Heterogeneous Graph Neural Networks (HGNNs) are widely used for deep learning on heterogeneous graphs. Typical end-to-end HGNNs require repetitive message passing during training, limiting efficiency for large-scale real-world graphs. Pre-computation-based HGNNs address this by performing message passing only once during preprocessing, collecting neighbor information into regular-shaped tensors, which enables efficient mini-batch training. Label-based pre-computation methods collect neighbors' label information but suffer from training label leakage, where a node's own label information propagates back to itself during multi-hop message passing—the echo effect. Existing mitigation strategies are memory-inefficient on large graphs or suffer from compatibility issues with advanced message passing methods. We propose Echoless Label-based Pre-computation (Echoless-LP), which eliminates training label leakage with Partition-Focused Echoless Propagation (PFEP). PFEP partitions target nodes and performs echoless propagation, where nodes in each partition collect label information only from neighbors in other partitions, avoiding echo while remaining memory-efficient and compatible with any message passing method. We also introduce an Asymmetric Partitioning Scheme (APS) and a PostAdjust mechanism to address information loss from partitioning and distributional shifts across partitions. Experiments on public datasets demonstrate that Echoless-LP achieves superior performance and maintains memory efficiency compared to baselines.

Isolated cold-start node classification on multimodal graphs is challenging because such nodes have no edges and often have missing modalities (e.g., absent text or image features). Existing methods address structural isolation by degrading graph learning models to multilayer perceptrons (MLPs) for isolated cold-start inference, using a teacher model (with graph access) to guide the MLP. However, this results in limited model capacity in the student, which is further challenged when modalities are missing. In this paper, we propose Neighbor-to-Self Graph Transformer (NTSFormer), a unified Graph Transformer framework that jointly tackles the isolation and missing-modality issues via a self-teaching paradigm. Specifically, NTSFormer uses a cold-start attention mask to simultaneously make two predictions for each node: a "student" prediction based only on self information (i.e., the node's own features), and a "teacher" prediction incorporating both self and neighbor information. This enables the model to supervise itself without degrading to an MLP, thereby fully leveraging the Transformer's capacity to handle missing modalities. To handle diverse graph information and missing modalities, NTSFormer performs a one-time multimodal graph pre-computation that converts structural and feature data into token sequences, which are then processed by Mixture-of-Experts (MoE) Input Projection and Transformer layers for effective fusion. Experiments on public datasets show that NTSFormer achieves superior performance for multimodal isolated cold-start node classification.

Real-world fraud detection applications benefit from graph learning techniques that jointly exploit node features—often rich in textual data—and graph structural information. Recently, Graph-Enhanced LLMs have emerged as a promising graph learning approach that converts graph information into prompts, exploiting LLMs' ability to reason over both textual and structural information. Among them, text-only prompting, which converts graph information into prompts consisting solely of text tokens, offers a solution that relies only on LLM tuning without requiring additional graph-specific encoders. However, text-only prompting struggles on heterogeneous fraud-detection graphs: multi-hop relations expand exponentially with each additional hop, leading to rapidly growing neighborhoods associated with dense textual information. These neighborhoods may overwhelm the model with long, irrelevant content in the prompt and suppress key signals from the target node, thereby degrading performance. To address this challenge, we propose Dual Granularity Prompting (DGP), which mitigates information overload by preserving fine-grained textual details for the target node while summarizing neighbor information into coarse-grained text prompts. DGP introduces tailored summarization strategies for different data modalities—bi-level semantic abstraction for textual fields and statistical aggregation for numerical features—enabling effective compression of verbose neighbor content into concise, informative prompts. Experiments across public and industry datasets demonstrate that DGP operates within a manageable token budget while improving fraud detection performance by up to 6.8% (AUPRC) over state-of-the-art methods, showing the potential of Graph-Enhanced LLMs for fraud detection.

Multimodal data plays a critical role in web-based recommendation systems, where information from diverse modalities such as vision and text enhances representation learning. However, real-world multimodal datasets often suffer from modality incompleteness due to sensor failures, annotation scarcity, or privacy constraints, which substantially degrade model performance and reliability. One effective solution to address this issue is modality completion, which reconstructs missing features to provide modality-complete graphs for downstream tasks. Given a query node with missing multimodal features, existing modality completion methods typically infer information from the node itself or its neighbors to reconstruct the missing modality. However, these methods may overlook semantically relevant context in the graph, which contains valuable cues that are non-trivial to capture through simple methods like neighborhood aggregation. In this work, we propose GRE-MC, a Graph Retrieval–Enhanced Modality Completion framework, to overcome these limitations. By introducing a modality-aware subgraph retrieval mechanism, GRE-MC selects semantically relevant subgraphs from the entire graph, providing richer contextual information for completing missing modalities. Subsequently, a graph transformer jointly encodes the query node and the retrieved subgraph via global attention to complete the missing features, while a learnable sparse-routing codebook regularizes latent embeddings into compact bases for improved robustness. Extensive experiments on multimodal recommendation benchmarks demonstrate that GRE-MC consistently outperforms state-of-the-art methods, validating the effectiveness of subgraph retrieval and joint-encoding graph transformer for robust modality completion.

Fine-tuning enables large language models (LLMs) to adapt to specific domains, but often compromises their previously established safety alignment. To mitigate the degradation of model safety during fine-tuning, we introduce LookAhead Tuning, a lightweight and effective data-driven approach that preserves safety during fine-tuning. The method introduces two simple strategies that modify training data by previewing partial answer prefixes, thereby minimizing perturbations to the model's initial token distributions and maintaining its built-in safety mechanisms. Comprehensive experiments demonstrate that LookAhead Tuning effectively maintains model safety without sacrificing robust performance on downstream tasks. Our findings position LookAhead Tuning https://github.com/zjunlp/LookAheadTuning as a reliable and efficient solution for the safe and effective adaptation of LLMs.

Graph pre-training has been concentrated on graph-level tasks involving small graphs (e.g., molecular graphs) or learning node representations on a fixed graph. Extending graph pre-trained models to web-scale graphs with billions of nodes in industrial scenarios, while avoiding negative transfer across graphs or tasks, remains a challenge. We aim to develop a general graph pre-trained model with inductive ability that can make predictions for unseen new nodes and even new graphs. In this work, we introduce a scalable transformer-based graph pre-training framework called (Pre-trained Graph Transformer). Based on the masked autoencoder architecture, we design two pre-training tasks: one for reconstructing node features and the other for reconstructing local structures. Unlike the original autoencoder architecture where the pre-trained decoder is discarded, we propose a novel strategy that utilizes the decoder for feature augmentation. Our framework, tested on the publicly available ogbn-papers100M dataset with 111 million nodes and 1.6 billion edges, achieves state-of-the-art performance, showcasing scalability and efficiency. We have deployed our framework on Tencent's online game data, confirming its capability to pre-train on real-world graphs with over 540 million nodes and 12 billion edges and to generalize effectively across diverse static and dynamic downstream tasks.

Multimodal recommendation systems can learn users' preferences from existing user-item interactions as well as the semantics of multimodal data associated with items. Many existing methods model this through a multimodal user-item graph, approaching multimodal recommendation as a graph learning task. Graph Neural Networks (GNNs) have shown promising performance in this domain. Prior research has capitalized on GNNs' capability to capture neighborhood information within certain receptive fields (typically denoted by the number of hops, K) to enrich user and item semantics. We observe that the optimal receptive fields for GNNs can vary across different modalities. In this paper, we propose GNNs with Modality-Independent Receptive Fields, which employ separate GNNs with independent receptive fields for different modalities to enhance performance. Our results indicate that the optimal K for certain modalities on specific datasets can be as low as 1 or 2, which may restrict the GNNs' capacity to capture global information. To address this, we introduce a Sampling-based Global Transformer, which utilizes uniform global sampling to effectively integrate global information for GNNs. We conduct comprehensive experiments that demonstrate the superiority of our approach over existing methods.

Phishing attacks are a major threat to online security, exploiting user vulnerabilities to steal sensitive information. Various methods have been developed to counteract phishing, each with varying levels of accuracy, but they also face notable limitations. In this study, we introduce PhishAgent, a multimodal agent that combines a wide range of tools, integrating both online and offline knowledge bases with Multimodal Large Language Models (MLLMs). This combination leads to broader brand coverage, which enhances brand recognition and recall. Furthermore, we propose a multimodal information retrieval framework designed to extract the relevant top k items from offline knowledge bases, using available information from a webpage, including logos and HTML. Our empirical results, based on three real-world datasets, demonstrate that the proposed framework significantly enhances detection accuracy and reduces both false positives and false negatives, while maintaining model efficiency. Additionally, PhishAgent shows strong resilience against various types of adversarial attacks.

With the advancement of technology, large language models (LLMs) have achieved remarkable performance across various natural language processing (NLP) tasks, powering LLMintegrated applications like Microsoft Copilot.However, as LLMs continue to evolve, new vulnerabilities, especially prompt injection attacks arise.These attacks trick LLMs into deviating from the original input instructions and executing the attacker's instructions injected in data content, such as retrieved results.Recent attack methods leverage LLMs' instruction-following abilities and their inabilities to distinguish instructions injected in the data content, and achieve a high attack success rate (ASR).When comparing the attack and defense methods, we interestingly find that they share similar design goals, of inducing the model to ignore unwanted instructions and instead to execute wanted instructions.Therefore, we raise an intuitive question: Could these attack techniques be utilized for defensive purposes?In this paper, we invert the intention of prompt injection methods to develop novel defense methods based on previous trainingfree attack methods, by repeating the attack process but with the original input instruction rather than the injected instruction.Our comprehensive experiments demonstrate that our defense techniques outperform existing defense approaches, achieving state-of-the-art results. 1

Prompt injection attacks manipulate large language models (LLMs) by misleading them to deviate from the original input instructions and execute maliciously injected instructions, because of their instruction-following capabilities and inability to distinguish between the original input instructions and maliciously injected instructions.To defend against such attacks, recent studies have developed various detection mechanisms.If we restrict ourselves specifically to works which perform detection rather than direct defense, most of them focus on direct prompt injection attacks, while there are few works for the indirect scenario, where injected instructions are indirectly from external tools, such as a search engine.Moreover, current works mainly investigate injection detection methods and pay less attention to the post-processing method that aims to mitigate the injection after detection.In this paper, we investigate the feasibility of detecting and removing indirect prompt injection attacks, and we construct a benchmark dataset for evaluation.For detection, we assess the performance of existing LLMs and open-source detection models, and we further train detection models using our crafted training datasets.For removal, we evaluate two intuitive methods: (1) the segmentation removal method, which segments the injected document and removes parts containing injected instructions, and (2) the extraction removal method, which trains an extraction model to identify and remove injected instructions.1

Large language models (LLMs) face significant challenges in handling long-context tasks because of their limited effective context window size during pretraining, which restricts their ability to generalize over extended sequences. Meanwhile, extending the context window in LLMs through post-pretraining is highly resource-intensive.To address this, we introduce LongRecipe, an efficient training strategy for extending the context window of LLMs, including impactful token analysis, position index transformation, and training optimization strategies. It simulates long-sequence inputs while maintaining training efficiency and significantly improves the model’s understanding of long-range dependencies. Experiments on three types of LLMs show that LongRecipe can utilize long sequences while requiring only 30% of the target context window size, and reduces computational training resource over 85% compared to full sequence training. Furthermore, LongRecipe also preserves the original LLM’s capabilities in general tasks. Ultimately, we can extend effective context window of open-source LLMs from 8k to 128k, achieving performance close to GPT-4 with just one day of dedicated training using a single GPU with 80G memory.Our code is released at https://github.com/zhiyuanhubj/LongRecipe.

Zhecheng Li, Yiwei Wang, Bryan Hooi, Yujun Cai, Zhen Xiong, Nanyun Peng, Kai-Wei Chang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Recent works integrating Knowledge Graphs (KGs) have shown promising improvements in enhancing the reasoning capabilities of Large Language Models (LLMs).However, existing benchmarks primarily focus on closedended tasks, leaving a gap in evaluating performance on more complex, real-world scenarios.This limitation also hinders a thorough assessment of KGs' potential to reduce hallucinations in LLMs.To address this, we introduce OKGQA 1 , a new benchmark specifically designed to evaluate LLMs augmented with KGs in open-ended, real-world question answering settings.OKGQA reflects practical complexities through diverse question types and incorporates metrics to quantify both hallucination rates and reasoning improvements in LLM+KG models.To consider the scenarios in which KGs may contain varying levels of errors, we propose a benchmark variant, OKGQA-P, to assess model performance when the semantics and structure of KGs are deliberately perturbed and contaminated.In this paper, we aims to (1) explore whether KGs can make LLMs more trustworthy in an open-ended setting, and (2) conduct a comparative analysis to shed light on method design.We believe this study can facilitate a more complete performance comparison and encourages continuous improvement in integrating KGs with LLMs to mitigate hallucination, and make LLMs more trustworthy.

Current unlearning methods for large language models usually rely on reverse optimization to reduce target token probabilities. However, this paradigm disrupts the subsequent tokens prediction, degrading model performance and linguistic coherence. Moreover, existing evaluation metrics overemphasize contextual forgetting while inadequately assessing response fluency and relevance. To address these challenges, we propose ReLearn, a data augmentation and fine-tuning pipeline for effective unlearning, along with a comprehensive evaluation framework. This framework introduces Knowledge Forgetting Ratio (KFR) and Knowledge Retention Ratio (KRR) to measure knowledge-level preservation, and Linguistic Score (LS) to evaluate generation quality. Our experiments show that ReLearn successfully achieves targeted forgetting while preserving high-quality outputs. Through mechanistic analysis, we further demonstrate how reverse optimization disrupts coherent text generation, while ReLearn preserves this essential capability.

The training data in large language models is key to their success, but it also presents privacy and security risks, as it may contain sensitive information. Detecting pre-training data is crucial for mitigating these concerns. Existing methods typically analyze target text in isolation or solely with non-member contexts, overlooking potential insights from simultaneously considering both member and non-member contexts. While previous work suggested that member contexts provide little information due to the minor distributional shift they induce, our analysis reveals that these subtle shifts can be effectively leveraged when contrasted with non-member contexts. In this paper, we propose Con-ReCall, a novel approach that leverages the asymmetric distributional shifts induced by member and non-member contexts through contrastive decoding, amplifying subtle differences to enhance membership inference. Extensive empirical evaluations demonstrate that Con-ReCall achieves state-of-the-art performance on the WikiMIA benchmark and is robust against various text manipulation techniques.

Vision-Language Models (VLMs) excel in integrating visual and textual information for vision-centric tasks, but their handling of inconsistencies between modalities is underexplored. We investigate VLMs’ modality preferences when faced with visual data and varied textual inputs in vision-centered settings. By introducing textual variations to four vision-centric tasks and evaluating ten Vision-Language Models (VLMs), we discover a "blind faith in text" phenomenon: VLMs disproportionately trust textual data over visual data when inconsistencies arise, leading to significant performance drops under corrupted text and raising safety concerns. We analyze factors influencing this text bias, including instruction prompts, language model size, text relevance, token order, and the interplay between visual and textual certainty. While certain factors, such as scaling up the language model size, slightly mitigate text bias, others like token order can exacerbate it due to positional biases inherited from language models. To address this issue, we explore supervised fine-tuning with text augmentation and demonstrate its effectiveness in reducing text bias. Additionally, we provide a theoretical analysis suggesting that the blind faith in text phenomenon may stem from an imbalance of pure text and multi-modal data during training. Our findings highlight the need for balanced training and careful consideration of modality interactions in VLMs to enhance their robustness and reliability in handling multi-modal data inconsistencies.

Despite inheriting security measures from underlying language models, Vision-Language Models (VLMs) may still be vulnerable to safety alignment issues. Through empirical analysis, we uncover two critical findings: scenario- matched images can significantly amplify harmful outputs, and contrary to common assumptions in gradient-based attacks, minimal loss values do not guarantee optimal attack effectiveness. Building on these insights, we introduce MLAI (Multi-Loss Adversarial Images), a novel jailbreak framework that leverages scenario-aware image generation for semantic alignment, exploits flat minima theory for robust adversarial image selection, and employs multi- image collaborative attacks for enhanced effectiveness. Extensive experiments demonstrate MLAI’s significant impact, achieving attack success rates of 77.75% on MiniGPT-4 and 82.80% on LLaVA-2, substantially outperforming existing methods by margins of 34.37% and 12.77% respectively. Furthermore, MLAI shows considerable transferability to commercial black-box VLMs, achieving up to 60.11% success rate. Our work reveals fundamental visual vulnerabilities in current VLMs safety mechanisms and underscores the need for stronger defenses. Warning: This paper contains potentially harmful example text.

Large language models (LLMs) have shown remarkable performance across a range of NLP tasks.However, their strong instructionfollowing capabilities and inability to distinguish instructions from data content make them vulnerable to indirect prompt injection attacks.In such attacks, instructions with malicious purposes are injected into external data sources, such as web documents.When LLMs retrieve this injected data through tools, such as a search engine and execute the injected instructions, they provide misled responses.Recent attack methods have demonstrated potential, but their abrupt instruction injection often undermines their effectiveness.Motivated by the limitations of existing attack methods, we propose Topi-cAttack, which prompts the LLM to generate a fabricated conversational transition prompt that gradually shifts the topic toward the injected instruction, making the injection smoother and enhancing the plausibility and success of the attack.Through comprehensive experiments, TopicAttack achieves state-of-the-art performance, with an attack success rate (ASR) over 90% in most cases, even when various defense methods are applied.We further analyze its effectiveness by examining attention scores.We find that a higher injected-to-original attention ratio leads to a greater success probability, and our method achieves a much higher ratio than the baseline methods. 1

Molecule generation is key to drug discovery and materials science, enabling the design of novel compounds with specific properties.Large language models (LLMs) can learn to perform a wide range of tasks from just a few examples.However, generating valid molecules using representations like SMILES is challenging for LLMs in few-shot settings.In this work, we explore how LLMs can generate 100% valid molecules.We evaluate whether LLMs can use SELFIES, a representation where every string corresponds to a valid molecule, for valid molecule generation but find that LLMs perform worse with SELFIES than with SMILES.We then examine LLMs' ability to correct invalid SMILES and find their capacity limited.Finally, we introduce SmiSelf, a cross-chemical language framework for invalid SMILES correction.SmiSelf converts invalid SMILES to SELFIES using grammatical rules, leveraging SELFIES' mechanisms to correct the invalid SMILES.Experiments show that SmiSelf ensures 100% validity while preserving molecular characteristics and maintaining or even enhancing performance on other metrics.SmiSelf helps expand LLMs' practical applications in biomedicine and is compatible with all SMILES-based generative models.Code is available at https: //github.com/wentao228/SmiSelf.

Recent progress in Multimodal Large Language Models (MLLMs) has unlocked powerful cross-modal reasoning abilities, but also raised new safety concerns, particularly when faced with adversarial multimodal inputs. To improve the safety of MLLMs during inference, we introduce a modular and adaptive inference-time intervention technology, AutoSteer, without requiring any fine-tuning of the underlying model. AutoSteer incorporates three core components: (1) a novel Safety Awareness Score (SAS) that automatically identifies the most safety-relevant distinctions among the model’s internal layers; (2) an adaptive safety prober trained to estimate the likelihood of toxic outputs from intermediate representations; and (3) a lightweight Refusal Head that selectively intervenes to modulate generation when safety risks are detected. Experiments on LLaVA-OV and Chameleon across diverse safety-critical benchmarks demonstrate that AutoSteer significantly reduces the Attack Success Rate (ASR) for textual, visual, and cross-modal threats, while maintaining general abilities. These findings position AutoSteer as a practical, interpretable, and effective framework for safer deployment of multimodal AI systems.

Given a heterogeneous multilayer network with various connections in pharmacology, how can we detect components with intensive interactions and strong dependencies? Can we accurately capture suspicious groups in a multi-lot transaction network under camouflage? These challenges related to dense subgraph detection have been extensively studied in simple graphs but remain under-explored in complex networks. Existing methods struggle to effectively handle the intricate dependencies, let alone accurately identify the interrelated dense connected patterns within a series of complex heterogeneous networks. Here, we introduce INDUEN, a novel algorithm designed to detect interrelated densest subgraphs in multilayer networks by leveraging joint optimization of coupled factorization and local search for an elaborate-designed joint density measure. Experimental results demonstrate that INDUEN outperforms the state-of-the-art baselines in accurately detecting interrelated densest sub graphs under various settings. Furthermore, INDUEN uncovers some intriguing patterns in real-world data; it is linearly scalable and achieves more than 35 × speedup compared to the state-of-the-art method Destine.

Multimodal Large Language Models (MLLMs) still struggle with hallucinations despite their impressive capabilities. Recent studies have attempted to mitigate this by applying Direct Preference Optimization (DPO) to multimodal scenarios using preference pairs from text-based responses. However, our analysis of representation distributions reveals that multimodal DPO struggles to align image and text representations and to distinguish between hallucinated and non-hallucinated descriptions. To address these challenges, In this work, we propose a Cross-modal Hierarchical Direct Preference Optimization (CHiP) to address these limitations. We introduce a visual preference optimization module within the DPO framework, enabling MLLMs to learn from both textual and visual preferences simultaneously. Furthermore, we propose a hierarchical textual preference optimization module that allows the model to capture preferences at multiple granular levels, including response, segment, and token levels. We evaluate CHiP through both quantitative and qualitative analyses, with results across multiple benchmarks demonstrating its effectiveness in reducing hallucinations. On the Object HalBench dataset, CHiP outperforms DPO in hallucination reduction, achieving improvements of 52.7% and 55.5% relative points based on the base model Muffin and LLaVA models, respectively. We make all our datasets and code publicly available.

Graphs are a complex and versatile data structure used across various domains, with possibly multi-label nodes playing a particularly crucial role. Examples include proteins in PPI networks with multiple functions and users in social or e-commerce networks exhibiting diverse interests. Tackling multi-label node classification (MLNC) on graphs has led to the development of various approaches. Some methods leverage graph neural networks (GNNs) to exploit label co-occurrence correlations, while others incorporate label embeddings to capture label proximity. However, these approaches fail to account for the intricate influences between labels in non-Euclidean graph data. To address this issue, we decompose the message passing process in GNNs into two operations: propagation and transformation. We then conduct a comprehensive analysis and quantification of the influence correlations between labels in each operation. Building on these insights, we propose a novel model, Label Influence Propagation (LIP). Specifically, we construct a label influence graph based on the integrated label correlations. Then, we propagate high-order influences through this graph, dynamically adjusting the learning process by amplifying labels with positive contributions and mitigating those with negative influence. Finally, our framework is evaluated on comprehensive benchmark datasets, consistently outperforming SOTA methods across various settings, demonstrating its effectiveness on MLNC tasks.

This paper proposes a simple yet effective jailbreak attack named FlipAttack against black-box LLMs. First, from the autoregressive nature, we reveal that LLMs tend to understand the text from left to right and find that they struggle to comprehend the text when the perturbation is added to the left side. Motivated by these insights, we propose to disguise the harmful prompt by constructing a left-side perturbation merely based on the prompt itself, then generalize this idea to 4 flipping modes. Second, we verify the strong ability of LLMs to perform the text-flipping task and then develop 4 variants to guide LLMs to understand and execute harmful behaviors accurately. These designs keep FlipAttack universal, stealthy, and simple, allowing it to jailbreak black-box LLMs within only 1 query. Experiments on 8 LLMs demonstrate the superiority of FlipAttack. Remarkably, it achieves $\sim$78.97% attack success rate across 8 LLMs on average and $\sim$98% bypass rate against 5 guard models on average.

Handcrafting heuristics for solving complex optimization tasks (e.g., route planning and task allocation) is a common practice but requires extensive domain knowledge. Recently, Large Language Model (LLM)-based automatic heuristic design (AHD) methods have shown promise in generating high-quality heuristics without manual interventions. Existing LLM-based AHD methods employ a population to maintain a fixed number of top-performing LLM-generated heuristics and introduce evolutionary computation (EC) to iteratively enhance the population. However, these population-based procedures cannot fully develop the potential of each heuristic and are prone to converge into local optima. To more comprehensively explore the space of heuristics, this paper proposes to use Monte Carlo Tree Search (MCTS) for LLM-based heuristic evolution. The proposed MCTS-AHD method organizes all LLM-generated heuristics in a tree structure and can better develop the potential of temporarily underperforming heuristics. In experiments, MCTS-AHD delivers significantly higher-quality heuristics on various complex tasks. Our code is available.

Graph Neural Networks (GNNs) face significant computational challenges when handling large-scale graphs. To address this, Graph Condensation (GC) methods aim to compress large graphs into smaller, synthetic ones that are more manageable for GNN training. Recently, trajectory matching methods have shown state-of-the-art (SOTA) performance for GC, aligning the model's training behavior on a condensed graph with that on the original graph by guiding the trajectory of model parameters. However, these approaches require repetitive GNN retraining during condensation, making them computationally expensive. To address the efficiency issue, we completely bypass trajectory matching and propose a novel two-stage framework. The first stage, a precomputation stage, performs one-time message passing to extract structural and semantic information from the original graph. The second stage, a diversity-aware adaptation stage, performs class-wise alignment while maximizing the diversity of synthetic features. Remarkably, even with just the precomputation stage, which takes only seconds, our method either matches or surpasses 5 out of 9 baseline results. Extensive experiments show that our approach achieves comparable or better performance while being 96× to 2,455× faster than SOTA methods, making it more practical for large-scale GNN applications. Our code and data are available at https://github.com/Xtra-Computing/GCPA.

Graph Foundation Models hold significant potential for advancing multi-domain graph learning, yet their full capabilities remain largely untapped. Existing works show promising task performance with the “pretrain-then-prompt” paradigm, which lacks theoretical foundations to understand why it works and how much knowledge can be transferred from source domains to the target. In this paper, we introduce BRIDGE, a bounded graph foundation model pre-trained on multi-domains with Generalization guarantees. To learn discriminative source knowledge, we align multi-domain graph features with domain-invariant aligners during pre-training. Then, a lightweight Mixture of Experts (MoE) network is proposed to facilitate downstream prompting through self-supervised selective knowledge assembly and transfer. Further, to determine the maximum amount of transferable knowledge, we derive an optimizable generalization error upper bound from a graph spectral perspective given the Lipschitz continuity. Extensive experiments demonstrate the superiority of BRIDGE on both node and graph classification compared with 15 state-of-the-art baselines.

In real-world applications, node classification on graphs often faces the challenge of class imbalance, where majority classes dominate training, resulting in biased model performance. Traditional Graph Neural Networks (GNNs) often struggle in such scenarios, as they tend to overfit to majority classes while underrepresenting minority classes. Existing solutions, which either prioritize nodes based on class size or synthesize new nodes for minority classes, often fall short of effectively addressing this imbalance issue. This paper introduces a novel approach to class-imbalanced node classification by utilizing a balanced meta-set for importance measurement, where a training node is considered significant if it enhances model performance under an unbiased setting. Our method identifies important nodes that can counteract class imbalance and utilizes them for model training, allowing for fine-grained and dynamic node selection throughout the training process. We theoretically derive a formula to directly assess node importance, reducing computational overhead and providing an intuitive threshold for node selection. Guided by this metric, we develop a novel framework that filters valuable labeled, unlabeled, and synthetic nodes that enhance model performance in an unbiased context. A key advantage of this framework is its separation of the synthetic node generation process from the filtering process, ensuring compatibility with various node generation techniques. Furthermore, we introduce a strategy to construct a high-quality meta-set that closely approximates the overall feature distribution, ensuring robust representation of each class. We evaluate our framework, NodeImport, across multiple benchmark datasets using popular GNN architectures, demonstrating its superiority over state-of-the-art baselines. Our results highlight the flexibility and effectiveness of the framework in mitigating class imbalance, leading to improved node classification outcomes. The source code is available at https://github.com/NanChanNN/NodeImport.

Foundation models like ChatGPT and GPT-4 have revolutionized artificial intelligence, exhibiting remarkable abilities to generalize across a wide array of tasks and applications beyond their initial training objectives. However, graph learning has predominantly focused on single-graph models, tailored to specific tasks or datasets, lacking the ability to transfer learned knowledge to different domains. This limitation stems from the inherent complexity and diversity of graph structures, along with the different feature and label spaces specific to graph data. In this paper, we recognize text as an effective unifying medium and employ Text-Attributed Graphs (TAGs) to leverage this potential. We present our UniGraph framework, designed to learn a foundation model for TAGs, which is capable of generalizing to unseen graphs and tasks across diverse domains. Unlike single-graph models that use pre-computed node features of varying dimensions as input, our approach leverages textual features for unifying node representations, even for graphs such as molecular graphs that do not naturally have textual features. We propose a novel cascaded architecture of Language Models (LMs) and Graph Neural Networks (GNNs) as backbone networks. Additionally, we propose the first pre-training algorithm specifically designed for large-scale self-supervised learning on TAGs, based on Masked Graph Modeling. We introduce graph instruction tuning using Large Language Models (LLMs) to enable zero-shot prediction ability. Our comprehensive experiments across various graph learning tasks and domains demonstrate the model's effectiveness in self-supervised representation learning on unseen graphs, few-shot in-context transfer, and zero-shot transfer, even surpassing or matching the performance of GNNs that have undergone supervised training on target datasets.

Cheng Wang, Yiwei Wang, Yujun Cai, Bryan Hooi. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

Recent advancements in AI agents have demonstrated their growing potential to drive and support scientific discovery. In this work, we introduce MLR-Bench, a comprehensive benchmark for evaluating AI agents on open-ended machine learning research. MLR-Bench includes three key components: (1) 201 research tasks sourced from NeurIPS, ICLR, and ICML workshops covering diverse ML topics; (2) MLR-Judge, an automated evaluation framework combining LLM-based reviewers with carefully designed review rubrics to assess research quality; and (3) MLR-Agent, a modular agent scaffold capable of completing research tasks through four stages: idea generation, proposal formulation, experimentation, and paper writing. Our framework supports both stepwise assessment across these distinct research stages, and end-to-end evaluation of the final research paper. We then use MLR-Bench to evaluate six frontier LLMs and an advanced coding agent, finding that while LLMs are effective at generating coherent ideas and well-structured papers, current coding agents frequently (e.g., in 80\% of the cases) produce fabricated or invalidated experimental results—posing a major barrier to scientific reliability. We validate MLR-Judge through human evaluation, showing high agreement with expert reviewers, supporting its potential as a scalable tool for research evaluation. We open-source MLR-Bench to help the community benchmark, diagnose, and improve AI research agents toward trustworthy and transparent scientific discovery.

Large Language Models (LLMs) are increasingly deployed in high-stakes domains such as science, law, and healthcare, where accurate expressions of uncertainty are essential for reliability and trust. However, current LLMs are often observed to generate incorrect answers with high confidence—a phenomenon known as "overconfidence". Recent efforts have focused on calibrating LLMs' verbalized confidence: i.e., their expressions of confidence in text form, such as "I am 80% confident that...". Existing approaches either rely on prompt engineering or fine-tuning with heuristically generated uncertainty estimates, both of which have limited effectiveness and generalizability. Motivated by the notion of proper scoring rules for calibration in classical machine learning models, we introduce ConfTuner, a simple and efficient fine-tuning method that introduces minimal overhead and does not require ground-truth confidence scores or proxy confidence estimates. ConfTuner relies on a new loss function, tokenized Brier score, which we theoretically prove to be a proper scoring rule, intuitively meaning that it "correctly incentivizes the model to report its true probability of being correct". ConfTuner improves calibration across diverse reasoning tasks and generalizes to black-box models such as GPT-4o. Our results further show that better-calibrated confidence enables downstream gains in self-correction and model cascade, advancing the development of trustworthy LLM systems. The code is available at https://github.com/liushiliushi/ConfTuner.

To enhance the safety of VLMs, this paper introduces a novel reasoning-based VLM guard model dubbed GuardReasoner-VL. The core idea is to incentivize the guard model to deliberatively reason before making moderation decisions via online RL. First, we construct GuardReasoner-VLTrain, a reasoning corpus with 123K samples and 631K reasoning steps, spanning text, image, and text-image inputs. Then, based on it, we cold-start our model's reasoning ability via SFT. In addition, we further enhance reasoning regarding moderation through online RL. Concretely, to enhance diversity and difficulty of samples, we conduct rejection sampling followed by data augmentation via the proposed safety-aware data concatenation. Besides, we use a dynamic clipping parameter to encourage exploration in early stages and exploitation in later stages. To balance performance and token efficiency, we design a length-aware safety reward that integrates accuracy, format, and token cost. Extensive experiments demonstrate the superiority of our model. Remarkably, it surpasses the runner-up by 19.27% F1 score on average, as shown in Figure 1. We release data, code, and models (3B/7B) of GuardReasoner-VL: https://github.com/yueliu1999/GuardReasoner-VL.

Inspired by the remarkable success of foundation models in language and vision, Graph Foundation Models (GFMs) hold significant promise for broad applicability across diverse graph tasks and domains. However, existing GFMs struggle with unstable few-shot fine-tuning, where both performance and adaptation efficiency exhibit significant fluctuations caused by the randomness in the support sample selection and structural discrepancies between the pre-trained and target graphs. How to fine-tune GFMs robustly and efficiently to enable trustworthy knowledge transfer across domains and tasks is the major challenge. In this paper, we propose **GRAVER**, a novel **G**enerative g**RA**ph **V**ocabulari**E**s for **R**obust GFM fine-tuning framework that tackles the aforementioned instability via generative augmentations. Specifically, to identify transferable units, we analyze and extract key class-specific subgraph patterns by ego-graph disentanglement and validate their transferability both theoretically and empirically. To enable effective pre-training across diverse domains, we leverage a universal task template based on ego-graph similarity and construct graph vocabularies via graphon-based generative experts. To facilitate robust and efficient prompt fine-tuning, we grave the support samples with in-context vocabularies, where the lightweight MoE-CoE network attentively routes knowledge from source domains. Extensive experiments demonstrate the superiority of GRAVER over **effectiveness**, **robustness**, and **efficiency** on downstream few-shot node and graph classification tasks compared with **15** state-of-the-art baselines.

Existing foundation models, such as CLIP, aim to learn a unified embedding space for multimodal data, enabling a wide range of downstream web-based applications like search, recommendation, and content classification. However, these models often overlook the inherent graph structures in multimodal datasets, where entities and their relationships are crucial. Multimodal graphs (MMGs) represent such graphs where each node is associated with features from different modalities, while the edges capture the relationships between these entities.On the other hand, existing graph foundation models primarily focus on text-attributed graphs (TAGs) and are not designed to handle the complexities of MMGs. To address these limitations, we propose UniGraph2, a novel cross-domain graph foundation model that enables general representation learning on MMGs, providing a unified embedding space. UniGraph2 employs modality-specific encoders alongside a graph neural network (GNN) to learn a unified low-dimensional embedding space that captures both the multimodal information and the underlying graph structure. We propose a new cross-domain multi-graph pre-training algorithm at scale to ensure effective transfer learning across diverse graph domains and modalities. Additionally, we adopt a Mixture of Experts (MoE) component to align features from different domains and modalities, ensuring coherent and robust embeddings that unify the information across modalities. Extensive experiments on a variety of multimodal graph tasks demonstrate that UniGraph2 significantly outperforms state-of-the-art models in tasks such as representation learning, transfer learning, and multimodal generative tasks, offering a scalable and flexible solution for learning on MMGs.

Graphs represent interconnected structures prevalent in a myriad of real-world scenarios. Effective graph analytics, such as graph learning methods, enables users to gain profound insights from graph data, underpinning various tasks including node classification and link prediction. However, these methods often suffer from data imbalance, a common issue in graph data where certain segments possess abundant data while others are scarce, thereby leading to biased learning outcomes. This necessitates the emerging field of imbalanced learning on graphs, which aims to correct these data distribution skews for more accurate and representative learning outcomes. In this survey, we embark on a comprehensive review of the literature on imbalanced learning on graphs. We begin by providing a definitive understanding of the concept and related terminologies, establishing a strong foundational understanding for readers. Following this, we propose two comprehensive taxonomies: (1) theproblem taxonomy, which describes the forms of imbalance we consider, the associated tasks, and potential solutions and (2) thetechnique taxonomy, which details key strategies for addressing these imbalances, and aids readers in their method selection process. Finally, we suggest prospective future directions for both problems and techniques within the sphere of imbalanced learning on graphs, fostering further innovation in this critical area.

Controllable text generation is a challenging and meaningful field in natural language generation (NLG). Especially, poetry generation is a typical one with well-defined and strict conditions for text generation which is an ideal playground for the assessment of current methodologies. While prior works succeeded in controlling either semantic or metrical aspects of poetry generation, simultaneously addressing both remains a challenge. In this paper, we pioneer the use of the Diffusion model for generating sonnets and Chinese SongCi poetry to tackle such challenges. In terms of semantics, our PoetryDiffusion model, built upon the Diffusion model, generates entire sentences or poetry by comprehensively considering the entirety of sentence information. This approach enhances semantic expression, distinguishing it from autoregressive and large language models (LLMs). For metrical control, its constraint control module which can be trained individually enables us to flexibly incorporate a novel metrical controller to manipulate and evaluate metrics (format and rhythm). The denoising process in PoetryDiffusion allows for the gradual enhancement of semantics and flexible integration of the metrical controller which can calculate and impose penalties on states that stray significantly from the target control distribution. Experimental results on two datasets demonstrate that our model outperforms existing models in terms of automatic evaluation of semantic, metrical, and overall performance as well as human evaluation. Codes are released to https://github.com/ChorlingLau/PoetryDiffusion.

As Natural Language Processing (NLP) systems are increasingly employed in intricate social environments, a pressing query emerges: Can these NLP systems mirror human-esque collaborative intelligence, in a multi-agent society consisting of multiple large language models (LLMs)? This paper probes the collaboration mechanisms among contemporary NLP systems by melding practical experiments with theoretical insights. We fabricate four unique ‘societies’ comprised of LLM agents, where each agent is characterized by a specific ‘trait’ (easy-going or overconfident) and engages in collaboration with a distinct ‘thinking pattern’ (debate or reflection). Through evaluating these multi-agent societies on three benchmark datasets, we discern that certain collaborative strategies not only outshine previous top-tier approaches but also optimize efficiency (using fewer API tokens). Moreover, our results further illustrate that LLM agents manifest human-like social behaviors, such as conformity and consensus reaching, mirroring foundational social psychology theories. In conclusion, we integrate insights from social psychology to contextualize the collaboration of LLM agents, inspiring further investigations into the collaboration mechanism for LLMs. We commit to sharing our code and datasets, hoping to catalyze further research in this promising avenue.

Graph Anomaly Detection (GAD) has surfaced as a significant field of research, predominantly due to its substantial influence in production environments. Although existing approaches for node anomaly detection have shown effectiveness, they have yet to fully address two major challenges: operating in settings with limited supervision and managing class imbalance effectively. In response to these challenges, we propose a novel model, ConsisGAD, which is tailored for GAD in scenarios characterized by limited supervision and is anchored in the principles of consistency training. Under limited supervision, ConsisGAD effectively leverages the abundance of unlabeled data for consistency training by incorporating a novel learnable data augmentation mechanism, thereby introducing controlled noise into the dataset. Moreover, ConsisGAD takes advantage of the variance in homophily distribution between normal and anomalous nodes to craft a simplified GNN backbone, enhancing its capability to distinguish effectively between these two classes. Comprehensive experiments on several benchmark datasets validate the superior performance of ConsisGAD in comparison to state-of-the-art baselines. Our code is available at https://github.com/Xtra-Computing/ConsisGAD.

Representation learning on text-attributed graphs (TAGs) has become a critical research problem in recent years. A typical example of a TAG is a paper citation graph, where the text of each paper serves as node attributes. Initial graph neural network (GNN) pipelines handled these text attributes by transforming them into shallow or hand-crafted features, such as skip-gram or bag-of-words features. Recent efforts have focused on enhancing these pipelines with language models (LMs), which typically demand intricate designs and substantial computational resources. With the advent of powerful large language models (LLMs) such as GPT or Llama2, which demonstrate an ability to reason and to utilize general knowledge, there is a growing need for techniques which combine the textual modelling abilities of LLMs with the structural learning capabilities of GNNs. Hence, in this work, we focus on leveraging LLMs to capture textual information as features, which can be used to boost GNN performance on downstream tasks. A key innovation is our use of \emph{explanations as features}: we prompt an LLM to perform zero-shot classification, request textual explanations for its decision-making process, and design an \emph{LLM-to-LM interpreter} to translate these explanations into informative features for downstream GNNs. Our experiments demonstrate that our method achieves state-of-the-art results on well-established TAG datasets, including \texttt{Cora}, \texttt{PubMed}, \texttt{ogbn-arxiv}, as well as our newly introduced dataset, \texttt{tape-arxiv23}. Furthermore, our method significantly speeds up training, achieving a 2.88 times improvement over the closest baseline on \texttt{ogbn-arxiv}. Lastly, we believe the versatility of the proposed method extends beyond TAGs and holds the potential to enhance other tasks involving graph-text data~\footnote{Our codes and datasets are available at: \url{https://github.com/XiaoxinHe/TAPE}}.

Graph Neural Networks (GNNs) have become the de facto standard for modeling graph-structured data in various applications. Among them, implicit GNNs have shown a superior ability to effectively capture long-range dependencies in underlying graphs. However, implicit GNNs tend to be computationally expensive and have high memory usage, due to 1) their use of full-batch training; and 2) they require a large number of iterations to solve a fixed-point equation. These compromise the scalability and efficiency of implicit GNNs especially on large graphs. In this paper, we aim to answer the question: how can we efficiently train implicit GNNs to provide effective predictions on large graphs? We propose a new scalable and effective implicit GNN (SEIGNN) with a mini-batch training method and a stochastic solver, which can be trained efficiently on large graphs. Specifically, SEIGNN can more effectively incorporate global and long-range information by introducing coarse-level nodes in the mini-batch training method. It also achieves reduced training time by obtaining unbiased approximate solutions with fewer iterations in the proposed solver. Comprehensive experiments on various large graphs demonstrate that SEIGNN outperforms baselines and achieves higher accuracy with less training time compared with existing implicit GNNs.

Empowering large language models (LLMs) to accurately express confidence in their answers is essential for reliable and trustworthy decision-making. Previous confidence elicitation methods, which primarily rely on *white-box access* to internal model information or model fine-tuning, have become less suitable for LLMs, especially closed-source commercial APIs. This leads to a growing need to explore the untapped area of *black-box* approaches for LLM uncertainty estimation. To better break down the problem, we define a systematic framework with three components: *prompting* strategies for eliciting verbalized confidence, *sampling* methods for generating multiple responses, and *aggregation* techniques for computing consistency. We then benchmark these methods on two key tasks—confidence calibration and failure prediction—across five types of datasets (e.g., commonsense and arithmetic reasoning) and five widely-used LLMs including GPT-4 and LLaMA 2 Chat. Our analysis uncovers several key insights: 1) LLMs, when verbalizing their confidence, tend to be *overconfident*, potentially imitating human patterns of expressing confidence. 2) As model capability scales up, both calibration and failure prediction performance improve, yet still far from ideal performance. 3) Employing our proposed strategies, such as human-inspired prompts, consistency among multiple responses, and better aggregation strategies can help mitigate this overconfidence from various perspectives. 4) Comparisons with white-box methods indicate that while white-box methods perform better, the gap is narrow, e.g., 0.522 to 0.605 in AUROC. Despite these advancements, none of these techniques consistently outperform others, and all investigated methods struggle in challenging tasks, such as those requiring professional knowledge, indicating significant scope for improvement. We believe this study can serve as a strong baseline and provide insights for eliciting confidence in black-box LLMs. The code is publicly available at https://github.com/MiaoXiong2320/llm-uncertainty.

Label imbalance and homophily-heterophily mixture are the fundamental problems encountered when applying Graph Neural Networks (GNNs) to Graph Fraud Detection (GFD) tasks. Existing GNN-based GFD models are designed to augment graph structure to accommodate the inductive bias of GNNs towards homophily, by excluding heterophilic neighbors during message passing. In our work, we argue that the key to applying GNNs for GFD is not to exclude but to {\em distinguish} neighbors with different labels. Grounded in this perspective, we introduce Partitioning Message Passing (PMP), an intuitive yet effective message passing paradigm expressly crafted for GFD. Specifically, in the neighbor aggregation stage of PMP, neighbors with different classes are aggregated with distinct node-specific aggregation functions. By this means, the center node can adaptively adjust the information aggregated from its heterophilic and homophilic neighbors, thus avoiding the model gradient being dominated by benign nodes which occupy the majority of the population. We theoretically establish a connection between the spatial formulation of PMP and spectral analysis to characterize that PMP operates an adaptive node-specific spectral graph filter, which demonstrates the capability of PMP to handle heterophily-homophily mixed graphs. Extensive experimental results show that PMP can significantly boost the performance on GFD tasks.

It is commonly perceived that fake news and real news exhibit distinct writing styles, such as the use of sensationalist versus objective language. However, we emphasize that style-related features can also be exploited for style-based attacks. Notably, the advent of powerful Large Language Models (LLMs) has empowered malicious actors to mimic the style of trustworthy news sources, doing so swiftly, cost-effectively, and at scale. Our analysis reveals that LLM-camouflaged fake news content significantly undermines the effectiveness of state-of-the-art text-based detectors (up to 38% decrease in F1 Score), implying a severe vulnerability to stylistic variations. To address this, we introduce SheepDog, a style-robust fake news detector that prioritizes content over style in determining news veracity. SheepDog achieves this resilience through (1) LLM-empowered news reframings that inject style diversity into the training process by customizing articles to match different styles; (2) a style-agnostic training scheme that ensures consistent veracity predictions across style-diverse reframings; and (3) content-focused veracity attributions that distill content-centric guidelines from LLMs for debunking fake news, offering supplementary cues and potential intepretability that assist veracity prediction. Extensive experiments on three real-world benchmarks demonstrate SheepDog's style robustness and adaptability to various backbones.

Given a graph with textual attributes, we enable users to `chat with their graph': that is, to ask questions about the graph using a conversational interface. In response to a user's questions, our method provides textual replies and highlights the relevant parts of the graph. While existing works integrate large language models (LLMs) and graph neural networks (GNNs) in various ways, they mostly focus on either conventional graph tasks (such as node, edge, and graph classification), or on answering simple graph queries on small or synthetic graphs. In contrast, we develop a flexible question-answering framework targeting real-world textual graphs, applicable to multiple applications including scene graph understanding, common sense reasoning, and knowledge graph reasoning. Toward this goal, we first develop a Graph Question Answering (GraphQA) benchmark with data collected from different tasks. Then, we propose our \textit{G-Retriever} method, introducing the first retrieval-augmented generation (RAG) approach for general textual graphs, which can be fine-tuned to enhance graph understanding via soft prompting. To resist hallucination and to allow for textual graphs that greatly exceed the LLM's context window size, \textit{G-Retriever} performs RAG over a graph by formulating this task as a Prize-Collecting Steiner Tree optimization problem. Empirical evaluations show that our method outperforms baselines on textual graph tasks from multiple domains, scales well with larger graph sizes, and mitigates hallucination.~\footnote{Our codes and datasets are available at: \url{https://github.com/XiaoxinHe/G-Retriever}}

In the face of uncertainty, the ability to *seek information* is of fundamental importance. In many practical applications, such as medical diagnosis and troubleshooting, the information needed to solve the task is not initially given, and has to be actively sought by asking follow-up questions (for example, a doctor asking a patient for more details about their symptoms). In this work, we introduce **Uncertainty of Thoughts (UoT)**, an algorithm to augment large language models with the ability to actively seek information by asking effective questions. UoT combines: 1. An *uncertainty-aware simulation approach* which enables the model to simulate possible future scenarios and how likely they are to occur, 2. *Uncertainty-based rewards* motivated by information gain which incentivizes the model to seek information, and 3. A *reward propagation scheme* to select the optimal question to ask in a way that maximizes the expected reward. In experiments on medical diagnosis, troubleshooting and the `20 Questions' game, UoT achieves an average performance improvement of 38.1% in the rate of successful task completion across multiple LLMs compared with direct prompting, and also improves efficiency (i.e., the number of questions needed to complete the task).

Multivariate time series forecasting plays a pivotal role in contemporary web technologies. In contrast to conventional methods that involve creating dedicated models for specific time series application domains, this research advocates for a unified model paradigm that transcends domain boundaries. However, learning an effective cross-domain model presents the following challenges. First, various domains exhibit disparities in data characteristics, e.g., the number of variables, posing hurdles for existing models that impose inflexible constraints on these factors. Second, the model may encounter difficulties in distinguishing data from various domains, leading to suboptimal performance in our assessments. Third, the diverse convergence rates of time series domains can also result in compromised empirical performance. To address these issues, we propose UniTime for effective cross-domain time series learning. Concretely, UniTime can flexibly adapt to data with varying characteristics. It also uses domain instructions and a Language-TS Transformer to offer identification information and align two modalities. In addition, UniTime employs masking to alleviate domain convergence speed imbalance issues. Our extensive experiments demonstrate the effectiveness of UniTime in advancing state-of-the-art forecasting performance and zero-shot transferability.

Given a heterogeneous multilayer network with various connections in pharmacology, how can we detect components with intensive interactions and strong dependencies? Can we accurately capture suspicious groups in a multi-lot transaction network under camouflage? These challenges related to dense subgraph detection have been extensively studied in simple graphs (such as bipartite graph, multi-view network) but remain under-explored on complex networks. Existing methods struggle to effectively handle theintricate dependencies, let alone accurately identify theinterrelated dense connected patternswithin a series of complex heterogeneous networks. In this paper, we proposeInDuen, a novel algorithm designed to detect interrelated densest subgraphs in multilayer networks through joint optimization of coupled factorization and local search for an elaborate-designed joint density measure. It is(a)effective for both large synthetic and real networks,(b)resistant to camouflage for anomaly detection, and(c)linearly scalable. Experimental results demonstrate thatInDuenoutperforms the state-of-the-art baselines in accurately detecting interrelated densest subgraphs under various settings. Furthermore,InDuenuncovers some intriguing patterns in real-world data, i.e., closely cooperated academic groups and interrelated dependent functional components in biology-net.InDuenachieves more than$35 \times$speedup compared to the SOTA methodDestine.

Heterogeneous Graph Neural Networks (HGNNs) are powerful tools for deep learning on heterogeneous graphs. Typical HGNNs require repetitive message passing during training, limiting efficiency for large-scale real-world graphs. Recent pre-computation-based HGNNs use one-time message passing to transform a heterogeneous graph into regular-shaped tensors, enabling efficient mini-batch training. Existing pre-computation-based HGNNs can be mainly categorized into two styles, which differ in how much information loss is allowed and efficiency. We propose a hybrid pre-computation-based HGNN, named Random Projection Heterogeneous Graph Neural Network (RpHGNN), which combines the benefits of one style's efficiency with the low information loss of the other style. To achieve efficiency, the main framework of RpHGNN consists of propagate-then-update iterations, where we introduce a Random Projection Squashing step to ensure that complexity increases only linearly. To achieve low information loss, we introduce a Relation-wise Neighbor Collection component with an Even-odd Propagation Scheme, which aims to collect information from neighbors in a finer-grained way. Experimental results indicate that our approach achieves state-of-the-art results on seven small and large benchmark datasets while also being 230% faster compared to the most effective baseline. Surprisingly, our approach not only surpasses pre-processing-based baselines but also outperforms end-to-end methods.

Graph Neural Networks (GNNs) have recently been utilized to build Collaborative Filtering (CF) models to predict user preferences based on historical user-item interactions. However, there is relatively little understanding of how GNN-based CF models relate to some traditional Network Representation Learning (NRL) approaches. In this paper, we show the equivalence between some state-of-the-art GNN-based CF models and a traditional 1-layer NRL model based on context encoding. Based on a Markov process that trades off two types of distances, we present Markov Graph Diffusion Collaborative Filtering (MGDCF) to generalize some state-of-the-art GNN-based CF models. Instead of considering the GNN as a trainable black box that propagates learnable user/item vertex embeddings, we treat GNNs as an untrainable Markov process that can construct constant context features of vertices for a traditional NRL model that encodes context features with a fully-connected layer. Such simplification can help us to better understand how GNNs benefit CF models. Especially, it helps us realize that ranking losses play crucial roles in GNN-based CF tasks. With our proposed simple yet powerful ranking loss InfoBPR, the NRL model can still perform well without the context features constructed by GNNs. We conduct experiments to perform detailed analysis on MGDCF.

In this article, a modified mutual information maximization (InfoMax) framework, named channel capacity maximization (CapMax), is proposed and applied to learn informative representations for dynamic networks with time-varying topology and/or time-evolving node attributes. The CapMax is based on the network information theory for multiuser communication, where the representation model is treated as a multiaccess communication channel with memory and feedback. Without requirements of the backbone structure, the learning objective of our CapMax is maximizing the channel capacity, which is measured by directed information (DI) rather than mutual information. For efficient implementation, we design an estimator of the channel capacity through the combination of graph neural networks (GNNs) and recurrent neural networks (RNNs). Under some mild conditions, we theoretically prove that DI is a better measure than mutual information in capturing useful information. The experiments are conducted on multiple real-world dynamic network datasets, and the outperformance of our CapMax on different backbone models on link detection and prediction validates the effectiveness of modeling the representation model as a communication channel.

Event-centric structured prediction involves predicting structured outputs of events.In most NLP cases, event structures are complex with manifold dependency, and it is challenging to effectively represent these complicated structured events.To address these issues, we propose Structured Prediction with Energybased Event-Centric Hyperspheres (SPEECH).SPEECH models complex dependency among event structured components with energybased modeling, and represents event classes with simple but effective hyperspheres.Experiments on two unified-annotated event datasets indicate that SPEECH is predominant in event detection and event-relation extraction tasks.

Dialogue systems and large language models (LLMs) have gained considerable attention. However, the direct utilization of LLMs as task-oriented dialogue (TOD) models has been found to underperform compared to smaller task-specific models. Nonetheless, it is crucial to acknowledge the significant potential of LLMs and explore improved approaches for leveraging their impressive abilities. Motivated by the goal of leveraging LLMs, we propose an alternative approach called User-Guided Response Optimization (UGRO) to combine it with a smaller TOD model. This approach uses LLM as an annotation-free user simulator to assess dialogue responses, combining them with smaller fine-tuned end-to-end TOD models. By utilizing the satisfaction feedback generated by LLMs, UGRO further optimizes the supervised fine-tuned TOD model. Specifically, the TOD model takes the dialogue history as input and, with the assistance of the user simulator's feedback, generates high-satisfaction responses that meet the user's requirements. Through empirical experiments on two TOD benchmarks, we validate the effectiveness of our method. The results demonstrate that our approach outperforms previous state-of-the-art (SOTA) results.

Despite considerable advances in automated fake news detection, due to the timely nature of news, it remains a critical open question how to effectively predict the veracity of news articles based on limited fact-checks. Existing approaches typically follow a "Train-from-Scratch" paradigm, which is fundamentally bounded by the availability of large-scale annotated data. While expressive pre-trained language models (PLMs) have been adapted in a "Pre-Train-and-Fine-Tune" manner, the inconsistency between pre-training and downstream objectives also requires costly task-specific supervision. In this paper, we propose "Prompt-and-Align" (P&A), a novel prompt-based paradigm for few-shot fake news detection that jointly leverages the pre-trained knowledge in PLMs and the social context topology. Our approach mitigates label scarcity by wrapping the news article in a task-related textual prompt, which is then processed by the PLM to directly elicit task-specific knowledge. To supplement the PLM with social context without inducing additional training overheads, motivated by empirical observation on user veracity consistency (i.e., social users tend to consume news of the same veracity type), we further construct a news proximity graph among news articles to capture the veracity-consistent signals in shared readerships, and align the prompting predictions along the graph edges in a confidence-informed manner. Extensive experiments on three real-world benchmarks demonstrate that P&A sets new states-of-the-art for few-shot fake news detection performance by significant margins.

Yiwei Wang, Bryan Hooi, Fei Wang, Yujun Cai, Yuxuan Liang, Wenxuan Zhou, Jing Tang, Manjuan Duan, Muhao Chen. Proceedings of the 27th Conference on Computational Natural Language Learning (CoNLL). 2023.

Mean ensemble (i.e. averaging predictions from multiple models) is a commonly-used technique in machine learning that improves the performance of each individual model. We formalize it as feature alignment for ensemble in open-set face recognition and generalize it into Bayesian Ensemble Averaging (BEA) through the lens of probabilistic modeling. This generalization brings up two practical benefits that existing methods could not provide: (1) the uncertainty of a face image can be evaluated and further decomposed into aleatoric uncertainty and epistemic uncertainty, the latter of which can be used as a measure for out-of-distribution detection of faceness; (2) a BEA statistic provably reflects the aleatoric uncertainty of a face image, acting as a measure for face image quality to improve recognition performance. To inherit the uncertainty estimation capability from BEA without the loss of inference efficiency, we propose BEA-KD, a student model to distill knowledge from BEA. BEA-KD mimics the overall behavior of ensemble members and consistently outperforms SOTA knowledge distillation methods on various challenging benchmarks.

Instruction-tuned large language models (LLMs), such as ChatGPT, have led to promising zero-shot performance in discriminative natural language understanding (NLU) tasks.This involves querying the LLM using a prompt containing the question, and the candidate labels to choose from.The question-answering capabilities of ChatGPT arise from its pre-training on large amounts of human-written text, as well as its subsequent fine-tuning on human preferences, which motivates us to ask: Does ChatGPT also inherit humans' cognitive biases?In this paper, we study the primacy effect of ChatGPT: the tendency of selecting the labels at earlier positions as the answer.We have two main findings: i) ChatGPT's decision is sensitive to the order of labels in the prompt; ii) ChatGPT has a clearly higher chance to select the labels at earlier positions as the answer.We hope that our experiments and analyses provide additional insights into building more reliable ChatGPT-based solutions.We release the source code at https: //github.com/wangywUST/PrimacyEffectGPT.

Road safety is a major global public health concern, and effective prediction of traffic accidents at a fine-grained spatial scale plays a critical role in reducing roadway deaths and serious injuries. However, previous studies have either overlooked implicit spatial correlations or inadequately simulated road structures due to the lack of graph-structured datasets. To bridge this gap, we introduce a graph-based Traffic Accident Prediction (TAP) data repository, along with two representative tasks: accident occurrence and severity prediction. With its real-world graph structures, comprehensive geographical coverage, and rich geospatial features, this repository has considerable potential to facilitate various traffic-related tasks. We extensively evaluate eleven Graph Neural Network (GNN) baselines using the constructed datasets. We also develop a novel GNN-based model, which can capture additional angular and directional information from road networks. We demonstrate that the proposed model consistently outperforms the baselines. The data and code are available at https://github.com/baixianghuang/travel.

Business Knowledge Graphs (KGs) are important to many enterprises today, providing factual knowledge and structured data that steer many products and make them more intelligent. Despite their promising benefits, building business KG necessitates solving prohibitive issues of deficient structure and multiple modalities. In this paper, we advance the understanding of the practical challenges related to building KG in non-trivial real-world systems. We introduce the process of building an open business knowledge graph (OpenBG) derived from a well-known enterprise, Alibaba Group. Specifically, we define a core ontology to cover various abstract products and consumption demands, with fine-grained taxonomy and multimodal facts in deployed applications. OpenBG is an open business KG of unprecedented scale: 2.6 billion triples with more than 88 million entities covering over 1 million core classes/concepts and 2,681 types of relations. We release all the open resources (OpenBG benchmarks) derived from it for the community and report experimental results of KG-centric tasks. We also run up an online competition based on OpenBG benchmarks, and has attracted thousands of teams. We further pre-train OpenBG and apply it to many KG-enhanced downstream tasks in business scenarios, demonstrating the effectiveness of billion-scale multimodal knowledge for e-commerce. All the resources with codes have been released at https://github.com/OpenBGBenchmark/OpenBG.

Reliable application of machine learning is of primary importance to the practical deployment of deep learning methods. A fundamental challenge is that models are often unreliable due to overconfidence. In this paper, we estimate a model’s reliability by measuring the agreement between its latent space, and the latent space of a foundation model. However, it is challenging to measure the agreement between two different latent spaces due to their incoherence, e.g., arbitrary rotations and different dimensionality. To overcome this incoherence issue, we design a neighborhood agreement measure between latent spaces and find that this agreement is surprisingly well-correlated with the reliability of a model’s predictions. Further, we show that fusing neighborhood agreement into a model’s predictive confidence in a post-hoc way significantly improves its reliability. Theoretical analysis and extensive experiments on failure detection across various datasets verify the effectiveness of our method on both in-distribution and out-of-distribution settings.

Graph Neural Networks (GNNs) have shown great potential in the field of graph representation learning. Standard GNNs define a local message-passing mechanism which propagates information over the whole graph domain by stacking multiple layers. This paradigm suffers from two major limitations, over-squashing and poor long-range dependencies, that can be solved using global attention but significantly increases the computational cost to quadratic complexity. In this work, we propose an alternative approach to overcome these structural limitations by leveraging the ViT/MLP-Mixer architectures introduced in computer vision. We introduce a new class of GNNs, called Graph ViT/MLP-Mixer, that holds three key properties. First, they capture long-range dependency and mitigate the issue of over-squashing as demonstrated on Long Range Graph Benchmark and TreeNeighbourMatch datasets. Second, they offer better speed and memory efficiency with a complexity linear to the number of nodes and edges, surpassing the related Graph Transformer and expressive GNN models. Third, they show high expressivity in terms of graph isomorphism as they can distinguish at least 3-WL non-isomorphic graphs. We test our architecture on 4 simulated datasets and 7 real-world benchmarks, and show highly competitive results on all of them. The source code is available for reproducibility at: https://github.com/XiaoxinHe/Graph-ViT-MLPMixer.

Label errors have been found to be prevalent in popular text, vision, and audio datasets, which heavily influence the safe development and evaluation of machine learning algorithms. Despite increasing efforts towards improving the quality of generic data types, such as images and texts, the problem of mislabel detection in graph data remains underexplored. To bridge the gap, we explore mislabelling issues in popular real-world graph datasets and propose GraphCleaner, a post-hoc method to detect and correct these mislabelled nodes in graph datasets. GraphCleaner combines the novel ideas of 1) Synthetic Mislabel Dataset Generation, which seeks to generate realistic mislabels; and 2) Neighborhood-Aware Mislabel Detection, where neighborhood dependency is exploited in both labels and base classifier predictions. Empirical evaluations on 6 datasets and 6 experimental settings demonstrate that GraphCleaner outperforms the closest baseline, with an average improvement of $0.14$ in F1 score, and $0.16$ in MCC. On real-data case studies, GraphCleaner detects real and previously unknown mislabels in popular graph benchmarks: PubMed, Cora, CiteSeer and OGB-arxiv; we find that at least 6.91% of PubMed data is mislabelled or ambiguous, and simply removing these mislabelled data can boost evaluation performance from 86.71% to 89.11%.

In Reinforcement Learning (RL), Laplacian Representation (LapRep) is a task-agnostic state representation that encodes the geometry of the environment. A desirable property of LapRep stated in prior works is that the Euclidean distance in the LapRep space roughly reflects the reachability between states, which motivates the usage of this distance for reward shaping. However, we find that LapRep does not necessarily have this property in general: two states having a small distance under LapRep can actually be far away in the environment. Such a mismatch would impede the learning process in reward shaping. To fix this issue, we introduce a Reachability-Aware Laplacian Representation (RA-LapRep), by properly scaling each dimension of LapRep. Despite the simplicity, we demonstrate that RA-LapRep can better capture the inter-state reachability as compared to LapRep, through both theoretical explanations and experimental results. Additionally, we show that this improvement yields a significant boost in reward shaping performance and benefits bottleneck state discovery.

Proposing an effective and flexible matrix to represent a graph is a fundamental challenge that has been explored from multiple perspectives, e.g., filtering in Graph Fourier Transforms. In this work, we develop a novel and general framework which unifies many existing GNN models from the view of parameterized decomposition and filtering, and show how it helps to enhance the flexibility of GNNs while alleviating the smoothness and amplification issues of existing models. Essentially, we show that the extensively studied spectral graph convolutions with learnable polynomial filters are constrained variants of this formulation, and releasing these constraints enables our model to express the desired decomposition and filtering simultaneously. Based on this generalized framework, we develop models that are simple in implementation but achieve significant improvements and computational efficiency on a variety of graph learning tasks. Code is available at https://github.com/qslim/PDF.

Given a stream of graph edges from a dynamic graph, how can we assign anomaly scores to edges and subgraphs in an online manner, for the purpose of detecting unusual behavior, using constant time and memory? For example, in intrusion detection, existing work seeks to detect either anomalous edges or anomalous subgraphs, but not both. In this paper, we first extend the count-min sketch data structure to a higher-order sketch. This higher-order sketch has the useful property of preserving the dense subgraph structure (dense subgraphs in the input turn into dense submatrices in the data structure). We then propose 4 online algorithms that utilize this enhanced data structure, which (a) detect both edge and graph anomalies; (b) process each edge and graph in constant memory and constant update time per newly arriving edge, and; (c) outperform state-of-the-art baselines on 4 real-world datasets. Our method is the first streaming approach that incorporates dense subgraph search to detect graph anomalies in constant memory and time.

Recent efforts in fake news detection have witnessed a surge of interest in using graph neural networks (GNNs) to exploit rich social context. Existing studies generally leverage fixed graph structures, assuming that the graphs accurately represent the related social engagements. However, edge noise remains a critical challenge in real-world graphs, as training on suboptimal structures can severely limit the expressiveness of GNNs. Despite initial efforts in graph structure learning (GSL), prior works often leverage node features to update edge weights, resulting in heavy computational costs that hinder the methods' applicability to large-scale social graphs. In this work, we approach the fake news detection problem with a novel aspect of social graph refinement. We find that the degrees of news article nodes exhibit distinctive patterns, which are indicative of news veracity. Guided by this, we propose DECOR, a novel application of Degree-Corrected Stochastic Blockmodels to the fake news detection problem. Specifically, we encapsulate our empirical observations into a lightweight social graph refinement component that iteratively updates the edge weights via a learnable degree correction mask, which allows for joint optimization with a GNN-based detector. Extensive experiments on two real-world benchmarks validate the effectiveness and efficiency of DECOR1.

Road traffic forecasting plays a critical role in smart city initiatives and has experienced significant advancements thanks to the power of deep learning in capturing non-linear patterns of traffic data. However, the promising results achieved on current public datasets may not be applicable to practical scenarios due to limitations within these datasets. First, the limited sizes of them may not reflect the real-world scale of traffic networks. Second, the temporal coverage of these datasets is typically short, posing hurdles in studying long-term patterns and acquiring sufficient samples for training deep models. Third, these datasets often lack adequate metadata for sensors, which compromises the reliability and interpretability of the data. To mitigate these limitations, we introduce the LargeST benchmark dataset. It encompasses a total number of 8,600 sensors in California with a 5-year time coverage and includes comprehensive metadata. Using LargeST, we perform in-depth data analysis to extract data insights, benchmark well-known baselines in terms of their performance and efficiency, and identify challenges as well as opportunities for future research. We release the datasets and baseline implementations at: https://github.com/liuxu77/LargeST.

Confidence calibration is central to providing accurate and interpretable uncertainty estimates, especially under safety-critical scenarios. However, we find that existing calibration algorithms often overlook the issue of proximity bias, a phenomenon where models tend to be more overconfident in low proximity data (i.e., data lying in the sparse region of the data distribution) compared to high proximity samples, and thus suffer from inconsistent miscalibration across different proximity samples. We examine the problem over $504$ pretrained ImageNet models and observe that: 1) Proximity bias exists across a wide variety of model architectures and sizes; 2) Transformer-based models are relatively more susceptible to proximity bias than CNN-based models; 3) Proximity bias persists even after performing popular calibration algorithms like temperature scaling; 4) Models tend to overfit more heavily on low proximity samples than on high proximity samples. Motivated by the empirical findings, we propose ProCal, a plug-and-play algorithm with a theoretical guarantee to adjust sample confidence based on proximity. To further quantify the effectiveness of calibration algorithms in mitigating proximity bias, we introduce proximity-informed expected calibration error (PIECE) with theoretical analysis. We show that ProCal is effective in addressing proximity bias and improving calibration on balanced, long-tail, and distribution-shift settings under four metrics over various model architectures. We believe our findings on proximity bias will guide the development of fairer and better-calibrated} models, contributing to the broader pursuit of trustworthy AI.

The power of DNNs relies heavily on the quantity and quality of training data. However, collecting and annotating data on a large scale is often expensive and time-consuming. To address this issue, we explore a new task, termed dataset expansion, aimed at expanding a ready-to-use small dataset by automatically creating new labeled samples. To this end, we present a Guided Imagination Framework (GIF) that leverages cutting-edge generative models like DALL-E2 and Stable Diffusion (SD) to "imagine" and create informative new data from the input seed data. Specifically, GIF conducts data imagination by optimizing the latent features of the seed data in the semantically meaningful space of the prior model, resulting in the creation of photo-realistic images with new content. To guide the imagination towards creating informative samples for model training, we introduce two key criteria, i.e., class-maintained information boosting and sample diversity promotion. These criteria are verified to be essential for effective dataset expansion: GIF-SD obtains 13.5% higher model accuracy on natural image datasets than unguided expansion with SD. With these essential criteria, GIF successfully expands small datasets in various scenarios, boosting model accuracy by 36.9% on average over six natural image datasets and by 13.5% on average over three medical datasets. The source code is available at https://github.com/Vanint/DatasetExpansion.

As the state-of-the-art graph learning models, the message passing based neural networks (MPNNs) implicitly use the graph topology as the "pathways" to propagate node features. This implicit use of graph topology induces the MPNNs' over-reliance on (node) features and high inference latency, which hinders their large-scale applications in industrial contexts. To mitigate these weaknesses, we propose the Graph Explicit Neural Network (GENN) framework. GENN can be flexibly applied to various MPNNs and improves them by providing more efficient and accurate inference that is robust in feature-constrained settings. Specifically, we carefully incorporate recent developments in network embedding methods to efficiently prioritize the graph topology for inference. From this vantage, GENN explicitly encodes the topology as an important source of information to mitigate the reliance on node features. Moreover, by adopting knowledge distillation (KD) techniques, GENN takes an MPNN as the teacher to supervise the training for better effectiveness while avoiding the teacher's high inference latency. Empirical results show that our GENN infers dramatically faster than its MPNN teacher by 40x-78x. In terms of accuracy, GENN yields significant gains (more than 40%) for its MPNN teacher when the node features are limited based on our explicit encoding. Moreover, GENN outperforms the MPNN teacher even in feature-rich settings thanks to our KD design.

Deep long-tailed learning, one of the most challenging problems in visual recognition, aims to train well-performing deep models from a large number of images that follow a long-tailed class distribution. In the last decade, deep learning has emerged as a powerful recognition model for learning high-quality image representations and has led to remarkable breakthroughs in generic visual recognition. However, long-tailed class imbalance, a common problem in practical visual recognition tasks, often limits the practicality of deep network based recognition models in real-world applications, since they can be easily biased towards dominant classes and perform poorly on tail classes. To address this problem, a large number of studies have been conducted in recent years, making promising progress in the field of deep long-tailed learning. Considering the rapid evolution of this field, this article aims to provide a comprehensive survey on recent advances in deep long-tailed learning. To be specific, we group existing deep long-tailed learning studies into three main categories (i.e., class re-balancing, information augmentation and module improvement), and review these methods following this taxonomy in detail. Afterward, we empirically analyze several state-of-the-art methods by evaluating to what extent they address the issue of class imbalance via a newly proposed evaluation metric, i.e., relative accuracy. We conclude the survey by highlighting important applications of deep long-tailed learning and identifying several promising directions for future research.

Time series data naturally exist in many domains including medical data analysis, infrastructure sensor monitoring, and motion tracking. However, a very small portion of anomalous time series can be observed, comparing to the whole data. Most existing approaches are based on the supervised classification model requiring representative labels for anomaly class(es), which is challenging in real-world problems. So can we learn how to detect anomalous time ticks in an effective yet efficient way, given mostly normal time series data? Therefore, we propose an unsupervised reconstruction model named BeatGAN which learns to detect anomalies based on normal data, or data which majority of samples are normal. BeatGAN provides a framework to adversarially learn to reconstruct, which can cooperate with both 1-d CNN and RNN. Rarely observed anomalies can result in larger reconstruction errors, which are then detected based on extreme value theory. Moreover, data augmentation with dynamic time warping regularizes reconstruction and provides robustness. In the experiments, effectiveness and sensitivity are studied in both synthetic data and various real-world time series. BeatGAN achieves better accuracy and fast inference.

Sparsity is an established problem for the next Point-of-Interest (POI) recommendation task, where it hinders effective learning of user preferences from the User-POI matrix. However, learning multiple hierarchically related spatial tasks, and visiting relations between users and POIs, can help to alleviate this sparsity problem. In this article, we propose our Hierarchical Multi-Task Graph Recurrent Network (HMT-GRN) approach, which alleviates the sparsity problem by learning different User-Region matrices of lower sparsities in a multi-task setting. We then perform a Hierarchical Beam Search (HBS) on the different region and POI distributions to hierarchically reduce the search space with increasing spatial granularity and predict the next POI. Our HBS provides efficiency gains by reducing the search space, resulting in speedups of 5 to 7 times over an exhaustive approach. In addition, we propose a selectivity layer to predict if the next POI has been visited before by the user to balance between personalization and exploration. Further, we propose a novel Joint Triplet Loss Learning (JTLL) module to learn visited and unvisited relations between users and POIs for the recommendation task. Experimental results on two real-world Location-Based Social Network (LBSN) datasets show that our proposed approach significantly outperforms baseline and the state-of-the-art methods.

Many well-established anomaly detection methods use the distance of a sample to those in its local neighbourhood: so-called `local outlier methods', such as LOF and DBSCAN. They are popular for their simple principles and strong performance on unstructured, feature-based data that is commonplace in many practical applications. However, they cannot learn to adapt for a particular set of data due to their lack of trainable parameters. In this paper, we begin by unifying local outlier methods by showing that they are particular cases of the more general message passing framework used in graph neural networks. This allows us to introduce learnability into local outlier methods, in the form of a neural network, for greater flexibility and expressivity: specifically, we propose LUNAR, a novel, graph neural network-based anomaly detection method. LUNAR learns to use information from the nearest neighbours of each node in a trainable way to find anomalies. We show that our method performs significantly better than existing local outlier methods, as well as state-of-the-art deep baselines. We also show that the performance of our method is much more robust to different settings of the local neighbourhood size.

Classification tasks on labeled graph-structured data have many important applications ranging from social recommendation to financial modeling. Deep neural networks are increasingly being used for node classification on graphs, wherein nodes with similar features have to be given the same label. Graph convolutional networks (GCNs) are one such widely studied neural network architecture that perform well on this task. However, powerful link-stealing attacks on GCNs have recently shown that even with black-box access to the trained model, inferring which links (or edges) are present in the training graph is practical. In this paper, we present a new neural network architecture called LPGNet for training on graphs with privacy-sensitive edges. LPGNet provides differential privacy (DP) guarantees for edges using a novel design for how graph edge structure is used during training. We empirically show that LPGNet models often lie in the sweet spot between providing privacy and utility: They can offer better utility than "trivially" private architectures which use no edge information (e.g., vanilla MLPs) and better resilience against existing link-stealing attacks than vanilla GCNs which use the full edge structure. LPGNet also offers consistently better privacy-utility tradeoffs than DPGCN, which is the state-of-the-art mechanism for retrofitting differential privacy into conventional GCNs, in most of our evaluated datasets.

Deep learning models are modern tools for spatio-temporal graph (STG) forecasting. Though successful, we argue that data scarcity is a key factor limiting their recent improvements. Meanwhile, contrastive learning has been an effective method for providing self-supervision signals and addressing data scarcity in various domains. In view of this, one may ask: can we leverage the additional signals from contrastive learning to alleviate data scarcity, so as to benefit STG forecasting? To answer this question, we present the first systematic exploration on incorporating contrastive learning into STG forecasting. Specifically, we first elaborate two potential schemes for integrating contrastive learning. We then propose two feasible and efficient designs of contrastive tasks that are performed on the node or graph level. The empirical study on STG benchmarks demonstrates that integrating graph-level contrast with the joint learning scheme achieves the best performance. In addition, we introduce four augmentations for STG data, which perturb the data in terms of graph structure, time domain, and frequency domain. Experimental results reveal that the model is not sensitive to the proposed augmentations' semantics. Lastly, we extend the classic contrastive loss via a rule-based strategy that filters out the most semantically similar negatives, yielding performance gains. We also provide explanations and insights based on the above experimental findings. Code is available at https://github.com/liuxu77/STGCL.

The space of value functions is a fundamental concept in reinforcement learning. Characterizing its geometric properties may provide insights for optimization and representation. Existing works mainly focus on the value space for Markov Decision Processes (MDPs). In this paper, we study the geometry of the robust value space for the more general Robust MDPs (RMDPs) setting, where transition uncertainties are considered. Specifically, since we find it hard to directly adapt prior approaches to RMDPs, we start with revisiting the non-robust case, and introduce a new perspective that enables us to characterize both the non-robust and robust value space in a similar fashion. The key of this perspective is to decompose the value space, in a state-wise manner, into unions of hypersurfaces. Through our analysis, we show that the robust value space is determined by a set of conic hypersurfaces, each of which contains the robust values of all policies that agree on one state. Furthermore, we find that taking only extreme points in the uncertainty set is sufficient to determine the robust value space. Finally, we discuss some other aspects about the robust value space, including its non-convexity and policy agreement on multiple states.

The detection of anomalous samples in large, high-dimensional datasets is a challenging task with numerous practical applications. Recently, state-of-the-art performance is achieved with deep learning methods: for example, using the reconstruction error from an autoencoder as anomaly scores. However, the scores are uncalibrated: that is, they follow an unknown distribution and lack a clear interpretation. Furthermore, the reconstruction error is highly influenced by the `hardness' of a given sample, which leads to false negative and false positive errors. In this paper, we empirically show the significance of this hardness bias present in a range of recent deep anomaly detection methods. To mitigate this, we propose an efficient and plug-and-play error calibration method which mitigates this hardness bias in the anomaly scoring without the need to retrain the model. We verify the effectiveness of our method on a range of image, time-series, and tabular datasets and against several baseline methods.

Of particular interest is to discover useful representations solely from observations in an unsupervised generative manner. However, the question of whether existing normalizing flows provide effective representations for downstream tasks remains mostly unanswered despite their strong ability for sample generation and density estimation. This paper investigates this problem for such a family of generative models that admits exact invertibility. We propose Neural Principal Component Analysis (Neural-PCA) that operates in full dimensionality while capturing principal components in descending order. Without exploiting any label information, the principal components recovered store the most informative elements in their leading dimensions and leave the negligible in the trailing ones, allowing for clear performance improvements of 5%-10% in downstream tasks. Such improvements are empirically found consistent irrespective of the number of latent trailing dimensions dropped. Our work suggests that necessary inductive bias be introduced into generative modeling when representation quality is of interest.

We present a new neighbor sampling method on temporal graphs. In a temporal graph, predicting different nodes' time-varying properties can require the receptive neigh-borhood of various temporal scales. In this work, we propose the TNS (Time-aware Neighbor Sampling) method: TNS learns from temporal information to provide an adaptive receptive neighborhood for every node at any time. Learning how to sample neighbors is non-trivial, since the neighbor indices in time order are discrete and not differentiable. To address this challenge, we transform neighbor indices from discrete values to continuous ones by interpolating the neighbors' messages. TNS can be flexibly incorporated into popular temporal graph networks to improve their effectiveness without increasing their time complexity. TNS can be trained in an end-to-end manner. It requires no extra supervision and is automatically and implicitly guided to sample the neighbors that are most beneficial for prediction. Empirical results on multiple standard datasets show that TNS yields significant gains on edge prediction and node classification.

Yiwei Wang, Muhao Chen, Wenxuan Zhou, Yujun Cai, Yuxuan Liang, Dayiheng Liu, Baosong Yang, Juncheng Liu, Bryan Hooi. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2022.

Recently, implicit graph neural networks (GNNs) have been proposed to capture long-range dependencies in underlying graphs. In this paper, we introduce and justify two weaknesses of implicit GNNs: the constrained expressiveness due to their limited effective range for capturing long-range dependencies, and their lack of ability to capture multiscale information on graphs at multiple resolutions. To show the limited effective range of previous implicit GNNs, we first provide a theoretical analysis and point out the intrinsic relationship between the effective range and the convergence of iterative equations used in these models. To mitigate the mentioned weaknesses, we propose a multiscale graph neural network with implicit layers (MGNNI) which is able to model multiscale structures on graphs and has an expanded effective range for capturing long-range dependencies. We conduct comprehensive experiments for both node classification and graph classification to show that MGNNI outperforms representative baselines and has a better ability for multiscale modeling and capturing of long-range dependencies.

Existing long-tailed recognition methods, aiming to train class-balanced models from long-tailed data, generally assume the models would be evaluated on the uniform test class distribution. However, practical test class distributions often violate this assumption (e.g., being either long-tailed or even inversely long-tailed), which may lead existing methods to fail in real applications. In this paper, we study a more practical yet challenging task, called test-agnostic long-tailed recognition, where the training class distribution is long-tailed while the test class distribution is agnostic and not necessarily uniform. In addition to the issue of class imbalance, this task poses another challenge: the class distribution shift between the training and test data is unknown. To tackle this task, we propose a novel approach, called Self-supervised Aggregation of Diverse Experts, which consists of two strategies: (i) a new skill-diverse expert learning strategy that trains multiple experts from a single and stationary long-tailed dataset to separately handle different class distributions; (ii) a novel test-time expert aggregation strategy that leverages self-supervision to aggregate the learned multiple experts for handling unknown test class distributions. We theoretically show that our self-supervised strategy has a provable ability to simulate test-agnostic class distributions. Promising empirical results demonstrate the effectiveness of our method on both vanilla and test-agnostic long-tailed recognition. The source code is available at https://github.com/Vanint/SADE-AgnosticLT.

Learning which Point-of-Interest (POI) a user will visit next is a challenging task for personalized recommender systems due to the large search space of possible POIs in the region. A recurring problem among existing works that makes it difficult to learn and perform well is the sparsity of the User-POI matrix. In this paper, we propose our Hierarchical Multi-Task Graph Recurrent Network (HMT-GRN) approach, which alleviates the data sparsity problem by learning different User-Region matrices of lower sparsities in a multi-task setting. We then perform a Hierarchical Beam Search (HBS) on the different region and POI distributions to hierarchically reduce the search space with increasing spatial granularity and predict the next POI. Our HBS provides efficiency gains by reducing the search space, resulting in speedups of 5 to 7 times over an exhaustive approach. In addition, we also propose a novel selectivity layer to predict if the next POI has been visited before by the user to balance between personalization and exploration. Experimental results on two real-world Location-Based Social Network (LBSN) datasets show that our model significantly outperforms baseline and the state-of-the-art methods.

Given a stream of money transactions between accounts in a bank, how can we accurately detect money laundering agent accounts and suspected behaviors in real-time? Money laundering agents try to hide the origin of illegally obtained money by dispersive multiple small transactions and evade detection by smart strategies. Therefore, it is challenging to accurately catch such fraudsters in an unsupervised manner. Existing approaches do not consider the characteristics of those agent accounts and are not suitable to the streaming settings. Therefore, we propose MonLAD and MonLAD-W to detect money laundering agent accounts in a transaction stream by keeping track of their residuals and other features; we devise AnoScore algorithm to find anomalies based on the robust measure of statistical deviation. Experimental results show that MonLAD outperforms the state-of-the-art baselines on real-world data and finds various suspicious behavior patterns of money laundering. Additionally, several detected suspected accounts have been manually-verified as agents in real money laundering scenario.

Given a stream of entries over time in a multi-dimensional data setting where concept drift is present, how can we detect anomalous activities? Most of the existing unsupervised anomaly detection approaches seek to detect anomalous events in an offline fashion and require a large amount of data for training. This is not practical in real-life scenarios where we receive the data in a streaming manner and do not know the size of the stream beforehand. Thus, we need a data-efficient method that can detect and adapt to changing data trends, or concept drift, in an online manner. In this work, we propose MemStream, a streaming anomaly detection framework, allowing us to detect unusual events as they occur while being resilient to concept drift. We leverage the power of a denoising autoencoder to learn representations and a memory module to learn the dynamically changing trend in data without the need for labels. We prove the optimum memory size required for effective drift handling. Furthermore, MemStream makes use of two architecture design choices to be robust to memory poisoning. Experimental results show the effectiveness of our approach compared to state-of-the-art streaming baselines using 2 synthetic datasets and 11 real-world datasets.

Real-time fraud detection is a challenge for most financial and electronic commercial platforms. To identify fraudulent communities, Grab, one of the largest technology companies in Southeast Asia, forms a graph from a set of transactions and detects dense subgraphs arising from abnormally large numbers of connections among fraudsters. Existing dense subgraph detection approaches focus on static graphs without considering the fact that transaction graphs are highly dynamic. Moreover, detecting dense subgraphs from scratch with graph updates is time consuming and cannot meet the real-time requirement in industry. Therefore, we introduce an incremental real-time fraud detection framework called Spade. Spade can detect fraudulent communities in hundreds of microseconds on million-scale graphs by incrementally maintaining dense subgraphs. Furthermore, Spade supports batch updates and edge grouping to reduce response latency. Lastly, Spade provides simple but expressive APIs for the design of evolving fraud detection semantics. Developers plug their customized suspiciousness functions into Spade which incrementalizes their semantics without recasting their algorithms. Extensive experiments show that Spade detects fraudulent communities in real time on million-scale graphs. Peeling algorithms incrementalized by Spade are up to a million times faster than the static version.

Given a stream of graph edges from a dynamic graph, how can we assign anomaly scores to edges in an online manner, for the purpose of detecting unusual behavior, using constant time and memory? Existing approaches aim to detect individually surprising edges. In this work, we propose Midas , which focuses on detecting microcluster anomalies , or suddenly arriving groups of suspiciously similar edges, such as lockstep behavior, including denial of service attacks in network traffic data. We further propose Midas -F, to solve the problem by which anomalies are incorporated into the algorithm’s internal states, creating a “poisoning” effect that can allow future anomalies to slip through undetected. Midas -F introduces two modifications: (1) we modify the anomaly scoring function, aiming to reduce the “poisoning” effect of newly arriving edges; (2) we introduce a conditional merge step, which updates the algorithm’s data structures after each time tick, but only if the anomaly score is below a threshold value, also to reduce the “poisoning” effect. Experiments show that Midas -F has significantly higher accuracy than Midas . In general, the algorithms proposed in this work have the following properties: (a) they detects microcluster anomalies while providing theoretical guarantees about the false positive probability; (b) they are online, thus processing each edge in constant time and constant memory, and also processes the data orders-of-magnitude faster than state-of-the-art approaches; and (c) they provides up to 62% higher area under the receiver operating characteristic curve than state-of-the-art approaches.

Mitigating the risk arising from extreme events is a fundamental goal with many applications, such as the modelling of natural disasters, financial crashes, epidemics, and many others. To manage this risk, a vital step is to be able to understand or generate a wide range of extreme scenarios. Existing approaches based on Generative Adversarial Networks (GANs) excel at generating realistic samples, but seek to generate typical samples, rather than extreme samples. Hence, in this work, we propose ExGAN, a GAN-based approach to generate realistic and extreme samples. To model the extremes of the training distribution in a principled way, our work draws from Extreme Value Theory (EVT), a probabilistic approach for modelling the extreme tails of distributions. For practical utility, our framework allows the user to specify both the desired extremeness measure, as well as the desired extremeness probability they wish to sample at. Experiments on real US Precipitation data show that our method generates realistic samples, based on visual inspection and quantitative measures, in an efficient manner. Moreover, generating increasingly extreme examples using ExGAN can be done in constant time (with respect to the extremeness probability τ), as opposed to the O(1/τ) time required by the baseline approach.

Given high-dimensional time series data (e.g., sensor data), how can we detect anomalous events, such as system faults and attacks? More challengingly, how can we do this in a way that captures complex inter-sensor relationships, and detects and explains anomalies which deviate from these relationships? Recently, deep learning approaches have enabled improvements in anomaly detection in high-dimensional datasets; however, existing methods do not explicitly learn the structure of existing relationships between variables, or use them to predict the expected behavior of time series. Our approach combines a structure learning approach with graph neural networks, additionally using attention weights to provide explainability for the detected anomalies. Experiments on two real-world sensor datasets with ground truth anomalies show that our method detects anomalies more accurately than baseline approaches, accurately captures correlations between sensors, and allows users to deduce the root cause of a detected anomaly.

A Graph Convolutional Network (GCN) stacks several layers and in each layer performs a PROPagation operation~(PROP) and a TRANsformation operation~(TRAN) for learning node representations over graph-structured data. Though powerful, GCNs tend to suffer performance drop when the model gets deep. Previous works focus on PROPs to study and mitigate this issue, but the role of TRANs is barely investigated. In this work, we study performance degradation of GCNs by experimentally examining how stacking only TRANs or PROPs works. We find that TRANs contribute significantly, or even more than PROPs, to declining performance, and moreover that they tend to amplify node-wise feature variance in GCNs, causing variance inflammation that we identify as a key factor for causing performance drop. Motivated by such observations, we propose a variance-controlling technique termed Node Normalization (NodeNorm), which scales each node's features using its own standard deviation. Experimental results validate the effectiveness of NodeNorm on addressing performance degradation of GCNs. Specifically, it enables deep GCNs to outperform shallow ones in cases where deep models are needed, and to achieve comparable results with shallow ones on 6 benchmark datasets. NodeNorm is a generic plug-in and can well generalize to other GNN architectures. Code is publicly available at https://github.com/miafei/NodeNorm.

An emerging line of research has found that spherical spaces better match the underlying geometry of facial im-ages, as evidenced by the state-of-the-art facial recognition methods which benefit empirically from spherical representations. Yet, these approaches rely on deterministic embeddings and hence suffer from the feature ambiguity dilemma, whereby ambiguous or noisy images are mapped into poorly learned regions of representation space, leading to inaccuracies. Probabilistic Face Embeddings (PFE) [17] is the first attempt to address this dilemma. However, we theoretically and empirically identify two main failures of PFE when it is applied to spherical deterministic embeddings aforementioned. To address these issues, in this paper, we propose a novel framework for face confidence learning in spherical space. Mathematically, we extend the von Mises Fisher density to its r-radius counterpart and derive a new optimization objective in closed form. Theoretically, the proposed probabilistic framework provably allows for better interpretability, leading to principled feature comparison and pooling. Extensive experimental results on multiple challenging benchmarks confirm our hypothesis and theory, and showcase the advantages of our framework over prior probabilistic methods and spherical deterministic embed-dings in various face recognition tasks.

The Laplacian representation recently gains increasing attention for reinforcement learning as it provides succinct and informative representation for states, by taking the eigenvectors of the Laplacian matrix of the state-transition graph as state embeddings. Such representation captures the geometry of the underlying state space and is beneficial to RL tasks such as option discovery and reward shaping. To approximate the Laplacian representation in large (or even continuous) state spaces, recent works propose to minimize a spectral graph drawing objective, which however has infinitely many global minimizers other than the eigenvectors. As a result, their learned Laplacian representation may differ from the ground truth. To solve this problem, we reformulate the graph drawing objective into a generalized form and derive a new learning objective, which is proved to have eigenvectors as its unique global minimizer. It enables learning high-quality Laplacian representations that faithfully approximate the ground truth. We validate this via comprehensive experiments on a set of gridworld and continuous control environments. Moreover, we show that our learned Laplacian representations lead to more exploratory options and better reward shaping.

We propose to organize the 6th ODD workshop at KDD 2021, following the successful series of the past five ODD Workshops that have been organized at KDD 2013, KDD 2014, KDD 2015, KDD 2016, and KDD 2018.

Given taxi-ride counts information between departure and destination locations, how can we forecast their future demands? In general, given a data stream of events with seasonal patterns that innovate over time, how can we effectively and efficiently forecast future events? In this paper, we propose Shifting Seasonal Matrix Factorization approach, namely SSMF, that can adaptively learn multiple seasonal patterns (called regimes), as well as switching between them. Our proposed method has the following properties: (a) it accurately forecasts future events by detecting regime shifts in seasonal patterns as the data stream evolves; (b) it works in an online setting, i.e., processes each observation in constant time and memory; (c) it effectively realizes regime shifts without human intervention by using a lossless data compression scheme. We demonstrate that our algorithm outperforms state-of-the-art baseline methods by accurately forecasting upcoming events on three real-world data streams.

Graph neural networks (GNNs) are widely used for modelling graph-structured data in numerous applications. However, with their inherently finite aggregation layers, existing GNN models may not be able to effectively capture long-range dependencies in the underlying graphs. Motivated by this limitation, we propose a GNN model with infinite depth, which we call Efficient Infinite-Depth Graph Neural Networks (EIGNN), to efficiently capture very long-range dependencies. We theoretically derive a closed-form solution of EIGNN which makes training an infinite-depth GNN model tractable. We then further show that we can achieve more efficient computation for training EIGNN by using eigendecomposition. The empirical results of comprehensive experiments on synthetic and real-world datasets show that EIGNN has a better ability to capture long-range dependencies than recent baselines, and consistently achieves state-of-the-art performance. Furthermore, we show that our model is also more robust against both noise and adversarial perturbations on node features.

Temporal Graph Networks (TGNs) are powerful on modeling temporal graph data based on their increased complexity. Higher complexity carries with it a higher risk of overfitting, which makes TGNs capture random noise instead of essential semantic information. To address this issue, our idea is to transform the temporal graphs using data augmentation (DA) with adaptive magnitudes, so as to effectively augment the input features and preserve the essential semantic information. Based on this idea, we present the MeTA (Memory Tower Augmentation) module: a multi-level module that processes the augmented graphs of different magnitudes on separate levels, and performs message passing across levels to provide adaptively augmented inputs for every prediction. MeTA can be flexibly applied to the training of popular TGNs to improve their effectiveness without increasing their time complexity. To complement MeTA, we propose three DA strategies to realistically model noise by modifying both the temporal and topological features. Empirical results on standard datasets show that MeTA yields significant gains for the popular TGN models on edge prediction and node classification in an efficient manner.

Contrastive self-supervised learning (CSL) has attracted increasing attention for model pre-training via unlabeled data. The resulted CSL models provide instance-discriminative visual features that are uniformly scattered in the feature space. During deployment, the common practice is to directly fine-tune CSL models with cross-entropy, which however may not be the best strategy in practice. Although cross-entropy tends to separate inter-class features, the resulting models still have limited capability for reducing intra-class feature scattering that exists in CSL models. In this paper, we investigate whether applying contrastive learning to fine-tuning would bring further benefits, and analytically find that optimizing the contrastive loss benefits both discriminative representation learning and model optimization during fine-tuning. Inspired by these findings, we propose Contrast-regularized tuning (Core-tuning), a new approach for fine-tuning CSL models. Instead of simply adding the contrastive loss to the objective of fine-tuning, Core-tuning further applies a novel hard pair mining strategy for more effective contrastive fine-tuning, as well as smoothing the decision boundary to better exploit the learned discriminative feature space. Extensive experiments on image classification and semantic segmentation verify the effectiveness of Core-tuning.

We study the hop-constrained s-t path enumeration (HcPE ) problem, which takes a graph G, two distinct vertices s,t and a hop constraint k as input, and outputs all paths from s to t whose length is at most k. The state-of-the-art algorithms suffer from severe performance issues caused by the costly pruning operations during enumeration for the workloads with the large search space. Consequently, these algorithms hardly meet the real-time constraints of many online applications. In this paper, we propose PathEnum, an efficient index-based algorithm towards real-time HcPE. For an input query, PathEnum first builds a light-weight index aiming to reduce the number of edges involved in the enumeration, and develops efficient index-based approaches for enumeration, one based on depth-first search and the other based on joins. We further develop a query optimizer based on a join-based cost model to optimize the search order. We conduct experiments with 15 real-world graphs. Our experiment results show that PathEnum outperforms the state-of-the-art approaches by orders of magnitude in terms of the query time, throughput and response time.

Next destination recommendation is an important task in the transportation domain of taxi and ride-hailing services, where users are recommended with personalized destinations given their current origin location. However, recent recommendation works do not satisfy this origin-awareness property, and only consider learning from historical destination locations, without origin information. Thus, the resulting approaches are unable to learn and predict origin-aware recommendations based on the user's current location, leading to sub-optimal performance and poor real-world practicality. Hence, in this work, we study the origin-aware next destination recommendation task. We propose the Spatial-Temporal Origin-Destination Personalized Preference Attention (STOD-PPA) encoder-decoder model to learn origin-origin (OO), destination-destination (DD), and origin-destination (OD) relationships by first encoding both origin and destination sequences with spatial and temporal factors in local and global views, then decoding them through personalized preference attention to predict the next destination. Experimental results on seven real-world user trajectory taxi datasets show that our model significantly outperforms baseline and state-of-the-art methods.

Given a stream of entries in a multi-aspect data setting i.e., entries having multiple dimensions, how can we detect anomalous activities in an unsupervised manner? For example, in the intrusion detection setting, existing work seeks to detect anomalous events or edges in dynamic graph streams, but this does not allow us to take into account additional attributes of each entry. Our work aims to define a streaming multi-aspect data anomaly detection framework, termed MStream which can detect unusual group anomalies as they occur, in a dynamic manner. MStream has the following properties: (a) it detects anomalies in multi-aspect data including both categorical and numeric attributes; (b) it is online, thus processing each record in constant time and constant memory; (c) it can capture the correlation between multiple aspects of the data. MStream is evaluated over the KDDCUP99, CICIDS-DoS, UNSW-NB 15 and CICIDS-DDoS datasets, and outperforms state-of-the-art baselines.

Graph neural networks (GNNs) have achieved state-of-the-art performance on graph classification tasks. Existing work usually feeds graphs to GNNs in random order for training. However, graphs can vary greatly in their difficulty for classification, and we argue that GNNs can benefit from an easy-to-difficult curriculum, similar to the learning process of humans. Evaluating the difficulty of graphs is challenging due to the high irregularity of graph data. To address this issue, we present the CurGraph (Curriculum Learning for Graph Classification) framework, that analyzes the graph difficulty in the high-level semantic feature space. Specifically, we use the infomax method to obtain graph-level embeddings and a neural density estimator to model the embedding distributions. Then we calculate the difficulty scores of graphs based on the intra-class and inter-class distributions of their embeddings. Given the difficulty scores, CurGraph first exposes a GNN to easy graphs, before gradually moving on to hard ones. To provide a soft transition from easy to hard, we propose a smooth-step method, which utilizes a time-variant smooth function to filter out hard graphs. Thanks to CurGraph, a GNN learns from the graphs at the border of its capability, neither too easy or too hard, to gradually expand its border at each training step. Empirically, CurGraph yields significant gains for popular GNN models on graph classification and enables them to achieve superior performance on miscellaneous graphs.

Mixup is an advanced data augmentation method for training neural network based image classifiers, which interpolates both features and labels of a pair of images to produce synthetic samples. However, devising the Mixup methods for graph learning is challenging due to the irregularity and connectivity of graph data. In this paper, we propose the Mixup methods for two fundamental tasks in graph learning: node and graph classification. To interpolate the irregular graph topology, we propose the two-branch graph convolution to mix the receptive field subgraphs for the paired nodes. Mixup on different node pairs can interfere with the mixed features for each other due to the connectivity between nodes. To block this interference, we propose the two-stage Mixup framework, which uses each node’s neighbors’ representations before Mixup for graph convolutions. For graph classification, we interpolate complex and diverse graphs in the semantic space. Qualitatively, our Mixup methods enable GNNs to learn more discriminative features and reduce over-fitting. Quantitative results show that our method yields consistent gains in terms of test accuracy and F1-micro scores on standard datasets, for both node and graph classification. Overall, our method effectively regularizes popular graph neural networks for better generalization without increasing their time complexity.

Given a stream of graph edges from a dynamic graph, how can we assign anomaly scores to edges in an online manner, for the purpose of detecting unusual behavior, using constant time and memory? Existing approaches aim to detect individually surprising edges. In this work, we propose Midas, which focuses on detecting microcluster anomalies, or suddenly arriving groups of suspiciously similar edges, such as lockstep behavior, including denial of service attacks in network traffic data. Midas has the following properties: (a) it detects microcluster anomalies while providing theoretical guarantees about its false positive probability; (b) it is online, thus processing each edge in constant time and constant memory, and also processes the data 108–505 times faster than state-of-the-art approaches; (c) it provides 46%-52% higher accuracy (in terms of AUC) than state-of-the-art approaches.

Suppose you visit an e-commerce site, and see that 50 users each reviewed almost all of the same 500 products several times each: would you get suspicious? Similarly, given a Twitter follow graph, how can we design principled measures for identifying surprisingly dense subgraphs? Dense subgraphs often indicate interesting structure, such as network attacks in network traffic graphs. However, most existing dense subgraph measures either do not model normal variation, or model it using an Erdős-Renyi assumption - but this assumption has been discredited decades ago. What is the right assumption then? We propose a novel application of extreme value theory to the dense subgraph problem, which allows us to propose measures and algorithms which evaluate the surprisingness of a subgraph probabilistically, without requiring restrictive assumptions (e.g. Erdős-Renyi). We then improve the practicality of our approach by incorporating empirical observations about dense subgraph patterns in real graphs, and by proposing a fast pruning-based search algorithm. Our approach (a) provides theoretical guarantees of consistency, (b) scales quasi-linearly, and (c) outperforms baselines in synthetic and ground truth settings.

Given a graph of the money transfers between accounts of a bank, how can we detect money laundering? Money laundering refers to criminals using the bank's services to move massive amounts of illegal money to untraceable destination accounts, in order to inject their illegal money into the legitimate financial system. Existing graph fraud detection approaches focus on dense subgraph detection, without considering the fact that money laundering involves high-volume flows of funds through chains of bank accounts, thereby decreasing their detection accuracy. Instead, we propose to model the transactions using a multipartite graph, and detect the complete flow of money from source to destination using a scalable algorithm, FlowScope. Theoretical analysis shows that FlowScope provides guarantees in terms of the amount of money that fraudsters can transfer without being detected. FlowScope outperforms state-of-the-art baselines in accurately detecting the accounts involved in money laundering, in both injected and real-world data settings.

Next Point-of-Interest (POI) recommendation is a longstanding problem across the domains of Location-Based Social Networks (LBSN) and transportation. Recent Recurrent Neural Network (RNN) based approaches learn POI-POI relationships in a local view based on independent user visit sequences. This limits the model's ability to directly connect and learn across users in a global view to recommend semantically trained POIs. In this work, we propose a Spatial-Temporal-Preference User Dimensional Graph Attention Network (STP-UDGAT), a novel explore-exploit model that concurrently exploits personalized user preferences and explores new POIs in global spatial-temporal-preference (STP) neighbourhoods, while allowing users to selectively learn from other users. In addition, we propose random walks as a masked self-attention option to leverage the STP graphs' structures and find new higher-order POI neighbours during exploration. Experimental results on six real-world datasets show that our model significantly outperforms baseline and state-of-the-art methods.

Graph Convolutional Networks (GCNs) have achieved state-of-the-art performance on node classification. However, recent works have shown that GCNs are vulnerable to adversarial attacks, such as additions or deletions of adversarially-chosen edges in the graph, in order to mislead the node classification algorithms. How can we design robust GCNs that are resistant to such adversarial attacks? More challengingly, how can we do this in a way that is provably robust? We propose a robust node classification approach based on a low-pass `message passing' mechanism, that (a) reduces the effectiveness of adversarial attacks in experiments, and (b) provides theoretical guarantees against adversarial attacks. Our approach can be embedded into the existing GCN architectures to enhance their robustness. Empirical results show that our loss-pass method effectively improves the performance of multiple GCNs under miscellaneous perturbations and helps them to achieve superior performance on various graphs.

Graph data is ubiquitous in academia and industry, from social networks to bioinformatics. The pervasiveness of graphs today has raised the demand for algorithms that can answer various questions: Which products would a user like to purchase given her order list? Which users are buying fake followers to increase their public reputation? Myriads of new graph mining algorithms are proposed every year to answer such questions - each with a distinct problem formulation, computational time, and memory footprint. This lack of unity makes it difficult for a practitioner to compare different algorithms and pick the most suitable one for a specific application. These challenges - even more severe for non-experts - create a gap in which state-of-the-art techniques developed in academic settings fail to be optimally deployed in real-world applications. To bridge this gap, we propose AutoGM, an automated system for graph mining algorithm development. We first define a unified framework UnifiedGM that integrates various message-passing based graph algorithms, ranging from conventional algorithms like PageRank to graph neural networks. Then UnifiedGM defines a search space in which five parameters are required to determine a graph algorithm. Under this search space, AutoGM explicitly optimizes for the optimal parameter set of UnifiedGM using Bayesian Optimization. AutoGM defines a novel budget-aware objective function for the optimization to incorporate a practical issue - finding the best speed-accuracy trade-off under a computation budget - into the graph algorithm generation problem. Experiments on real-world benchmark datasets demonstrate that AutoGM generates novel graph mining algorithms with the best speed/accuracy trade-off compared to existing models with heuristic parameters.

Detecting anomalies is an important task in a wide variety of applications and domains. Deep learning methods have achieved state-of-the-art performance in anomaly detection in recent years; unsupervised methods being particularly popular. However, deep learning methods can be fragile to small perturbations in the input data. This can be exploited by an adversary to deliberately hinder model performance; an adversarial attack. This phenomena has been widely studied in the context of supervised image classification since its discovery, however such studies for an anomaly detection setting are sorely lacking. Moreover, the plethora of defense mechanisms that have been proposed are often not applicable to unsupervised anomaly detection models. In this work, we study the effect of adversarial attacks on the performance of anomaly-detecting autoencoders using real data from a Cyber physical system (CPS) testbed with intervals of controlled, physical attacks as anomalies. An adversary would attempt to disguise these points as normal through adversarial perturbations. To combat this, we propose the Approximate Projection Autoencoder (APAE), which incorporates two defenses against such attacks into a general autoencoder. One of these involves a novel technique to improve robustness under adversarial impact by optimising latent representations for better reconstruction outputs.

Graph convolutional networks (GCNs) have achieved state-of-the-art performance on the task of node classification. However, the performance of GCNs is prone to implementation bugs that do not explicitly produce compile-time or run-time errors but degrade their effectiveness heavily. These bugs are hard to detect, since the way in which the node attributes and graph structures contribute to the outputs is complicated, non-transparent, and not traceable by humans. To address this issue, we propose a systematic approach with formal justifications to detect implementation bugs in GCN based node classifiers. Our approach is based on the idea of Metamorphic Testing, which does not check input-output relations for a single input, but for input-output pairs. To speed up our approach, we design a pipeline system, which synchronizes the workload on CPUs and GPUs adaptively and processes them simultaneously. Our empirical study shows that our approach is able to identify over 80% of the synthetic mutants and two real-world bugs in GCN implementations. In addition, our pipeline system can achieve more than 10× speedup over the sequential system that leaves the CPU/GPU idle when using the other.

Graphs have been utilized as a powerful tool to model pairwise relationships between people or objects. Such structure is a special type of a broader concept referred to as hypergraph, in which each hyperedge may consist of an arbitrary number of nodes, rather than just two. A large number of real-world datasets are of this form - for example, lists of recipients of emails sent from an organization, users participating in a discussion thread or subject labels tagged in an online question. However, due to complex representations and lack of adequate tools, little attention has been paid to exploring the underlying patterns in these interactions.

By using Data Augmentation (DA), we present a new method to enhance Graph Convolutional Networks (GCNs), that are the state-of-the-art models for semi-supervised node classification. DA for graph data remains under-explored. Due to the connections built by edges, DA for different nodes influence each other and lead to undesired results, such as uncontrollable DA magnitudes and changes of ground-truth labels. To address this issue, we present the NodeAug (Node-Parallel Augmentation) scheme, that creates a 'parallel universe' for each node to conduct DA, to block the undesired effects from other nodes. NodeAug regularizes the model prediction of every node (including unlabeled) to be invariant with respect to changes induced by Data Augmentation (DA), so as to improve the effectiveness. To augment the input features from different aspects, we propose three DA strategies by modifying both node attributes and the graph structure. In addition, we introduce the subgraph mini-batch training for the efficient implementation of NodeAug. The approach takes the subgraph corresponding to the receptive fields of a batch of nodes as the input per iteration, rather than the whole graph that the prior full-batch training takes. Empirically, NodeAug yields significant gains for strong GCN models on the Cora, Citeseer, Pubmed, and two co-authorship networks, with a more efficient training process thanks to the proposed subgraph mini-batch training approach.

Given a stream of edge additions and deletions, how can we estimate the count of triangles in it? If we can store only a subset of the edges, how can we obtain unbiased estimates with small variances? Counting triangles (i.e., cliques of size three) in a graph is a classical problem with applications in a wide range of research areas, including social network analysis, data mining, and databases. Recently, streaming algorithms for triangle counting have been extensively studied since they can naturally be used for large dynamic graphs. However, existing algorithms cannot handle edge deletions or suffer from low accuracy. Can we handle edge deletions while achieving high accuracy? We propose T hink D, which accurately estimates the counts of global triangles (i.e., all triangles) and local triangles associated with each node in a fully dynamic graph stream with additions and deletions of edges. Compared to its best competitors, T hink D is (a) Accurate: up to 4.3 × more accurate within the same memory budget, (b) Fast: up to 2.2 × faster for the same accuracy requirements, and (c) Theoretically sound: always maintaining estimates with zero bias (i.e., the difference between the true triangle count and the expected value of its estimate) and small variance. As an application, we use T hink D to detect suddenly emerging dense subgraphs, and we show its advantages over state-of-the-art methods.

Given a large-scale rhythmic time series containing mostly normal data segments (or `beats'), can we learn how to detect anomalous beats in an effective yet efficient way? For example, how can we detect anomalous beats from electrocardiogram (ECG) readings? Existing approaches either require excessively high amounts of labeled and balanced data for classification, or rely on less regularized reconstructions, resulting in lower accuracy in anomaly detection. Therefore, we propose BeatGAN, an unsupervised anomaly detection algorithm for time series data. BeatGAN outputs explainable results to pinpoint the anomalous time ticks of an input beat, by comparing them to adversarially generated beats. Its robustness is guaranteed by its regularization of reconstruction error using an adversarial generation approach, as well as data augmentation using time series warping. Experiments show that BeatGAN accurately and efficiently detects anomalous beats in ECG time series, and routes doctors' attention to anomalous time ticks, achieving accuracy of nearly 0.95 AUC, and very fast inference (2.6 ms per beat). In addition, we show that BeatGAN accurately detects unusual motions from multivariate motion-capture time series data, illustrating its generality.

Given a dynamic graph stream, how can we detect the sudden appearance of anomalous patterns, such as link spam, follower boosting, or denial of service attacks? Additionally, can we categorize the types of anomalies that occur in practice, and theoretically analyze the anomalous signs arising from each type? In this work, we propose AnomRank, an online algorithm for anomaly detection in dynamic graphs. AnomRank uses a two-pronged approach defining two novel metrics for anomalousness. Each metric tracks the derivatives of its own version of a 'node score' (or node importance) function. This allows us to detect sudden changes in the importance of any node. We show theoretically and experimentally that the two-pronged approach successfully detects two common types of anomalies: sudden weight changes along an edge, and sudden structural changes to the graph. AnomRank is (a) Fast and Accurate: up to 49.5x faster or 35% more accurate than state-of-the-art methods, (b) Scalable: linear in the number of edges in the input graph, processing millions of edges within 2 seconds on a stock laptop/desktop, and (c) Theoretically Sound: providing theoretical guarantees of the two-pronged approach.

Given multivariate time series data, how do we detect changes in the behavior of the time series: for example, the onset of illnesses or complications in patients? Can we do this without making strong assumptions about the data? We propose BnB (Branch and Border), an online, nonparametric change detection method that detects multiple changes in multivariate data. Unlike existing methods, BnB approaches change detection by separating points before and after the change using an ensemble of random partitions. BnB is (a) scalable: it scales linearly in the number of time ticks and dimensions, and is online, thus using bounded memory and bounded time per iteration; (b) effective: providing theoretical guarantees on the false positive rate, and achieving 70% or more increased F-measure over baselines in experiments averaged over 11 datasets; (c) general: it is nonparametric, and works on mixed data, including numerical, categorical, and ordinal data.

Consider a stream of time-stamped events, such as taxi rides, where we record the start and end locations of each ride. How do we learn a matrix factorization model which takes into account seasonal patterns (such as: rides toward office areas occur more frequently in the morning), and use it to forecast taxi rides tomorrow? Also, how can we model drift (such as population growth), and detect sudden changes (or anomalies)? Existing matrix factorization algorithms do not take seasonal patterns into account. We propose SMF (Seasonal Matrix Factorization), a matrix factorization model for seasonal data, and a streaming algorithm for fitting it. SMF is (a) accurate in forecasting: outperforming baselines by 13% to 60% in RMSE; (b) online: requiring fixed memory even as more data is received over time, and scaling linearly; (c) effective: providing interpretable results. In addition, we propose SMF-A, an algorithm which detects anomalies in a computationally feasible way, without forecasting every observation in the matrix.

How can we detect fraud in a big graph with rich properties, as online fraudsters invest more resources, including purchasing large pools of fake user accounts and dedicated IPs, to hide their fraudulent attacks? To achieve robustness, existing approaches detected dense sub-graphs as suspicious patterns in an unsupervised way, such as average degree maximization. However, such approaches suffer from the bias of including more nodes than necessary, resulting in lower accuracy and increased need for manual verification. Therefore, we propose HoloScope, which introduces a novel metric “contrast suspiciousness” integrating information from graph topology and spikes to more accurately detect fraudulent users and objects. Contrast suspiciousness dynamically emphasizes the contrasting patterns between fraudsters and normal users, making HoloScope capable of distinguishing the synchronized and strange behaviors of fraudsters by means of topology, bursts and drops, and rating scores. In addition, we provide theoretical bounds for how much this method increases the time cost needed for fraudsters to conduct adversarial attacks. Moreover, HoloScope has a concise framework and sub-quadratic time complexity, making the algorithm reproducible and scalable. In extensive experiments, HoloScope achieved significant accuracy improvements on real data with injected labels and true labels, when compared with state-of-the-art fraud detection methods.

Given electrical sensors placed on the power grid, how can we automatically determine when electrical components (e.g. power lines) fail? Or, given traffic sensors which measure the speed of vehicles passing over them, how can we determine when traffic accidents occur? Both these problems involve detecting change points in a set of sensors on the nodes or edges of a graph. To this end, we propose ChangeDAR (Change Detection And Resolution), which detects changes in an online manner, and reports when and where the change occurred in the graph.

In the smart power grid, short-term load forecasting (STLF) is a crucial step in scheduling and planning for future load, so as to improve the reliability, cost, and emissions of the power grid. Different from traditional time series forecast, STLF is a more challenging task, because of the complex demand of active and reactive power from numerous categories of electrical loads and the effects of environment. Therefore, we propose NeuCast, a seasonal neural forecasting method, which dynamically models various loads as co-evolving time series in a hidden space, as well as extra weather conditions, in a neural network structure. NeuCast captures seasonality and patterns of the time series by integrating factor modeling and hidden state recognition. NeuCast can also detect anomalies and forecast under different temperature assumptions. Extensive experiments on 134 real-word datasets show the improvements of NeuCast over the stateof-the-art methods.

How can we efficiently forecast the power consumption of a location for the next few days? More challengingly, how can we forecast the power consumption if the temperature increases by 10° C, the number of appliances in the grid increase by 20%, and voltage levels increase by 5%? Such ‘what-if scenarios' are crucial for future planning, to ensure that the grid remains reliable even under extreme conditions. Our contributions are as follows: 1) Domain knowledge infusion: we propose a novel Temporal BIG model that extends the physics-based BIG model, allowing it to capture changes over time, trends, and seasonality, and temperature effects. 2) Forecasting: our StreamCast algorithm forecasts multiple steps ahead and outperforms baselines in accuracy. Our algorithm is online, requiring constant update time per new data point and bounded memory. 3) What-if scenarios and anomaly detection: our approach can handle scenarios in which the voltage levels, temperature, or number of appliances change. It also spots anomalies in real data, and provides confidence intervals for its forecasts, to assist in planning for various scenarios. Experimental results show that StreamCast has 27% lower forecasting error than baselines on real data, scales linearly, and runs in 4 minutes on a time sequence of 40 million points.

Rating platforms enable large-scale collection of user opinion about items(e.g., products or other users). However, untrustworthy users give fraudulent ratings for excessive monetary gains. In this paper, we present REV2, a system to identify such fraudulent users. We propose three interdependent intrinsic quality metrics---fairness of a user, reliability of a rating and goodness of a product. The fairness and reliability quantify the trustworthiness of a user and rating, respectively, and goodness quantifies the quality of a product. Intuitively, a user is fair if it provides reliable scores that are close to the goodness of products. We propose six axioms to establish the interdependency between the scores, and then, formulate a mutually recursive definition that satisfies these axioms. We extend the formulation to address cold start problem and incorporate behavior properties. We develop the REV2 algorithm to calculate these intrinsic quality scores for all users, ratings, and products. We show that this algorithm is guaranteed to converge and has linear time complexity. By conducting extensive experiments on five rating datasets, we show that REV2 outperforms nine existing algorithms in detecting fair and unfair users. We reported the 150 most unfair users in the Flipkart network to their review fraud investigators, and 127 users were identified as being fraudulent(84.6% accuracy). The REV2 algorithm is being deployed at Flipkart.

Given a large-scale and high-order tensor, how can we detect dense subtensors in it? Can we spot them in near-linear time but with quality guarantees? Extensive previous work has shown that dense subtensors, as well as dense subgraphs, indicate anomalous or fraudulent behavior (e.g., lockstep behavior in social networks). However, available algorithms for detecting dense subtensors are not satisfactory in terms of speed, accuracy, and flexibility. In this work, we propose two algorithms, called M-Z oom and M-B iz , for fast and accurate dense-subtensor detection with various density measures. M-Z oom gives a lower bound on the density of detected subtensors, while M-B iz guarantees the local optimality of detected subtensors. M-Z oom and M-B iz can be combined, giving the following advantages: (1) Scalable: scale near-linearly with all aspects of tensors and are up to 114× faster than state-of-the-art methods with similar accuracy, (2) Provably accurate : provide a guarantee on the lowest density and local optimality of the subtensors they find, (3) Flexible: support multi-subtensor detection and size bounds as well as diverse density measures, and (4) Effective: successfully detected edit wars and bot activities in Wikipedia, and spotted network attacks from a TCP dump with near-perfect accuracy (AUC = 0.98).

As online fraudsters invest more resources, including purchasing large pools of fake user accounts and dedicated IPs, fraudulent attacks become less obvious and their detection becomes increasingly challenging. Existing approaches such as average degree maximization suffer from the bias of including more nodes than necessary, resulting in lower accuracy and increased need for manual verification. Hence, we propose HoloScope, which introduces a novel metric "contrast suspiciousness" integrating information from graph topology and spikes to more accurately detect fraudulent users and objects. Contrast suspiciousness dynamically emphasizes the contrast patterns between fraudsters and normal users, making HoloScope capable of distinguishing the synchronized and anomalous behaviors of fraudsters on topology, bursts and drops, and rating scores. In addition, we provide theoretical bounds for how much this increases the time cost needed for fraudsters to conduct adversarial attacks. Moreover, HoloScope has a concise framework and sub-quadratic time complexity, making the algorithm reproducible and scalable. Extensive experiments showed that HoloScope achieved significant accuracy improvements on synthetic and real data, compared with state-of-the-art fraud detection methods.

Given a retinal image, can we automatically determine whether it is of high quality (suitable for medical diagnosis)? Can we also explain our decision, pinpointing the region or regions that led to our decision? Images from human retinas are vital for the diagnosis of multiple health issues, like hypertension, diabetes, and Alzheimer's; low quality images may force the patient to come back again for a second scanning, wasting time and possibly delaying treatment. However, existing retinal image quality assessment methods are either black boxes without explanations of the results or depend heavily on feature engineering or on complex and error-prone anatomical structures' segmentation. Therefore, we propose EyeQual, that solves exactly this problem. EyeQual is novel, fast for inference, accurate and explainable, pinpointing low-quality regions on the image. We evaluated EyeQual on two real datasets where it achieved 100% accuracy taking just 36 milliseconds for each image.

Consider a stream of retweet events - how can we spot fraudulent lock-step behavior in such multi-aspect data (i.e., tensors) evolving over time? Can we detect it in real time, with an accuracy guarantee? Past studies have shown that dense subtensors tend to indicate anomalous or even fraudulent behavior in many tensor data, including social media, Wikipedia, and TCP dumps. Thus, several algorithms have been proposed for detecting dense subtensors rapidly and accurately. However, existing algorithms assume that tensors are static, while many real-world tensors, including those mentioned above, evolve over time.

How can we detect fraudulent lockstep behavior in large-scale multi-aspect data (i.e., tensors)? Can we detect it when data are too large to fit in memory or even on a disk? Past studies have shown that dense blocks in real-world tensors (e.g., social media, Wikipedia, TCP dumps, etc.) signal anomalous or fraudulent behavior such as retweet boosting, bot activities, and network attacks. Thus, various approaches, including tensor decomposition and search, have been used for rapid and accurate dense-block detection in tensors. However, all such methods have low accuracy, or assume that tensors are small enough to fit in main memory, which is not true in many real-world applications such as social media and web.

Given a collection of seasonal time-series, how can we find regular (cyclic) patterns and outliers (i.e. rare events)? These two types of patterns are hidden and mixed in the time-varying activities. How can we robustly separate regular patterns and outliers, without requiring any prior information?

Given a bipartite graph of users and the products that they review, or followers and followees, how can we detect fake reviews or follows? Existing fraud detection methods (spectral, etc.) try to identify dense subgraphs of nodes that are sparsely connected to the remaining graph. Fraudsters can evade these methods using camouflage , by adding reviews or follows with honest targets so that they look “normal.” Even worse, some fraudsters use hijacked accounts from honest users, and then the camouflage is indeed organic. Our focus is to spot fraudsters in the presence of camouflage or hijacked accounts. We propose FRAUDAR, an algorithm that (a) is camouflage resistant, (b) provides upper bounds on the effectiveness of fraudsters, and (c) is effective in real-world data. Experimental results under various attacks show that FRAUDAR outperforms the top competitor in accuracy of detecting both camouflaged and non-camouflaged fraud. Additionally, in real-world experiments with a Twitter follower--followee graph of 1.47 billion edges, FRAUDAR successfully detected a subgraph of more than 4, 000 detected accounts, of which a majority had tweets showing that they used follower-buying services.

Given a bipartite graph of users and the products that they review, or followers and followees, how can we detect fake reviews or follows? Existing fraud detection methods (spectral, etc.) try to identify dense subgraphs of nodes that are sparsely connected to the remaining graph. Fraudsters can evade these methods using camouflage, by adding reviews or follows with honest targets so that they look "normal". Even worse, some fraudsters use hijacked accounts from honest users, and then the camouflage is indeed organic. Our focus is to spot fraudsters in the presence of camouflage or hijacked accounts. We propose FRAUDAR, an algorithm that (a) is camouflage-resistant, (b) provides upper bounds on the effectiveness of fraudsters, and (c) is effective in real-world data. Experimental results under various attacks show that FRAUDAR outperforms the top competitor in accuracy of detecting both camouflaged and non-camouflaged fraud. Additionally, in real-world experiments with a Twitter follower-followee graph of 1.47 billion edges, FRAUDAR successfully detected a subgraph of more than 4000 detected accounts, of which a majority had tweets showing that they used follower-buying services.

Review fraud is a pervasive problem in online commerce, in which fraudulent sellers write or purchase fake reviews to manipulate perception of their products and services. Fake reviews are often detected based on several signs, including 1) they occur in short bursts of time; 2) fraudulent user accounts have skewed rating distributions. However, these may both be true in any given dataset. Hence, in this paper, we propose an approach for detecting fraudulent reviews which combines these 2 approaches in a principled manner, allowing successful detection even when one of these signs is not present. To combine these 2 approaches, we formulate our Bayesian Inference for Rating Data (BIRD) model, a flexible Bayesian model of user rating behavior. Based on our model we formulate a likelihood-based suspiciousness metric, Normalized Expected Surprise Total (NEST). We propose a linear-time algorithm for performing Bayesian inference using our model and computing the metric. Experiments on real data show that BIRDNEST successfully spots review fraud in large, real-world graphs: the 50 most suspicious users of the Flipkart platform flagged by our algorithm were investigated and all identified as fraudulent by domain experts at Flipkart.

Many commercial products and academic research activities are embracing behavior analysis as a technique for improving detection of attacks of many sorts-from retweet boosting, hashtag hijacking to link advertising. Traditional approaches focus on detecting dense blocks in the adjacency matrix of graph data, and recently, the tensors of multimodal data. No method gives a principled way to score the suspiciousness of dense blocks with different numbers of modes and rank them to draw human attention accordingly. In this paper, we first give a list of axioms that any metric of suspiciousness should satisfy; we propose an intuitive, principled metric that satisfies the axioms, and is fast to compute; moreover, we propose CrossSpot, an algorithm to spot dense blocks that are worth inspecting, typically indicating fraud or some other noteworthy deviation from the usual, and sort them in the order of importance (“suspiciousness”). Finally, we apply CrossSpot to the real data, where it improves the F1 score over previous techniques by 68 percent and finds suspicious behavioral patterns in social datasets spanning 0.3 billion posts.

Which seems more suspicious: 5,000 tweets from 200 users on 5 IP addresses, or 10,000 tweets from 500 users on 500 IP addresses but all with the same trending topic and all in 10 minutes? The literature has many methods that try to find dense blocks in matrices, and, recently, tensors, but no method gives a principled way to score the suspiciouness of dense blocks with different numbers of modes and rank them to draw human attention accordingly. Dense blocks are worth inspecting, typically indicating fraud, emerging trends, or some other noteworthy deviation from the usual. Our main contribution is that we show how to unify these methods and how to give a principled answer to questions like the above. Specifically, (a) we give a list of axioms that any metric of suspicousness should satisfy, (b) we propose an intuitive, principled metric that satisfies the axioms, and is fast to compute, (c) we propose CROSSSPOT, an algorithm to spot dense regions, and sort them in importance ("suspiciousness") order. Finally, we apply CROSSSPOT to real data, where it improves the F1 score over previous techniques by 68% and finds retweet-boosting in a real social dataset spanning 0.3 billion posts.