Large Reasoning Models (LRMs) have recently demonstrated impressive performance across a range of reasoning tasks by generating intermediate thoughts. However, these models can suffer from overthinking—generating excessive tokens that contribute little to final accuracy while increasing inference cost. To mitigate this, we propose TIV (Thought Injection via Vectors), an innovative framework that compresses token-level reasoning into compact vectors without sacrificing performance. Rather than generating explicit thoughts, TIV injects learnable vectors into the post-attention hidden states of the final token across Transformer layers, enabling implicit and lightweight reasoning. We further introduce a two-stage reinforcement learning strategy: the first stage calibrates the model's reasoning distribution, and the second distills it into a vector-based policy optimized for both accuracy and brevity. Experiments on three reasoning benchmarks show that TIV preserves over 99% of the original accuracy while reducing output length by more than 65% on average, reaching up to 80% in some cases. Moreover, TIV consistently achieves superior trade-offs between accuracy and efficiency compared to existing methods, distinguishing itself as a state-of-the-art (SOTA) approach for efficient reasoning in LRMs.

Multimodal fake news detection plays a crucial role in combating online misinformation. The inherent domain diversity of news in the real world has driven the development of cross-domain detection methods. However, these detection methods either suffer from significant performance degradation due to semantic and deception pattern shifts between the training (source) and test (target) domains or heavily rely on annotated labels. To address the problems, we propose ADOSE, an active multi-source domain adaptation framework for multimodal fake news detection which actively annotates a small subset of target samples to improve detection performance. Specifically, for domain shifts, we design a multi-expert classifier network based on refined features to comprehensively capture and adapt to the semantic space and deception patterns of news across different domains. To maximize adaptation performance with limited annotation cost, we propose a least-disagree uncertainty selector equipped with a diversity calculator for selecting the most informative samples. The selector leverages the uncertainty of inconsistent predictions before and after perturbations by multiple classifiers as an indicator of unfamiliar samples. It further incorporates diversity scores derived from multi-view features to ensure the chosen samples achieve maximal coverage of target domain features. The extensive experiments on multiple datasets show that ADOSE outperforms existing domain adaptation methods by 2.45% ~ 9.1%, indicating the superiority of our model.

Retrieval-augment generation is a prevalent strategy to mitigate hallucinations of LLMs. The attributable RAG (RAGQ) generates quotes for its answers. The quotes indicate which input contexts support the RAG to derive the answers, enhancing the answer's verifiability and trustworthiness. However, existing RAGQs exhibit significant degradation when dealing with questions that require multi-hop reasoning and multi-modal understanding, suffering from over-citation, implicit entity identification failure, and poor generalization. In this paper, we propose a novel RAGQ framework, namely QDRAG. QDRAG breaks down the input question into atomic subquestions to identify the implicit entities. Then, the reranker prunes context distractors to eliminate the downstream over-citation. To facilitate query decomposition, we propose two zero-shot approaches: QD-C and QD-R, which guide the QD MLLM to decompose the question based on context knowledge and retrieval rewards, respectively. One interesting finding is that finetuning on the QD task shows better generalizability compared to directly finetuning on the downstream RAGQ task. Experiments on four multi-modal QA benchmarks demonstrate QDRAG's efficacy in grounding answers and generating faithful citations. The framework significantly outperforms all the baselines on both in-domain and out-of-domain tests, even surpassing Gemini-Pro.

Remote sensing imagery presents vast, inherently unstructured spatial data, necessitating sophisticated reasoning to interpret complex user intents and contextual relationships beyond simple recognition tasks. In this paper, we aim to construct an Earth observation workflow to handle complex queries by reasoning about spatial context and user intent. As a reasoning workflow, it should autonomously explore and construct its own inference paths, rather than being confined to predefined ground‑truth sequences. Ideally, its architecture ought to be unified yet generalized, possessing capabilities to perform diverse reasoning tasks through one model without requiring additional fine-tuning. Existing remote sensing approaches rely on supervised fine-tuning paradigms and task‑specific heads, limiting both autonomous reasoning and unified generalization. To this end, we propose RemoteReasoner, a unified workflow for geospatial reasoning. The design of RemoteReasoner integrates a multi-modal large language model (MLLM) for interpreting user instructions and localizing targets, together with task transformation strategies that enable multi-granularity tasks, including object-, region-, and pixel-level. In contrast to existing methods, our framework is trained with reinforcement learning (RL) to endow the MLLM sufficient reasoning autonomy. At the inference stage, our transformation strategies enable diverse task output formats without requiring task-specific decoders or further fine-tuning. Experiments demonstrated that RemoteReasoner achieves state-of-the-art performance across multi-granularity reasoning tasks. Furthermore, it retains the MLLM's inherent generalization capability, demonstrating robust performance on unseen tasks and categories.

Recommending event schedules is a key issue in Event-based Social Networks (EBSNs) in order to maintain user activity. An effective recommendation is required to maximize the user's preference, subjecting to both time and geographical constraints. Existing methods face an inherent trade-off among efficiency, effectiveness, and generalization, due to the NP-hard nature of the problem. This paper proposes the Chain-of-Scheduling (CoS) framework, which activates the event scheduling capability of Large Language Models (LLMs) through a guided, efficient scheduling process. CoS enhances LLM by formulating the schedule task into three atomic stages, i.e., exploration, verification and integration. Then we enable the LLMs to generate CoS autonomously via Knowledge Distillation (KD). Experimental results show that CoS achieves near-theoretical optimal effectiveness with high efficiency on three real-world datasets in a interpretable manner. Moreover, it demonstrates strong zero-shot learning ability on out-of-domain data.

Yidan Liang, Jia Zhu, Weijie Shi, Hanghui Guo, Yue Cui, Jiawei Shen, Guoqing Ma, Jingjiang Liu, Qingyu Niu, Yilin Wang, Shimin Di, Jiajie Xu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Jingjiang Liu, Jia Zhu, Hanghui Guo, Weijie Shi, Yue Cui, Xiaokang Jin, Yilin Wang, Qingyu Niu, Jiawei Shen, Guoqing Ma, Yidan Liang, Shimin Di, Jiajie Xu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

High-level automation is increasingly critical in AI, driven by rapid advances in large language models (LLMs) and AI agents. However, LLMs, despite their general reasoning power, struggle significantly in specialized, data-sensitive tasks such as designing Graph Neural Networks (GNNs). This difficulty arises from (1) the inherent knowledge gaps in modeling the intricate, varying relationships between graph properties and suitable architectures and (2) the external noise from misleading descriptive inputs, often resulting in generic or even misleading model suggestions. Achieving proficiency in designing data-aware models—defined as the meta-level capability to systematically accumulate, interpret, and apply data-specific design knowledge—remains challenging for existing automated approaches, due to their inefficient construction and application of meta-knowledge. To achieve meta-level proficiency, we propose DesiGNN, a knowledge-centered framework that systematically converts past model design experience into structured, fine-grained knowledge priors well-suited for meta-learning with LLMs. To account for the inherent variability and external noise, DesiGNN aligns empirical property filtering from extensive benchmarks with adaptive elicitation of literature insights via LLMs. By constructing a solid meta-knowledge between unseen graph understanding and known effective architecture patterns, DesiGNN can deliver top-5.77% initial model proposals for unseen datasets within seconds and achieve consistently superior performance with minimal search cost compared to baselines.

Dynamic Retrieval-augmented Generation (RAG) has shown great success in mitigating hallucinations in large language models (LLMs) during generation.However, existing dynamic RAG methods face significant limitations in two key aspects: 1) Lack of an effective mechanism to control retrieval triggers, and 2) Lack of effective scrutiny of retrieval content.To address these limitations, we propose an innovative dynamic RAG method, DioR (Adaptive Cognitive Detection and Contextual Retrieval Optimization), which consists of two main components: adaptive cognitive detection and contextual retrieval optimization, specifically designed to determine when retrieval is needed and what to retrieve for LLMs is useful.Experimental results demonstrate that DioR achieves superior performance on all tasks, demonstrating the effectiveness of our work.

Recently, Hyper-relational Knowledge Graphs (HKGs) have been proposed as an extension of traditional Knowledge Graphs (KGs) to better represent real-world facts with additional qualifiers. As a result, researchers have attempted to adapt classical Knowledge Graph Embedding (KGE) models for HKGs by designing extra qualifier processing modules. However, it remains unclear whether the superior performance of Hyper-relational KGE (HKGE) models arises from their base KGE model or the specially designed extension module. In this paper, we data-wise convert HKGs to KG format using decomposition methods and then evaluate several classical KGE models' performance on HKGs. Our results show that some KGE models achieve comparable performance to HKGE models. Upon further analysis, we find that the decomposition methods alter the original HKG topology and fail to fully preserve HKG information. Moreover, we observe that current HKGE models are either insufficient in capturing the graph's long-range dependency or struggle to integrate main-triple and qualifier information due to the information compression issue. To further justify our findings and provide a direction for HKGE research, we propose FormerGNN, which employs a qualifier integrator to preserve the original HKG topology, a GNN-based graph encoder to capture the graph's long-range dependencies, and an improved approach for integrating main-triple and qualifier information to mitigate compression issues. Our experimental results demonstrate that FormerGNN outperforms existing HKGE models.

In recent years, the design and transfer of neural network models have been widely studied due to their exceptional performance and capabilities. However, the complex nature of datasets and the vast architecture space pose significant challenges for both manual and automated algorithms in creating high-performance models. Inspired by researchers who design, train, and document the performance of various models across different datasets, this paper introduces a novel schema that transforms the benchmark data into a Knowledge Benchmark Graph (KBG), which primarily stores the facts in the form of performance(data, model). Constructing the KBG facilitates the structured storage of design knowledge, aiding subsequent model design and transfer. However, it is a non-trivial task to retrieve or design suitable neural networks based on the KBG, as real-world data are often off the records. To tackle this challenge, we propose transferring existing models stored in KBG by establishing correlations between unseen and previously seen datasets. Given that measuring dataset similarity is a complex and open-ended issue, we explore the potential for evaluating the correctness of the similarity function. Then, we further integrate the KBG with Large Language Models (LLMs), assisting LLMs to think and retrieve existing model knowledge in a manner akin to humans when designing or transferring models. We demonstrate our method specifically in the context of Graph Neural Network (GNN) architecture design, constructing a KBG (with 26,206 models, 211,669 performance records, and 2,540,064 facts) and validating the effectiveness of leveraging the KBG to promote GNN architecture design.

Node classification is a key task in temporal graph learning (TGL). Real-life temporal graphs often introduce new node classes over time, but existing TGL methods assume a fixed set of classes. This assumption brings limitations, as updating models with full data is costly, while focusing only on new classes results in forgetting old ones. Graph continual learning (GCL) methods mitigate forgetting using old-class subsets but fail to account for their evolution. We define this novel problem as temporal graph continual learning (TGCL), which focuses on efficiently maintaining up-to-date knowledge of old classes. To tackle TGCL, we propose a selective learning framework that substitutes the old-class data with its subsets, Learning Towards the Future (LTF). We derive an upper bound on the error caused by such replacement and transform it into objectives for selecting and learning subsets that minimize classification error while preserving the distribution of the full old-class data. Experiments on three real-world datasets show that LTF effectively addresses the TGCL challenge.

Single-cell RNA sequencing (scRNA-seq) technology has emerged as a powerful tool for studying complex biological systems and diseases with high resolution. However, generating sufficient high quality scRNA-seq samples for accurate downstream analyses and ensuring reproducibility remains a challenge. Previous methods often produce low-quality samples or a limited number of useful specific cell subpopulations. To address this issue, we propose a novel Efficient Single-Cell RNA-seq Flow Diffusion (named ESCFD) method based on the Diffusion Model, capable of synthesizing large-scale, high-quality scRNA-seq samples and generating either ''holistic'' or specific cellular subpopulations within a single framework. A pre-guidance mechanism is designed for synthesizing specific cellular subpopulations of high quality, and a post-guidance mechanism was designed to accelerate the sampling of scRNA-seq samples. The ESCFD can efficiently synthesize large-scale and high-quality scRNA-seq samples for various downstream tasks, surpassing previous work with faster sampling speed. Experimental results demonstrate that the ESCFD model achieves state-of-the-art performance in cell classification and cell distribution similarity when evaluated on three scRNA-seq benchmarks, and verified the effectiveness of accelerated sampling. Visualization experiments also reveal the ESCFD's ability to synthesize specific cellular subpopulations. This innovative approach has the potential to significantly advance our understanding of complex biological systems and accelerate the development of novel therapeutic interventions and personalized medicine strategies.

We aim to develop a robust yet flexible visual foundation model for Earth observation. It should possess strong capabilities in recognizing and localizing diverse visual targets while providing compatibility with various input-output interfaces required across different task scenarios. Current systems cannot meet these requirements, as they typically utilize task-specific architecture trained on narrow data domains with limited semantic coverage. Our study addresses these limitations from two aspects: data and modeling. We first introduce an automatic data engine that enjoys significantly better scalability compared to previous human annotation or rule-based approaches. It has enabled us to create the largest dataset of its kind to date, comprising 270K image-text-mask triplets covering an unprecedented range of diverse semantic categories and attribute specifications. Based on this data foundation, we further propose a task unification paradigm that centers around referring expression segmentation. It effectively handles a wide range of vision-centric perception tasks, including classification, detection, segmentation, grounding, etc, using a single model without any task-specific heads. Combining these innovations on data and modeling, we present RemoteSAM, a foundation model that establishes new SoTA on several earth observation perception benchmarks, outperforming other foundation models such as Falcon, GeoChat, and LHRS-Bot with significantly higher efficiency. Models and data are publicly available at https://github.com/1e12Leon/RemoteSAM.

The development of multi-modal Unmanned Aerial Vehicles (UAVs) environment perception systems is hindered by three critical gaps in existing datasets: (1) insufficient modalities and pixel misalignment, (2) noisy labels, and (3) limited task types. To address these gaps, we propose an automatic data construction approach and construct a multi-modal UAV-based environment perception dataset, UEMM-Air. Its synthetic nature ensures scalability, reproducibility, and rare-event coverage, making it suitable for large-scale model pre-training. Benefiting from our automated data collection and annotation pipeline, UEMM-Air encompasses 120k data pairs across 6 aligned modalities and supports 4 perception tasks, significantly exceeding existing datasets (max 60k data, 3 modalities, 2 tasks). Compared to existing synthetic datasets like SynDrone, UEMM-Air provides more accurate annotations by avoiding noisy labels from direct coordinate computation. Notably, models pre-trained on UEMM-Air achieve a 5.8% accuracy improvement compared to those utilizing other synthetic datasets, while requiring less than half the data. This benchmark establishes performance evaluation of UAV multi-modal environmental perception models, and hopefully encourages more research efforts towards enabling UAVs to undertake more multi-modal tasks. The dataset and its generation engine are openly accessible under a permissive license at https://github.com/1e12Leon/UEMM-Air.

Designing a proper scoring function is the key to ensuring the excellent performance of knowledge base (KB) embedding. Recently, the scoring function search method introduces the automated machine learning technique to design the data-aware scoring function for the given binary relational data (a.k.a. knowledge graph, KG), which can consistently achieve good performance on different data sets. However, the current data-aware search method is still not as good as desired. First, the existing model can only search scoring functions on the given binary relational data, which is a special form of N-ary relational KBs. Second, observing that existing scoring functions can exhibit distinct performance on different semantic patterns, we are motivated to explore such semantics by searching pattern-aware scoring functions. Unfortunately, it is hard to extend existing search approaches to the scenarios of N-ary and pattern-aware due to the search efficiency and effectiveness issues. In this paper, we propose latent-based factors to model relational patterns and an efficient search algorithm on the N-ary scenario, i.e., efficient LA tent-based SCO ring function search for N-ary relational KBs (LASCO). The empirical results of LASCO on binary and N-ary relational data sets demonstrate that the proposed method can efficiently search pattern-aware scoring functions and achieve better embedding performance than advanced baselines.

The prediction of stock prices is a highly sought-after topic in the data mining field. In recent decades, many promising methods have been proposed and widely adopted for stock price prediction. However, these methods have inherent limitations, such as low accuracy, lack of transparency, and failure to consider the interactions among stock factors. To address these issues, we propose a UNIversal and interpretable framework for enhancing Stock Price Prediction (abbreviated to UniSPP), which is capable of modeling the interactions among stock factors. UniSPP first builds a fully connected graph, where the nodes and edges are the stock factors and interactions between them, respectively. However, it is a non-trivial task to discover a proper feature interaction subgraph from a large space, especially in discrete graph modeling. Therefore, UniSPP proposes a novel idea to mine the real factor interactions by iteratively sampling subgraphs and optimizing the sampling controller. Empirical studies show that our framework can be incorporated with many popular forecasting models and can effectively discover the suitable factor interaction, which can significantly improve the prediction results of existing models.

Graph self-supervised learning aiming to learn the graph representation without much label information is an important tasks in data mining and machine learning since labeled graph data is scarce and expensive to obtain in the real world. Contrastive learning emerges as a promising solution. However, we show existing graph contrastive learning (GCL) models have a significant issue: they generate representations that collapse into a low-dimensional subspace, resulting in a loss of information and diversity. We believe this issue arises from the strong assumption in current GCL methods that all positive samples should be close and all negative samples should be far in the representation space. From a data engineering view, this assumption fails to deeply mine the graph data and oversimplifies the complexity and heterogeneity of graph data, leading to clustered and redundant representations. To address this issue, we propose GradGCL, a novel method that leverages intrinsic gradient information as an additional input signal to regularize GCL training. The gradient information reflects the optimization process of the representations with respect to the contrastive loss, providing a complementary perspective to the representations. Furthermore, we have designed a soft separation strategy that relaxes the hard separation strategy between positive and negative samples, allowing for more flexibility and diversity in the representation space. We have conducted extensive experiments on various graph-related tasks, using different types of contrastive losses, datasets, and model architectures. We demonstrate that gradients alone can learn graph information and achieve competitive results with representation-based GCL methods. We also show that GradGCL can enhance existing GCL models and prevent the issue of dimensional collapse.

Recently, graph contrastive learning proposes to learn node representations from the unlabeled graph to alleviate the heavy reliance on node labels in graph neural networks (GNNs). The core idea is to generate diverse positive views and negative views according to local subgraphs. Then, GNNs take these views as supervised signals and train the model by maximizing the similarity between positive view pairs of each node and minimizing the similarity between positive and negative views. Regardless of the fruitful progress, existing graph contrastive learning approaches still suffer from low-efficiency, insufficient-expressivity, and unpreserved-locality issues. First, they train GNNs by all nodes, reducing the efficiency due to similar and redundant nodes. Second, they only use limited operations (e.g., edge deletion and feature masking) to generate positive views, thereby restricting their expressivity. Third, they uniformly delete edges and mask node features and may modify important edges and features, thereby damaging the important locality information of nodes. In this paper, we propose an efficient and expressive contrastive learning framework for GNNs, namely E2GCL. Specifically, given a limited node budget, we select a set of representative nodes instead of all nodes to accelerate the GNNs training. Besides, we use three general operations (edge deletion, edge addition, and feature perturbation) to generate expressive and locality-preserved positive views based on edge and feature importance. Extensive experiments on various real-world datasets demonstrate the superior effectiveness and efficiency of our proposed E2GCL.

Recently, continual learning (CL) has attracted much attention due to its widespread applications in the real world. Given a set of data sets sequentially, continual learning aims to achieve good performance on the new data sets while avoiding deterioration in performance on the old data sets. Despite the success, most CL models follow the supervised setting, which limits their potential in data scarcity cases. Thus, some pioneering works study unsupervised CL (UCL) to discuss what CL tricks suit the unsupervised setting. However, their advancements lack in-depth analysis of the characteristics of UCL, especially the lack of attention to the use of old data. We identify that using old data sets is essential for improving the UCL model performance while existing works ignore them. Unfortunately, given a limited data storage budget, it is a nontrivial task to select representative data and effectively replay them without label assistance. To further improve the UCL performance, we present a new method in this paper, named Effective Data Selection and Replay (EDSR) for UCL. Specifi-cally, we analyze that entropy can be an effective data selection metric, where representative data usually exhibit the highest entropy in the representation space. Then, to balance the model stability for old data and the plasticity for new data, we adopt a strategy of replaying those stored representative data with a noise-enhanced knowledge distillation process. The empirical study demonstrates the outstanding performance of EDSR on benchmark computer vision data sets. Especially, EDSR shows strong resistance to forgetting old data knowledge while maintaining high accuracy. The implementation is publicly available at https://github.com/LeeJarvis996/edsr_project/tree/main/EDSR.

Recently, graph neural networks (GNNs) have shown its unprecedented success in many graph-related tasks. However, GNNs face the label scarcity issue as other neural networks do. Thus, recent efforts try to pre-train GNNs on a large-scale unlabeled graph and adapt the knowledge from the unlabeled graph to the target downstream task. The adaptation is generally achieved by fine-tuning the pre-trained GNNs with a limited number of labeled data. However, current GNNs pre-training works focus more on how to better pre-train a GNN, but ignore the importance of fine-tuning to better leverage the transferred knowledge. Only a few works start to investigate a better fine-tuning strategy for pre-trained GNNs. But their designs either have strong assumptions or overlook the data-aware issue behind various downstream domains. To further boost pre-trained GNNs, we propose to search to fine-tune pre-trained GNNs for graph-level tasks (S2PGNN), which can adaptively design a suitable fine-tuning framework for the given pre-trained GNN and downstream data. Unfortunately, it is a non-trivial task to achieve this goal due to two technical challenges. First is the hardness of fine-tuning space design since there lack a systematic and unified exploration in existing literature. Second is the enormous computational overhead required for discovering suitable fine-tuning strategies from the discrete space. To tackle these challenges, S2PGNN first carefully summarizes a search space of fine-tuning strategies that is suitable for GNNs, which is expressive enough to enable powerful strategies to be searched. Then, S2PGNN integrates an efficient search algorithm to solve the computationally expensive search problem from a discrete and large space. The empirical studies show that S2PGNN can be implemented on the top of 10 famous pre-trained GNNs and consistently improve their performance by 9 % to 17 %. Our code is publicly available at https://github.com/zwangeo/icde2024.

In this paper, we address the challenges of data augmentation in Multi-Modal Knowledge Graphs (MMKGs), a relatively under-explored area. We propose a novel diffusion-based generative model, the Simple Denoising Probabilistic Latent Diffusion Model (SimDiff). SimDiff is capable of handling different data modalities including the graph topology in a unified manner by the same diffusion model in the latent space. It enhances the utilization of multi-modal data and encourage the multi-modal fusion and reduces the dependency on limited training data. We validate our method in downstream Entity Alignment (EA) tasks in MMKGs, demonstrating that even when using only half of the seed entities in training, our methods can still achieve superior performance. This work contributes to the field by providing a new data generation or augmentation method for MMKGs, potentially paving the way for more effective use of MMKGs in various applications. Code is made available at https://github.com/ranlislz/SimDiff.

Recently, emergence has received widespread attention from the research community along with the success of large-scale models.Different from the literature, we hypothesize a key factor that promotes the performance during the increase of scale: the reduction of monosemantic neurons that can only form one-to-one correlations with specific features.Monosemantic neurons tend to be sparser and have negative impacts on the performance in large models.Inspired by this insight, we propose an intuitive idea to identify monosemantic neurons and inhibit them.However, achieving this goal is a non-trivial task as there is no unified quantitative evaluation metric and simply banning monosemantic neurons does not promote polysemanticity in neural networks.Therefore, we first propose a new metric to measure the monosemanticity of neurons with the guarantee of efficiency for online computation, then introduce a theoretically supported method to suppress monosemantic neurons and proactively promote the ratios of polysemantic neurons in training neural networks.We validate our conjecture that monosemanticity brings about performance change at different model scales on a variety of neural networks and benchmark datasets in different areas, including language, image, and physics simulation tasks.Further experiments validate our analysis and theory regarding the inhibition of monosemanticity.

Graph neural networks (GNNs) have achieved great success on various graph tasks. However, recent studies have revealed that GNNs are vulnerable to injective attacks. Due to the openness of platforms, attackers can inject malicious nodes with carefully designed edges and node features, making GNNs misclassify the labels of target nodes. To resist such adversarial attacks, recent researchers propose GNN defenders. They assume that the attack patterns have been known, e.g., attackers tend to add edges between dissimilar nodes. Then, they remove edges between dissimilar nodes from attacked graphs, aiming to alleviate the negative impact of adversarial attacks. Nevertheless, on dynamic graphs, attackers can change their attack strategies at different times, making existing passive GNN defenders that are passively designed for specific attack patterns fail to resist attacks. In this paper, we propose a novel active GNN defender for dynamic graphs, namely ADGNN, which actively injects guardian nodes to protect target nodes from effective attacks. Specifically, we first formulate an active defense objective to design guardian node behaviors. This objective targets to disrupt the prediction of attackers and protect easily attacked nodes, thereby preventing attackers from generating effective attacks. Then, we propose a gradient-based algorithm with two acceleration techniques to optimize this objective. Extensive experiments on four real-world graph datasets demonstrate the effectiveness of our proposed defender and its capacity to enhance existing GNN defenders.

Unsupervised image denoising has been proposed to alleviate the widespread noise problem without requiring clean images. Existing works mainly follow the self-supervised way, which tries to reconstruct each pixel x of noisy images without the knowledge of x. More recently, some pioneer works further emphasize the importance of x and propose to weigh the information extracted from x and other pixels when recovering x. However, such a method is highly sensitive to the standard deviation σnof noise injected to clean images, where σnis inaccessible without knowing clean images. Thus, it is unrealistic to assume that σnis known for pursuing high model performance.To alleviate this issue, we propose Noise2Info to extract the critical information, the standard deviation σnof injected noise, only based on the noisy images. Specifically, we first theoretically provide an upper bound on σn, while the bound requires clean images. Then, we propose a novel method to estimate the bound of σnby only using noisy images. Besides, we prove that the difference between our estimation with the true deviation goes smaller as the model training. Empirical studies show that Noise2Info is effective and robust on benchmark data sets and closely estimates the standard deviation of noise during model training.

Recently, the message passing neural network (MPNN) has attracted a lot of attention, which learns node representations based on the receptive field of the given node. Despite its success in many graph-related tasks, recent studies find that conventional MPNNs are incapable of handling variant receptive fields required in different graphs, and thereby some upgraded MPNNs have been developed. However, these methods are limited to designing a common solution for different graphs, which fails to capture the impact of different graph properties on the receptive fields. To alleviate such issues, we propose a novel MPNN space for data-dependent receptive fields (MpnnDRF), which enables us to dynamically design suitable MPNNs to capture the receptive field for the given graph. More concretely, we systemically investigate the capability of existing designs and propose several key design dimensions to improve them. Then, to fully explore the proposed designs and useful designs in existing works, we propose a novel search space to incorporate them and formulate a search framework. In the empirical study, the proposed MpnnDRF shows very strong robustness against the increased receptive field, which allows MpnnDRF to learn node representations based on a larger perceptual field. Therefore, MpnnDRF consistently achieves outstanding performance on benchmark node and graph classification tasks.

Recently, many promising embedding models have been proposed to embed knowledge graphs (KGs) and their more general forms, such as n-ary relational data (NRD) and hyper-relational KG (HKG). To promote the data adaptability and performance of embedding models, KG searching methods propose to search for suitable models for a given KG data set. But they are restricted to a single KG form, and the searched models are restricted to a single type of embedding model. To tackle such issues, we propose to build a search space for the message function in graph neural networks (GNNs). However, it is a non-trivial task. Existing message function designs fix the structures and operators, which makes them difficult to handle different KG forms and data sets. Therefore, we first design a novel message function space, which enables both structures and operators to be searched for the given KG form (including KG, NRD, and HKG) and data. The proposed space can flexibly take different KG forms as inputs and is expressive to search for different types of embedding models. Especially, some existing message function designs and some classic KG embedding models can be instantiated as special cases of our space. We empirically show that the searched message functions are data-dependent, and can achieve leading performance on benchmark KGs, NRD, and HKGs.

Recently, incremental learning has attracted a lot of interest in both research communities and industries. Generally, given a series of data sets sequentially, it tries to achieve good performance on the new data set while maintaining not bad performance on the old ones. Despite the recent success of incremental learning, existing works mainly assume that the coming data set is from the feature space of old ones, i.e., homogeneous feature space. And they adopt one feature extractor to forcibly project different feature spaces into one space. However, this assumption is hard to hold in real-world scenarios. Especially, the attributes of tables may sequentially increase in tabular learning. Thus, classic incremental learning models may hinder their effectiveness. In this paper, we propose a new method, incremental tabular learning on heterogeneous feature space (ILEAHE) to solve this issue. We first propose the ideas that feature extractors should be decomposed into shared and specific extractors to process the shared and specific features across different data sets respectively. Then, we propose a novel measurement named discriminative ability to measure specific extractors. Thus, two kinds of extractors can be discriminated and the specific extractor will more focus on those domain-specific features. We further demonstrate the effectiveness of ILEAHE through empirical studies.

Graph neural networks (GNNs) have achieved great success on various graph tasks. However, recent studies have re-vealed that GNNs are vulnerable to adversarial attacks, including topology modifications and feature perturbations. Regardless of the fruitful progress, existing attackers require node labels and GNN parameters to optimize a bi-level problem, or cannot cover both topology modifications and feature perturbations, which are not practical, efficient, or effective. In this paper, we propose a black-box attacker PEEGA, which is restricted to access node features and graph topology for practicability. Specifically, we propose to measure the negative impact of various adversarial attacks from the perspective of node representations, thereby we formulate a single-level problem that can be efficiently solved. Furthermore, we observe that existing attackers tend to blur the context of nodes through adding edges between nodes with different labels. As a result, GNNs are unable to recognize nodes. Based on this observation, we propose a GNN defender GNAT, which incorporates three augmented graphs, i.e., a topology graph, a feature graph, and an ego graph, to make the context of nodes more distinguishable. Extensive experiments on three real-world datasets demonstrate the effectiveness and efficiency of our proposed attacker, despite the fact that we do not access node labels and GNN parameters. Moreover, the effectiveness and efficiency of our proposed defender are also validated by substantial experiments.

Recent studies have demonstrated that recommender systems (RecSys) are vulnerable to injective attacks.Given a limited fake user budget, attackers can inject fake users with carefully designed behaviors into the open platforms, making RecSys recommend a target item to more real users for profits. In this paper, we first revisit existing attackers and reveal that they suffer from the difficulty-agnostic and diversity-deficit issues. Existing attackers concentrate their efforts on difficult users who have low tendencies toward the target item, thus reducing their effectiveness. Moreover, they are incapable of affecting the target RecSys to recommend the target item to real users in a diverse manner, because their generated fake user behaviors are dominated by large communities. To alleviate these two issues, we propose a difficulty and diversity aware attacker, namely DADA. We design the difficulty-aware and diversity-aware objectives to enable easy users from various communities to contribute more weights when optimizing attackers. By incorporating these two objectives, the proposed attacker DADA can concentrate on easy users while also affecting a broader range of real users simultaneously, thereby boosting the effectiveness. Extensive experiments on three real-world datasets demonstrate the effectiveness of our proposed attacker.

The scoring function, which measures the plausibility of triplets in knowledge graphs (KGs), is the key to ensure the excellent performance of KG embedding, and its design is also an important problem in the literature. Automated machine learning (AutoML) techniques have recently been introduced into KG to design task-aware scoring functions, which achieve the state-of-the-art performance in KG embedding. However, the effectiveness of searched scoring functions is still not as good as desired. In this paper, observing that existing scoring functions can exhibit distinct performance on different semantic patterns, we are motivated to explore such semantics by searching relationa-ware scoring functions. But the relation-aware search requires a much larger search space than the previous one. Hence, we propose to encode the space as a supernet and propose an efficient alternative minimization algorithm to search through the supernet in a one-shot manner. Finally, experimental results on benchmark datasets demonstrate that the proposed method can efficiently search relation-aware scoring functions, and achieve better embedding performance than state-of-the-art methods1.

Recently, Graph Neural Networks (GNNs) have gained popularity in a variety of real-world scenarios. Despite the great success, the architecture design of GNNs heavily relies on manual labor. Thus, automated graph neural network (AutoGNN) has attracted interest and attention from the research community, which makes significant performance improvements in recent years. However, existing AutoGNN works mainly adopt an implicit way to model and leverage the link information in the graphs, which is not well regularized to the link prediction task on graphs, and limits the performance of AutoGNN for other graph tasks. In this paper, we present a novel AutoGNN work that explicitly models the link information, abbreviated to AutoGEL. In such a way, AutoGEL can handle the link prediction task and improve the performance of AutoGNNs on the node classification and graph classification task. Moreover, AutoGEL proposes a novel search space containing various design dimensions at both intra-layer and inter-layer designs and adopts a more robust differentiable search algorithm to further improve efficiency and effectiveness. Experimental results on benchmark data sets demonstrate the superiority of AutoGEL on several tasks.

Anomaly detection in time series is a research area of increasing importance. In order to safeguard the availability and stability of services, large companies need to monitor various time-series data to detect anomalies in real time for troubleshooting, thereby reducing potential economic losses. However, in many practical applications, time-series anomaly detection is still an intractable problem due to the huge amount of data, complex data patterns, and limited computational resources. SPOT is an efficient streaming algorithm for anomaly detection, but it is only sensitive to extreme values in the whole data distribution. In this paper, we propose FluxEV, a fast and effective unsupervised anomaly detection framework. By converting the non-extreme anomalies to extreme values, our framework addresses the limitation of SPOT and achieves a huge improvement in the detection accuracy. Moreover, Method of Moments is adopted to speed up the parameter estimation in the automatic thresholding. Extensive experiments show that FluxEV greatly outperforms the state-of-the-art baselines on two large public datasets while ensuring high efficiency.

Tensor, an extension of the vector and matrix to the multi-dimensional case, is a natural way to describe the N-ary relational data. Recently, tensor decomposition methods have been introduced into N-ary relational data and become state-of-the-art on embedding learning. However, the performance of existing tensor decomposition methods is not as good as desired. First, they suffer from the data-sparsity issue since they can only learn from the N-ary relational data with a specific arity, i.e., parts of common N-ary relational data. Besides, they are neither effective nor efficient enough to be trained due to the over-parameterization problem. In this paper, we propose a novel method, i.e., S2S, for effectively and efficiently learning from the N-ary relational data. Specifically, we propose a new tensor decomposition framework, which allows embedding sharing to learn from facts with mixed arity. Since the core tensors may still suffer from the over-parameterization, we propose to reduce parameters by sparsifying the core tensors while retaining their expressive power using neural architecture search (NAS) techniques, which can search for data-dependent architectures. As a result, the proposed S2S not only guarantees to be expressive but also efficiently learns from mixed arity. Finally, empirical results have demonstrated that S2S is efficient to train and achieves state-of-the-art performance. 1

Relation extraction in knowledge base construction has been researched for the last decades due to its applicability to many problems. Most classical works, such as supervised information extraction and distant supervision, focus on how to construct the knowledge base (KB) by utilizing the large number of labels or certain related KBs. However, in many real-world scenarios, the existing methods may not perform well when a new knowledge base is required but only scarce labels or few related KBs available. In this paper, we propose a novel approach called, Relation Extraction via Domain-aware Transfer Learning (ReTrans), to extract relation mentions from a given text corpus by exploring the experience from a large amount of existing KBs which may not be closely related to the target relation. We first propose to initialize the representation of relation mentions from the massive text corpus and update those representations according to existing KBs. Based on the representations of relation mentions, we investigate the contribution of each KB to the target task and propose to select useful KBs for boosting the effectiveness of the proposed approach. Based on selected KBs, we develop a novel domain-aware transfer learning framework to transfer knowledge from source domains to the target domain, aiming to infer the true relation mentions in the unstructured text corpus. Most importantly, we give the stability and generalization bound of ReTrans. Experimental results on the real world datasets well demonstrate that the effectiveness of our approach, which outperforms all the state-of-the-art baselines.

Transfer learning has gained increasing attention due to the inferior performance of machine learning algorithms with insufficient training data. Most of the previous homogeneous or heterogeneous transfer learning works aim to learn a mapping function between feature spaces based on the inherent correspondence across the source and target domains or labeled instances. However, in many real world applications, existing methods may not be robust when the correspondence across domains is noisy or labeled instances are not representative. In this paper, we develop a novel transfer learning framework called Transfer Learning via Feature Isomorphism Discovery (abbreviated to TLFid), which owns high tolerance for noisy correspondence between domains as well as scarce or non-existing labeled instances. More specifically, we propose a feature isomorphism approach to discovering common substructures across feature spaces and learning a feature mapping function from the target domain to the source domain. We evaluate the performance of TLFid on the cross-lingual sentiment classification tasks. The results show that our method achieves significant improvement in terms of accuracy compared with the state-of-the-art methods.