product reviews, have been well studied. However, it is still challenging to predict the readers’ sentiment after reading online news articles, since news articles often have more complicated structures, e.g., longer text and more images. To address this problem, we propose a layout-driven multimodal attention network (LD-MAN) to recognize news sentiment in an end-to-end manner. Rather than modeling text and images individually, LD-MAN uses the layout of online news to align images with the corresponding text. Specifically, it exploits a set of distance-based coefficients to model the image locations and measure the contextual relationship between images and text. LD-MAN then learns the affective representations of the articles from the aligned text and images using a multimodal attention mechanism. Considering the lack of relevant datasets in this field, we collect two multimodal online news datasets, containing a total of 14,566 articles with 56,260 images and 251,202 words. Experimental results demonstrate that the proposed method performs favorably compared with state-of-the-art approaches. We will release all the codes, models and datasets to the community.

Visual Question Answering~(VQA) requires a simultaneous understanding of images and questions. Existing methods achieve well performance by focusing on both key objects in images and key words in questions. However, the answer also contains rich information which can help to better describe the image and generate more accurate attention maps. In this paper, to utilize the information in answer, we propose a re-attention framework for the VQA task. We first associate image and question by calculating the similarity of each object-word pairs in the feature space. Then, based on the answer, the learned model re-attends the corresponding visual objects in images and reconstructs the initial attention map to produce consistent results. Benefiting from the re-attention procedure, the question can be better understood, and the satisfactory answer is generated. Extensive experiments on the benchmark dataset demonstrate the proposed method performs favorably against the state-of-the-art approaches.

Cloud gaming has been very popular recently, but providing satisfactory gaming experiences to players at a modest cost is still challenging. Colocating several games onto one server could improve server utilization. To enable efficient colocations while providing Quality of Service (QoS) guarantees, a precise quantification of performance interference among colocated games is required. However, achieving such precise interference prediction is very challenging for games due to the complexity introduced by the contention on many shared resources across CPU and GPU. Moreover, the distinctive properties of cloud gaming require that the prediction model should be constructed beforehand and the prediction should be made instantaneously at request arrivals, which further increases the difficulty. The existing solutions are either not applicable or not effective due to many limitations. In this paper, we present GAugur, a novel methodology that enables highly accurate prediction of the performance interference among games arbitrarily colocated. By leveraging machine learning technologies, GAugur is able to capture the complex relationship between the interference and the contention features of colocated games. We evaluate GAugur through extensive experiments using a large number of real popular games. The results show that GAugur is able to identify whether a colocated game satisfies QoS requirement within an average error of 5%, and is able to quantify the performance degradation of a colocated game within an average error of 7.9%, which significantly outperforms the alternatives. Moreover, GAugur incurs an offline profiling cost linear to the number of games, and negligible overhead for online prediction. We apply GAugur to guiding efficient game colocations for cloud gaming. Experimental results show that GAugur is able to increase the resource utilization by 20% to 60%, and improve the overall performance by up to 15%, compared to the state-of-the-art solutions.

Improving the captioning performance on low-resource languages by leveraging English caption datasets has received increasing research interest in recent years. Existing works mainly fall into two categories: translation-based and alignment-based approaches. In this paper, we propose to combine the merits of both approaches in one unified architecture. Specifically, we use a pre-trained English caption model to generate high-quality English captions, and then take both the image and generated English captions to generate low-resource language captions. We improve the captioning performance by adding the cycle consistency constraint on the cycle of image regions, English words, and low-resource language words. Moreover, our architecture has a flexible design which enables it to benefit from large monolingual English caption datasets. Experimental results demonstrate that our approach outperforms the state-of-the-art methods on common evaluation metrics. The attention visualization also shows that the proposed approach really improves the fine-grained alignment between words and image regions.

The missing values, appear in most of multivariate time series, prevent advanced analysis of multivariate time series data. Existing imputation approaches try to deal with missing values by deletion, statistical imputation, machine learning based imputation and generative imputation. However, these methods are either incapable of dealing with temporal information or multi-stage. This paper proposes an end-to-end generative model E²GAN to impute missing values in multivariate time series. With the help of the discriminative loss and the squared error loss, E²GAN can impute the incomplete time series by the nearest generated complete time series at one stage. Experiments on multiple real-world datasets show that our model outperforms the baselines on the imputation accuracy and achieves state-of-the-art classification/regression results on the downstream applications. Additionally, our method also gains better time efficiency than multi-stage method on the training of neural networks.

Embeddings of medical concepts such as medication, procedure and diagnosis codes in Electronic Medical Records (EMRs) are central to healthcare analytics. Previous work on medical concept embedding takes medical concepts and EMRs as words and documents respectively. Nevertheless, such models miss out the temporal nature of EMR data. On the one hand, two consecutive medical concepts do not indicate they are temporally close, but the correlations between them can be revealed by the time gap. On the other hand, the temporal scopes of medical concepts often vary greatly (e.g., common cold and diabetes). In this paper, we propose to incorporate the temporal information to embed medical codes. Based on the Continuous Bag-of-Words model, we employ the attention mechanism to learn a ``soft'' time-aware context window for each medical concept. Experiments on public and proprietary datasets through clustering and nearest neighbour search tasks demonstrate the effectiveness of our model, showing that it outperforms five state-of-the-art baselines.

Emotion analysis of on-line user generated textual content is important for natural language processing and social media analytics tasks. Most of previous emotion analysis approaches focus on identifying users’ emotional states from text by classifying emotions into one of the finite categories, e.g., joy, surprise, anger and fear. However, there exists ambiguity characteristic for the emotion analysis, since a single sentence can evoke multiple emotions with different intensities. To address this problem, we introduce emotion distribution learning and propose a multi-task convolutional neural network for text emotion analysis. The end-to-end framework optimizes the distribution prediction and classification tasks simultaneously, which is able to learn robust representations for the distribution dataset with annotations of different voters. While most work adopt the majority voting scheme for the ground truth labeling, we also propose a lexiconbased strategy to generate distributions from a single label, which provides prior information for the emotion classification. Experiments conducted on five public text datasets (i.e., SemEval, Fairy Tales, ISEAR, TEC, CBET) demonstrate that our proposed method performs favorably against the state-of-the-art approaches.

Multivariate time series usually contain a large number of missing values, which hinders the application of advanced analysis methods on multivariate time series data. Conventional approaches to addressing the challenge of missing values, including mean/zero imputation, case deletion, and matrix factorization-based imputation, are all incapable of modeling the temporal dependencies and the nature of complex distribution in multivariate time series. In this paper, we treat the problem of missing value imputation as data generation. Inspired by the success of Generative Adversarial Networks (GAN) in image generation, we propose to learn the overall distribution of a multivariate time series dataset with GAN, which is further used to generate the missing values for each sample. Different from the image data, the time series data are usually incomplete due to the nature of data recording process. A modified Gate Recurrent Unit is employed in GAN to model the temporal irregularity of the incomplete time series. Experiments on two multivariate time series datasets show that the proposed model outperformed the baselines in terms of accuracy of imputation. Experimental results also showed that a simple model on the imputed data can achieve state-of-the-art results on the prediction tasks, demonstrating the benefits of our model in downstream applications.

Similarity search is an essential operation in many applications. Given a collection of set records and a query, the exact set similarity search aims at finding all the records that are similar to the query from the collection. Existing methods adopt a filter-and-verify framework, which make use of inverted indexes. However, as the complexity of verification is rather low for setbased similarity metrics, they always fail to make a good tradeoff between filter power and filter cost. In this paper, we proposed an efficient framework for exact set similarity search based on tree index structure. We defined a hash-based ordering to effectively import data into the index structure and then make optimizations to reduce the filter cost. To further improve the filter power, we proposed a dynamic algorithm to partition the dataset into several parts and propose a multiple-index framework. Experimental results on real-world datasets show that our method significantly outperform the state-of-the-art algorithms.

This paper studies an emotion classification problem, which aims to classify online news comments to one of fine-grained emotion categories, e.g. happy, sad, and angry, etc. Neural networks have been widely used and achieved great success in sentiment classification. However, there must be sufficient labeled comments available for training neural networks, which usually requires labor-intensive and time-consuming manual labeling. One of the most effective solutions is to apply transfer learning, which uses abundant labeled comments from a source news domain to help the classification for another target domain with limited amount of labeled data. Still, the comments from different domains can have very different word distributions, which makes it difficult to transfer knowledge from one domain to another. In this paper, we accomplish cross-domain emotion tagging based on an advanced neural network BLSTM (bidirectional long short-term memory) with "domain translation'', which can overcome the difference between domains. A weighted linear transformation is utilized to "translate'' knowledge from source to target domain. An extensive set of experimental results on four datasets from popular online news services demonstrates the effectiveness of our proposed models.

Sentiment analysis of such opinionated online texts as reviews and comments has received increasingly close attention, yet most of the work is intended to deal with the detection of authors’ emotion. In contrast, this article presents our study of the social emotion detection problem, the objective of which is to identify the evoked emotions of readers by online documents such as news articles. A novel Latent Discriminative Model (LDM) is proposed for this task. LDM works by introducing intermediate hidden variables to model the latent structure of input text corpora. To achieve this, it defines a joint distribution over emotions and latent variables, conditioned on the observed text documents. Moreover, we assume that social emotions are not independent but correlated with one another, and the dependency of them is capable of providing additional guidance to LDM in the training process. The inclusion of this emotional dependency into LDM gives rise to a new Emotional Dependency-based LDM (eLDM). We evaluate the proposed models through a series of empirical evaluations on two real-world corpora of news articles. Experimental results verify the effectiveness of LDM and eLDM in social emotion detection.

In many online news services, users often write comments towards news in subjective emotions such as sadness, happiness or anger. Knowing such emotions can help understand the preferences and perspectives of individual users, and therefore may facilitate online publishers to provide more relevant services to users. Although building emotion classifiers is a practical task, it highly depends on sufficient training data that is not easy to be collected directly and the manually labeling work of comments can be quite labor intensive. Also, online news has different domains, which makes the problem even harder as different word distributions of the domains require different classifiers with corresponding distinct training data.

This paper is concerned with the joint allocation of bid price and campaign budget in sponsored search. In this application, an advertiser can create a number of campaigns and set a budget for each of them. In a campaign, he/she can further create several ad groups with bid keywords and bid prices. Data analysis shows that many advertisers are dealing with a very large number of campaigns, bid keywords, and bid prices at the same time, which poses a great challenge to the optimality of their campaign management. As a result, the budgets of some campaigns might be too low to achieve the desired performance goals while those of some other campaigns might be wasted; the bid prices for some keywords may be too low to win competitive auctions while those of some other keywords may be unnecessarily high. In this paper, we propose a novel algorithm to automatically address this issue. In particular, we model the problem as a constrained optimization problem, which maximizes the expected advertiser revenue subject to the constraints of the total budget of the advertiser and the ranges of bid price change. By solving this optimization problem, we can obtain an optimal budget allocation plan as well as an optimal bid price setting. Our simulation results based on the sponsored search log of a commercial search engine have shown that by employing the proposed method, we can effectively improve the performances of the advertisers while at the same time we also see an increase in the revenue of the search engine. In addition, the results indicate that this method is robust to the second-order effects caused by the bid fluctuations from other advertisers.

With the rapid growth of online news services, users can actively respond to online news by making comments. Users often express subjective emotions in comments such as sadness, surprise and anger. Such emotions can help understand the preferences and perspectives of individual users, and therefore may facilitate online publishers to provide users with more relevant services. This paper tackles the task of predicting emotions for the comments of online news. To the best of our knowledge, this is the first research work for addressing the task. In particular, this paper proposes a novel Meta classification approach that exploits heterogeneous information sources such as the content of the comments and the emotion tags of news articles generated by users. The experiments on two datasets from online news services demonstrate the effectiveness of the proposed approach.

This paper proposes a new method for computing page importance, referred to as BrowseRank. The conventional approach to compute page importance is to exploit the link graph of the web and to build a model based on that graph. For instance, PageRank is such an algorithm, which employs a discrete-time Markov process as the model. Unfortunately, the link graph might be incomplete and inaccurate with respect to data for determining page importance, because links can be easily added and deleted by web content creators. In this paper, we propose computing page importance by using a 'user browsing graph' created from user behavior data. In this graph, vertices represent pages and directed edges represent transitions between pages in the users' web browsing history. Furthermore, the lengths of staying time spent on the pages by users are also included. The user browsing graph is more reliable than the link graph for inferring page importance. This paper further proposes using the continuous-time Markov process on the user browsing graph as a model and computing the stationary probability distribution of the process as page importance. An efficient algorithm for this computation has also been devised. In this way, we can leverage hundreds of millions of users' implicit voting on page importance. Experimental results show that BrowseRank indeed outperforms the baseline methods such as PageRank and TrustRank in several tasks.