The integration of Large Language Models (LLMs) into critical societal and scientific functions has intensified the urgent demand for transparency, reliability, and trust. While post-hoc attribution methods and Chain-of-Thought reasoning currently serve as the dominant approaches to explainability, growing evidence shows that they are often unreliable, producing brittle, misleading, or illusory explanations that fail to reflect true model behavior. This tutorial aims to unpack why these limitations arise. We first establish the theoretical intractability of complete mechanistic explanations for modern LLMs and clarify the intrinsic barriers to achieving full transparency in overparameterized models. We then pivot to a principled alternative: user-centric explainability, with a focus on concept-based interpretability and controlled data attribution. We review the theoretical foundations of these methods and survey their modern extensions that enable comprehensive explanation, inference-time intervention, and model editability. Finally, we demonstrate how such approaches support effective human--AI collaboration in high-stakes scientific and decision-critical applications. By synthesizing foundational theory, critical analysis of existing methods, and emerging techniques, this tutorial offers a coherent framework for developing the next generation of explainable and trustworthy AI systems.

Single-cell RNA sequencing (scRNA-seq) enables high-resolution characterization of cellular heterogeneity, but its rich, complementary structure across cells and genes remains underexploited, especially in the presence of technical noise and sparsity. Effectively leveraging this multi-scale structure is essentially an information fusion problem that requires integrating heterogeneous graph-based views of cells and genes into robust low-dimensional representations. In this paper, we introduce GatorSC, a unified representation learning framework that models scRNA-seq data through multi-scale cell and gene graphs and fuses them with a mixture-of-experts architecture. GatorSC constructs a global cell-cell graph, a global gene-gene graph, and a local gene-gene graph derived from neighborhood-specific subgraphs, and learns graph neural network embeddings that are adaptively fused by a gating network. To learn noise-robust and structure-preserving embeddings without labels, we couple graph reconstruction and graph contrastive learning in a unified self-supervised objective applied to both cell- and gene-level graphs. We evaluate GatorSC on 19 publicly available scRNA-seq datasets covering diverse tissues, species, and sequencing platforms. Experiments showed that GatorSC consistently outperforms state-of-the-art deep generative, graph-based, and contrastive methods for cell clustering, gene expression imputation, and cell-type annotation. The learned embeddings are used for accurate trajectory inference, recovery of canonical marker gene programs, and cell-type-specific pathway signatures in an Alzheimer's disease single-nucleus dataset. GatorSC provides a flexible foundation for comprehensive single-cell transcriptomic analysis and can be readily extended to multi-omic and spatial modalities.

Distribution shifts on graphs—the discrepancies in data distribution between training and employing a graph machine learning model—are ubiquitous and often unavoidable in real-world applications. These shifts may severely deteriorate model performance, posing significant challenges for reliable graph machine learning. In recent years, there has been a surge in research on graph machine learning specifically designed to tackle such distribution shifts, aiming to train models to achieve satisfactory performance on Out-of-Distribution (OOD) test data. This survey provides an up-to-date and forward-looking review of deep graph learning under distribution shifts. We categorize the field into three primary scenarios: graph OOD generalization, training-time graph OOD adaptation, and test-time graph OOD adaptation. We begin by formally formulating the problems and discussing various types of distribution shifts that can affect graph learning, such as covariate shifts and concept shifts. To provide a structured understanding of the literature, we introduce a systematic taxonomy that classifies existing methods into model-centric and data-centric approaches, investigating the techniques used in each category. We also summarize commonly used datasets in this research area to facilitate further investigation. Finally, we point out promising research directions and the corresponding challenges to encourage further study in this vital domain. Additionally, we provide a continuously updated reading list at https://github.com/kaize0409/Awesome-Graph-OOD .

The superiority of graph contrastive learning (GCL) has prompted its application to anomaly detection tasks for more powerful risk warning systems. Unfortunately, existing GCL-based models tend to excessively prioritize overall detection performance while neglecting robustness to structural imbalance, which can be problematic for many real-world networks following power-law degree distributions. Particularly, GCL-based methods may fail to capture tail anomalies (abnormal nodes with low degrees). This raises concerns about the security and robustness of current anomaly detection algorithms and therefore hinders their applicability in a variety of realistic high-risk scenarios. To the best of our knowledge, research on the robustness of graph anomaly detection to structural imbalance has received little scrutiny. To address the above issues, this paper presents a novel GCL-based framework named AD-GCL. It devises the neighbor pruning strategy to filter noisy edges for head nodes and facilitate the detection of genuine tail nodes by aligning from head nodes to forged tail nodes. Moreover, AD-GCL actively explores potential neighbors to enlarge the receptive field of tail nodes through anomaly-guided neighbor completion. We further introduce intra- and inter-view consistency loss of the original and augmentation graph for enhanced representation. The performance evaluation of the whole, head, and tail nodes on multiple datasets validates the comprehensive superiority of the proposed AD-GCL in detecting both head anomalies and tail anomalies.

Large Language Models (LLMs) have brought significant advances across various NLP tasks through few-shot or zero-shot prompting, bypassing the need for parameter tuning. However, the "black-box" nature behind their massive parameter sizes increases the "hallucination" concerns, especially in high-stakes applications (e.g., healthcare), where decision mistakes can lead to severe consequences. In contrast, human decision-making relies on complex cognitive processes, such as the ability to sense and adaptively correct mistakes through conceptual understanding. Drawing inspiration from human cognition, we propose an innovative metacognitive approach CLEAR, to equip LLMs with capabilities for self-aware error identification and correction. Our framework constructs concept-specific sparse subnetworks that indicate decision processes. This provides a novel interface for model {intervention} after deployment. The benefits include: (i) at inference time, our metacognitive LLMs can self-consciously identify potential mispredictions with minimum human involvement, (ii) the model can self-correct its errors efficiently without additional tuning, and (iii) the correction procedure is not only self-explanatory but also user-friendly, enhancing model interpretability and accessibility. With these metacognitive features, our approach pioneers a new path toward the trustworthiness of LLMs.

Federated Graph Learning (FGL) enables multiple clients to jointly train powerful graph learning models, e.g., Graph Neural Networks (GNNs), without sharing their local graph data for graph-related downstream tasks, such as graph property prediction. In the real world, however, the graph data can suffer from significant distribution shifts across clients as the clients may collect their graph data for different purposes. In particular, graph properties are usually associated with invariant label-relevant substructures (i.e., subgraphs) across clients, while label-irrelevant substructures can appear in a client-specific manner. The issue of distribution shifts of graph data hinders the efficiency of GNN training and leads to serious performance degradation in FGL. To tackle the aforementioned issue, we propose a novel FGL framework entitled FedVN that eliminates distribution shifts through client-specific graph augmentation strategies with multiple learnable Virtual Nodes (VNs). Specifically, FedVN lets the clients jointly learn a set of shared VNs while training a global GNN model. To eliminate distribution shifts, each client trains a personalized edge generator that determines how the VNs connect local graphs in a client-specific manner. Furthermore, we provide theoretical analyses indicating that FedVN can eliminate distribution shifts of graph data across clients. Comprehensive experiments on four datasets under five settings demonstrate the superiority of our proposed FedVN over nine baselines.

Functional Magnetic Resonance Image (fMRI) is commonly employed to study human brain activity, since it offers insight into the relationship between functional fluctuations and human behavior. To enhance analysis and comprehension of brain activity, Graph Neural Networks (GNNs) have been widely applied to the analysis of functional connectivities (FC) derived from fMRI data, due to their ability to capture the synergistic interactions among brain regions. However, in the human brain, performing complex tasks typically involves the activation of certain pathways, which could be represented as paths across graphs. As such, conventional GNNs struggle to learn from these pathways due to the long-range dependencies of multiple pathways. To address these challenges, we introduce a novel framework BrainMAP to learn multiple pathways in brain networks. BrainMAP leverages sequential models to identify long-range correlations among sequentialized brain regions and incorporates an aggregation module based on Mixture of Experts (MoE) to learn from multiple pathways. Our comprehensive experiments highlight BrainMAP's superior performance. Furthermore, our framework enables explanatory analyses of crucial brain regions involved in tasks.

Graph generation aims to create new graphs that closely align with a target graph distribution. Existing works often implicitly capture this distribution by aligning the output of a generator with each training sample. As such, the overview of the entire distribution is not explicitly captured and used for graph generation. In contrast, in this work, we propose a novel self-conditioned graph generation framework designed to explicitly model graph distributions and employ these distributions to guide the generation process. We first perform self-conditioned modeling to capture the graph distributions by transforming each graph sample into a low-dimensional representation and optimizing a representation generator to create new representations reflective of the learned distribution. Subsequently, we leverage these bootstrapped representations as self-conditioned guidance for the generation process, thereby facilitating the generation of graphs that more accurately reflect the learned distributions. We conduct extensive experiments on generic and molecular graph datasets. Our framework, GraphRCG, demonstrates superior performance over existing state-of-the-art graph generation methods in terms of graph quality and fidelity to training data.

The widespread application of graph data in various high-risk scenarios has increased attention to graph anomaly detection (GAD). Faced with real-world graphs that often carry node descriptions in the form of raw text sequences, termed text-attributed graphs (TAGs), existing graph anomaly detection pipelines typically involve shallow embedding techniques to encode such textual information into features, and then rely on complex self-supervised tasks within the graph domain to detect anomalies. However, this text encoding process is separated from the anomaly detection training objective in the graph domain, making it difficult to ensure that the extracted textual features focus on GAD-relevant information, seriously constraining the detection capability. How to seamlessly integrate raw text and graph topology to unleash the vast potential of cross-modal data in TAGs for anomaly detection poses a challenging issue. This paper presents a novel end-to-end paradigm for text-attributed graph anomaly detection, named CMUCL. We simultaneously model data from both text and graph structures, and jointly train text and graph encoders by leveraging cross-modal and uni-modal multi-scale consistency to uncover potential anomaly-related information. Accordingly, we design an anomaly score estimator based on inconsistency mining to derive node-specific anomaly scores. Considering the lack of benchmark datasets tailored for anomaly detection on TAGs, we release 8 datasets to facilitate future research. Extensive evaluations show that CMUCL significantly advances in text-attributed graph anomaly detection, delivering an 11.13% increase in average accuracy (AP) over the suboptimal.

Advances in large language models (LLMs) significantly enhance reasoning capabilities but their deployment is restricted in resourceconstrained scenarios.Knowledge distillation addresses this by transferring knowledge from powerful teacher models to compact and transparent students.However, effectively capturing the teacher's comprehensive reasoning is challenging due to conventional token-level supervision's limited scope.Using multiple reasoning paths per query alleviates this problem, but treating each path identically is suboptimal as paths vary widely in quality and suitability across tasks and models.We propose Qualityfiltered Routing with Cooperative Distillation (QR-Distill), combining path quality filtering, conditional routing, and cooperative peer teaching.First, quality filtering retains only correct reasoning paths scored by an LLM-based evaluation.Second, conditional routing dynamically assigns paths tailored to each student's current learning state.Finally, cooperative peer teaching enables students to mutually distill diverse insights, addressing knowledge gaps and biases toward specific reasoning styles.Experiments demonstrate QR-Distill's superiority over traditional single-and multi-path distillation methods.Ablation studies further highlight the importance of each component-quality filtering, conditional routing, and peer teaching-in effective knowledge transfer.

Retrieval-augmented generation (RAG) enhances large language models (LLMs) by integrating external knowledge sources to address their limitations in accessing up-to-date or specialized information.A natural strategy to increase the likelihood of retrieving relevant information is to expand the number of retrieved documents.However, involving more documents could introduce significant noise, as many documents may be irrelevant or misleading, thereby reducing the overall accuracy of the generated responses.To overcome the challenge associated with handling a larger number of documents, we propose WinnowRAG, a novel RAG framework designed to systematically filter out noisy documents while preserving valuable content -a process we refer to as winnowing.WinnowRAG operates in two stages: In Stage I, we perform queryaware clustering to group similar documents and form distinct topic clusters.Each cluster is assigned to an LLM agent for generating a unique answer.In Stage II, we perform winnowing, wherein a critic LLM evaluates the outputs of multiple agents and iteratively separates useful documents from noisy ones.To retain useful documents when discarding agents, we propose two strategic merging techniques to ensure that only relevant knowledge is used for generating the final response.Crucially, WinnowRAG is model-agnostic and does not require any model fine-tuning, making it easily adaptable to various tasks.Extensive experiments on various realistic datasets demonstrate the effectiveness of WinnowRAG over state-ofthe-art baselines.

Recent progress in large language model (LLM)-based multi-agent collaboration highlights the power of structured communication in enabling collective intelligence.However, existing methods largely rely on static or graphbased inter-agent topologies, lacking the potential adaptability and flexibility in communication.In this work, we propose a new framework that rethinks multi-agent coordination through a sequential structure rather than a graph structure, offering a significantly larger topology space for multi-agent communication.Our method focuses on two key directions: (1) Next-Agent Prediction, which selects the most suitable agent role at each step, and (2) Next-Context Selection (NCS), which enables each agent to selectively access relevant information from any previous step.Together, these components construct task-adaptive communication pipelines that support both role flexibility and global information flow.Extensive evaluations across multiple benchmarks demonstrate that our approach achieves superior performance while substantially reducing communication overhead.

Graph Neural Networks (GNNs) have achieved remarkable success in various graph-based learning tasks. While their performance is often attributed to the powerful neighborhood aggregation mechanism, recent studies suggest that other components such as non-linear layers may also significantly affecting how GNNs process the input graph data in the spectral domain. Such evidence challenges the prevalent opinion that neighborhood aggregation mechanisms dominate the behavioral characteristics of GNNs in the spectral domain. To demystify such a conflict, this paper introduces a comprehensive benchmark to measure and evaluate GNNs' capability in capturing and leveraging the information encoded in different frequency components of the input graph data. Specifically, we first conduct an exploratory study demonstrating that GNNs can flexibly yield outputs with diverse frequency components even when certain frequencies are absent or filtered out from the input graph data. We then formulate a novel research problem of measuring and benchmarking the performance of GNNs from a spectral perspective. To take an initial step towards a comprehensive benchmark, we design an evaluation protocol supported by comprehensive theoretical analysis. Finally, we introduce a comprehensive benchmark on real-world datasets, revealing insights that challenge prevalent opinions from a spectral perspective. We believe that our findings will open new avenues for future advancements in this area. Our implementations can be found at: https://github.com/yushundong/Spectral-benchmark.

While large language models (LLMs) have made significant progress in processing and reasoning over knowledge graphs, current methods suffer from a high non-retrieval rate. This limitation reduces the accuracy of answering questions based on these graphs. Our analysis reveals that the combination of greedy search and forward reasoning is a major contributor to this issue. To overcome these challenges, we introduce the concept of super-relations, which enables both forward and backward reasoning by summarizing and connecting various relational paths within the graph. This holistic approach not only expands the search space, but also significantly improves retrieval efficiency. In this paper, we propose the ReKnoS framework, which aims to Reason over Knowledge Graphs with Super-Relations. Our framework’s key advantages include the inclusion of multiple relation paths through super-relations, enhanced forward and backward reasoning capabilities, and increased efficiency in querying LLMs. These enhancements collectively lead to a substantial improvement in the successful retrieval rate and overall reasoning performance. We conduct extensive experiments on a variety of datasets to evaluate ReKnoS, and the results demonstrate the superior performance of ReKnoS over existing state-of-the-art baselines, with an average accuracy gain of 2.92% across nine real-world datasets.

As Large Language Models (LLMs) are increasingly deployed to handle various natural language processing (NLP) tasks, concerns regarding the potential negative societal impacts of LLM-generated content have also arisen. To evaluate the biases exhibited by LLMs, researchers have recently proposed a variety of datasets. However, existing bias evaluation efforts often focus on only a particular type of bias and employ inconsistent evaluation metrics, leading to difficulties in comparison across different datasets and LLMs. To address these limitations, we collect a variety of datasets designed for the bias evaluation of LLMs, and further propose CEB, a Compositional Evaluation Bechmark that covers different types of bias across different social groups and tasks. The curation of CEB is based on our newly proposed compositional taxonomy, which characterizes each dataset from three dimensions: bias types, social groups, and tasks. By combining the three dimensions, we develop a comprehensive evaluation strategy for the bias in LLMs. Our experiments demonstrate that the levels of bias vary across these dimensions, thereby providing guidance for the development of specific bias mitigation methods.

In-Context Learning (ICL) empowers Large Language Models (LLMs) to tackle diverse tasks by incorporating multiple input-output examples, known as demonstrations, into the input of LLMs. More recently, advancements in the expanded context windows of LLMs have led to many-shot ICL, which uses hundreds of demonstrations and outperforms few-shot ICL, which relies on fewer examples. However, this approach is often hindered by the high cost of obtaining large amounts of labeled data. To address this challenge, we propose Many-Shot Adaptive Pseudo-LabEling, namely MAPLE, a novel influence-based many-shot ICL framework that utilizes pseudo-labeled samples to compensate for the lack of label information. We first identify a subset of impactful unlabeled samples and perform pseudo-labeling on them by querying LLMs. These pseudo-labeled samples are then adaptively selected and tailored to each test query as input to improve the performance of many-shot ICL, without significant labeling costs. Extensive experiments on real-world datasets demonstrate the effectiveness of our framework, showcasing its ability to enhance LLM adaptability and performance with limited labeled data. Our code is provided at https://github.com/Chen-1031/MAPLE_ICL.

Pretrained language models (PLMs) achieve state-of-the-art results but often function as ``black boxes'', hindering interpretability and responsible deployment. While methods like attention analysis exist, they often lack clarity and intuitiveness. We propose interpreting PLMs through high-level, human-understandable concepts using Concept Bottleneck Models (CBMs). This extended abstract introduces C3M (ChatGPT-guided Concept augmentation with Concept-level Mixup), a novel framework for training Concept-Bottleneck-Enabled PLMs (CBE-PLMs). C3M leverages Large Language Models (LLMs) like ChatGPT to augment concept sets and generate noisy concept labels, combined with a concept-level MixUp mechanism to enhance robustness and effectively learn from both human-annotated and machine-generated concepts. Empirical results show our approach provides intuitive explanations, aids model diagnosis via test-time intervention, and improves the interpretability-utility trade-off, even with limited or noisy concept annotations. This is an concise version of [Tan et al., 2024b], recipient of the Best Paper Award at PAKDD 2024. Code and data are released at https://github.com/Zhen-Tan-dmml/CBM_NLP.git.

Fairness-aware graph learning has gained increasing attention in recent years. Nevertheless, there lacks a comprehensive benchmark to evaluate and compare different fairness-aware graph learning methods, which blocks practitioners from choosing appropriate ones for broader real-world applications. In this paper, we present an extensive benchmark on ten representative fairness-aware graph learning methods. Specifically, we design a systematic evaluation protocol and conduct experiments on seven real-world datasets to evaluate these methods from multiple perspectives, including group fairness, individual fairness, the balance between different fairness criteria, and computational efficiency. Our in-depth analysis reveals key insights into the strengths and limitations of existing methods. Additionally, we provide practical guidance for applying fairness-aware graph learning methods in applications. To the best of our knowledge, this work serves as an initial step towards comprehensively understanding representative fairness-aware graph learning methods to facilitate future advancements in this area. Open-source code can be found at: https://github.com/yushundong/Fairness-Aware-Graph-Learning-Benchmark.

In-Context Learning (ICL) is a technique where large language models (LLMs) leverage multiple demonstrations (i.e., examples) to perform tasks. With the recent expansion of LLM context windows, many-shot ICL (generally with more than 50 demonstrations) can lead to significant performance improvements on a variety of language tasks such as text classification and question answering. Nevertheless, ICL faces the issue of demonstration order instability (ICL-DOI), which means that performance varies significantly depending on the order of demonstrations. Moreover, ICL-DOI persists in many-shot ICL, validated by our thorough experimental investigation. Current strategies for handling ICL-DOI are not applicable to many-shot ICL due to two critical challenges: (1) Most existing methods assess demonstration order quality by first prompting the LLM, then using heuristic metrics based on the LLM's predictions. In the many-shot scenarios, these metrics without theoretical grounding become unreliable, where the LLMs struggle to effectively utilize information from long input contexts, making order distinctions less clear. The requirement to examine all orders for the large number of demonstrations is computationally infeasible due to the super-exponential complexity of the order space in many-shot ICL. To tackle the first challenge, we design a demonstration order evaluation metric based on information theory for measuring order quality, which effectively quantifies the usable information gain of a given demonstration order. To address the second challenge, we propose a hierarchical demonstration order optimization method named \texttt{HIDO} that enables a more refined exploration of the order space, achieving high ICL performance without the need to evaluate all possible orders. Extensive experiments on multiple LLMs and real-world datasets demonstrate that our \texttt{HIDO} method consistently and efficiently outperforms other baselines. Our code project can be found at https://github.com/YinhanHe123/HIDO/.

With advancements in reasoning capabilities, Large Language Models (LLMs) are increasingly employed for automated judgment tasks. While LLMs-as-Judges offer promise in automating evaluations, current approaches often rely on simplistic aggregation methods (e.g., majority voting), which can fail even when individual agents provide correct answers. To address this, we propose a multi-agent debate judge framework where agents collaboratively reason and iteratively refine their responses. We formalize the debate process mathematically, analyzing agent interactions and proving that debate amplifies correctness compared to static ensembles. To enhance efficiency, we introduce a stability detection mechanism that models judge consensus dynamics via a time-varying Beta-Binomial mixture, with adaptive stopping based on distributional similarity (Kolmogorov-Smirnov test). This mechanism models the judges' collective correct rate dynamics using a time-varying mixture of Beta-Binomial distributions and employs an adaptive stopping criterion based on distributional similarity (Kolmogorov-Smirnov statistic). Experiments across multiple benchmarks and models demonstrate that our framework improves judgment accuracy over majority voting while maintaining computational efficiency.

Epidemic containment has long been a crucial task in many high-stake application domains, ranging from public health to misinformation dissemination. Existing studies for epidemic containment are primarily focused on undirected networks, assuming that the infection rate is constant throughout the contact network regardless of the strength and direction of contact. However, such an assumption can be unrealistic given the asymmetric nature of the real-world infection process. To tackle the epidemic containment problem in directed networks, simply grafting the methods designed for undirected network can be problematic, as most of the existing methods rely on the orthogonality and Lipschitz continuity in the eigensystem of the underlying contact network, which do not hold for directed networks. In this work, we derive a theoretical analysis on the general epidemic threshold condition for directed networks and show that such threshold condition can be used as an optimization objective to control the spread of the disease. Based on the epidemic threshold, we propose an asymptotically greedy algorithm DINO (DIrected NetwOrk epidemic containment) to identify the most critical nodes for epidemic containment. The proposed algorithm is evaluated on real-world directed networks, and the results validate its effectiveness and efficiency. The code is available at https://github.com/YinhanHe123/DINO/.

Large Language Models (LLMs) have shown impressive performance in various tasks, including knowledge graph completion (KGC). However, current studies mostly apply LLMs to classification tasks, like identifying missing triplets, rather than ranking-based tasks, where the model ranks candidate entities based on plausibility. This focus limits the practical use of LLMs in KGC, as real-world applications prioritize highly plausible triplets. Additionally, while graph paths can help infer the existence of missing triplets and improve completion accuracy, they often contain redundant information. To address these issues, we propose KG-CF, a framework tailored for ranking-based KGC tasks. KG-CF leverages LLMs’ reasoning abilities to filter out irrelevant contexts, achieving superior results on real-world datasets. The code and datasets are available at https://anonymous.4open.science/r/KG-CF.

Job marketplace is a heterogeneous graph composed of interactions among members (job-seekers), companies, and jobs. Understanding and modeling job marketplace can benefit both job seekers and employers, ultimately contributing to the greater good of the society. However, existing graph neural network (GNN)-based methods have shallow understandings of the associated textual features and heterogeneous relations. To address the above challenges, we propose PLM4Job, a job marketplace foundation model that tightly couples pretrained language models (PLM) with job market graph, aiming to fully utilize the pretrained knowledge and reasoning ability to model member/job textual features as well as various member-job relations simultaneously. In the pretraining phase, we propose a heterogeneous ego-graph-based prompting strategy to model and aggregate member/job textual features based on the topological structure around the target member/job node, where entity type embeddings and graph positional embeddings are introduced accordingly to model different entities and their heterogeneous relations. Meanwhile, a proximity-aware attention alignment strategy is designed to dynamically adjust the attention of the PLM on ego-graph node tokens in the prompt, such that the attention can be better aligned with job marketplace semantics. Extensive experiments at LinkedIn demonstrate the effectiveness of PLM4Job.

Zhen Tan, Dawei Li, Song Wang, Alimohammad Beigi, Bohan Jiang, Amrita Bhattacharjee, Mansooreh Karami, Jundong Li, Lu Cheng, Huan Liu. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

Retrieval-Augmented Generative (RAG) models enhance Large Language Models (LLMs) by integrating external knowledge bases, improving their performance in applications like fact-checking and information searching.In this paper, we demonstrate a security threat where adversaries can exploit the openness of these knowledge bases by injecting deceptive content into the retrieval database, intentionally changing the model's behavior.This threat is critical as it mirrors real-world usage scenarios where RAG systems interact with publicly accessible knowledge bases, such as web scrapings and user-contributed data pools.To be more realistic, we target a realistic setting where the adversary has no knowledge of users' queries, knowledge base data, and the LLM parameters.We demonstrate that it is possible to exploit the model successfully through crafted content uploads with access to the retriever.Our findings emphasize an urgent need for security measures in the design and deployment of RAG systems to prevent potential manipulation and ensure the integrity of machinegenerated content.

To deal with distribution shifts in graph data, various graph out-of-distribution (OOD) generalization techniques have been recently proposed. These methods often employ a two-step strategy that first creates augmented environments and subsequently identifies invariant subgraphs to improve generalizability. Nevertheless, this approach could be suboptimal from the perspective of consistency. First, the process of augmenting environments by altering the graphs while preserving labels may lead to graphs that are not realistic or meaningfully related to the origin distribution, thus lacking distribution consistency. Second, the extracted subgraphs are obtained from directly modifying graphs, and may not necessarily maintain a consistent predictive relationship with their labels, thereby impacting label consistency. In response to these challenges, we introduce an innovative approach that aims to enhance these two types of consistency for graph OOD generalization. We propose a modifier to obtain both augmented and invariant graphs in a unified manner. With the augmented graphs, we enrich the training data without compromising the integrity of label-graph relationships. The label consistency enhancement in our framework further preserves the supervision information in the invariant graph. We conduct extensive experiments on real-world datasets to demonstrate the superiority of our framework over other state-of-the-art baselines.

Haochen Liu, Song Wang, Chen Chen, Jundong Li. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

In-Context Learning (ICL) empowers Large Language Models (LLMs) to tackle various tasks by providing input-output examples as additional inputs, referred to as demonstrations. Nevertheless, the performance of ICL could be easily impacted by the quality of selected demonstrations. Existing efforts generally learn a retriever model to score each demonstration for selecting suitable demonstrations, however, the effect is suboptimal due to the large search space and the noise from unhelpful demonstrations. In this study, we introduce MoD, which partitions the demonstration pool into groups, each governed by an expert to reduce search space. We further design an expert-wise training strategy to alleviate the impact of unhelpful demonstrations when optimizing the retriever model. During inference, experts collaboratively retrieve demonstrations for the input query to enhance the ICL performance. We validate MoD via experiments across a range of NLP datasets and tasks, demonstrating its state-of-the-art performance and shedding new light on the future design of retrieval methods for ICL.

The problem of few-shot graph classification targets at assigning class labels for graph samples, where only limited labeled graphs are provided for each class. To solve the problem brought by label scarcity, recent studies have proposed to adopt the prevalent few-shot learning framework to achieve fast adaptations to graph classes with limited labeled graphs. In particular, these studies typically propose to accumulate meta-knowledge across a large number of meta-training tasks, and then generalize such meta-knowledge to meta-test tasks sampled from a disjoint class set. Nevertheless, existing studies generally ignore the crucial task correlations among meta-training tasks and treat them independently. In fact, such task correlations can help promote the model generalization to meta-test tasks and result in better classification performance. On the other hand, it remains challenging to capture and utilize task correlations due to the complex components and interactions in meta-training tasks. To deal with this, we propose a novel few-shot graph classification framework FAITH to capture task correlations via learning a hierarchical task structure at different granularities. We further propose a task-specific classifier to incorporate the learned task correlations into the few-shot graph classification process. Moreover, we derive FAITH+, a variant of FAITH that can improve the sampling process for the hierarchical task structure. The extensive experiments on four prevalent graph datasets further demonstrate the superiority of FAITH and FAITH+ over other state-of-the-art baselines.

Graph Neural Networks (GNNs) have emerged as the leading paradigm for solving graph analytical problems in various real-world applications. Nevertheless, GNNs could potentially render biased predictions towards certain demographic subgroups. Understanding how the bias in predictions arises is critical, as it guides the design of GNN debiasing mechanisms. However, most existing works overwhelmingly focus on GNN debiasing, but fall short on explaining how such bias is induced. In this paper, we study a novel problem of interpreting GNN unfairness through attributing it to the influence of training nodes. Specifically, we propose a novel strategy named Probabilistic Distribution Disparity (PDD) to measure the bias exhibited in GNNs, and develop an algorithm to efficiently estimate the influence of each training node on such bias. We verify the validity of PDD and the effectiveness of influence estimation through experiments on real-world datasets. Finally, we also demonstrate how the proposed framework could be used for debiasing GNNs. Open-source code can be found at https://github.com/yushundong/BIND.

Recently, there has been a growing interest in developing machine learning (ML) models that can promote fairness, i.e., eliminating biased predictions towards certain populations (e.g., individuals from a specific demographic group). Most existing works learn such models based on well-designed fairness constraints in optimization. Nevertheless, in many practical ML tasks, only very few labeled data samples can be collected, which can lead to inferior fairness performance. This is because existing fairness constraints are designed to restrict the prediction disparity among different sensitive groups, but with few samples, it becomes difficult to accurately measure the disparity, thus rendering ineffective fairness optimization. In this paper, we define the fairness-aware learning task with limited training samples as the fair few-shot learning problem. To deal with this problem, we devise a novel framework that accumulates fairness-aware knowledge across different meta-training tasks and then generalizes the learned knowledge to meta-test tasks. To compensate for insufficient training samples, we propose an essential strategy to select and leverage an auxiliary set for each meta-test task. These auxiliary sets contain several labeled training samples that can enhance the model performance regarding fairness in meta-test tasks, thereby allowing for the transfer of learned useful fairness-oriented knowledge to meta-test tasks. Furthermore, we conduct extensive experiments on three real-world datasets to validate the superiority of our framework against the state-of-the-art baselines.

Federated Learning (FL) enables multiple clients to collaboratively learn a machine learning model without exchanging their own local data. In this way, the server can exploit the computational power of all clients and train the model on a larger set of data samples among all clients. Although such a mechanism is proven to be effective in various fields, existing works generally assume that each client preserves sufficient data for training. In practice, however, certain clients can only contain a limited number of samples (i.e., few-shot samples). For example, the available photo data taken by a specific user with a new mobile device is relatively rare. In this scenario, existing FL efforts typically encounter a significant performance drop on these clients. Therefore, it is urgent to develop a few-shot model that can generalize to clients with limited data under the FL scenario. In this paper, we refer to this novel problem as federated few-shot learning. Nevertheless, the problem remains challenging due to two major reasons: the global data variance among clients (i.e., the difference in data distributions among clients) and the local data insufficiency in each client (i.e., the lack of adequate local data for training). To overcome these two challenges, we propose a novel federated few-shot learning framework with two separately updated models and dedicated training strategies to reduce the adverse impact of global data variance and local data insufficiency. Extensive experiments on four prevalent datasets that cover news articles and images validate the effectiveness of our framework compared with the state-of-the-art baselines.

Few-shot node classification, which aims to predict labels for nodes on graphs with only limited labeled nodes as references, is of great significance in real-world graph mining tasks. To tackle such a label shortage issue, existing works generally leverage the meta-learning framework, which utilizes a number of episodes to extract transferable knowledge from classes with abundant labeled nodes and generalizes the knowledge to other classes with limited labeled nodes. In essence, the primary aim of few-shot node classification is to learn node embeddings that are generalizable across different classes. To accomplish this, the GNN encoder must be able to distinguish node embeddings between different classes, while also aligning embeddings for nodes in the same class. Thus, in this work, we propose to consider both the intra-class and inter-class generalizability of the model. We create a novel contrastive meta-learning framework on graphs, named COSMIC, with two key designs. First, we propose to enhance the intra-class generalizability by involving a contrastive two-step optimization in each episode to explicitly align node embeddings in the same classes. Second, we strengthen the inter-class generalizability by generating hard node classes for classification via a novel similarity-sensitive mix-up strategy. Extensive experiments on prevalent few-shot node classification datasets verify the effectiveness of our framework and demonstrate its superiority over other state-of-the-art baselines.

Few-shot node classification aims at classifying nodes with limited labeled nodes as references. Recent few-shot node classification methods typically learn from classes with abundant labeled nodes (i.e., meta-training classes) and then generalize to classes with limited labeled nodes (i.e., meta-test classes). Nevertheless, on real-world graphs, it is usually difficult to obtain abundant labeled nodes for many classes. In practice, each meta-training class can only consist of several labeled nodes, known as the extremely weak supervision problem. In few-shot node classification, with extremely limited labeled nodes for meta-training, the generalization gap between meta-training and meta-test will become larger and thus lead to suboptimal performance. To tackle this issue, we study a novel problem of few-shot node classification with extremely weak supervision and propose a principled framework X-FNC under the prevalent meta-learning framework. Specifically, our goal is to accumulate meta-knowledge across different meta-training tasks with extremely weak supervision and generalize such knowledge to meta-test tasks. To address the challenges resulting from extremely scarce labeled nodes, we propose two essential modules to obtain pseudo-labeled nodes as extra references and effectively learn from extremely limited supervision information. We further conduct extensive experiments on four node classification datasets with extremely weak supervision to validate the superiority of our framework compared to the state-of-the-art baselines.

Graph mining algorithms have been playing a significant role in myriad fields over the years. However, despite their promising performance on various graph analytical tasks, most of these algorithms lack fairness considerations. As a consequence, they could lead to discrimination towards certain populations when exploited in human-centered applications. Recently, algorithmic fairness has been extensively studied in graph-based applications. In contrast to algorithmic fairness on independent and identically distributed (i.i.d.) data, fairness in graph mining has exclusive backgrounds, taxonomies, and fulfilling techniques. In this survey, we provide a comprehensive and up-to-date introduction of existing literature under the context of fair graph mining. Specifically, we propose a novel taxonomy of fairness notions on graphs, which sheds light on their connections and differences. We further present an organized summary of existing techniques that promote fairness in graph mining. Finally, we discuss current research challenges and open questions, aiming at encouraging cross-breeding ideas and further advances.

Few-shot graph classification aims at predicting classes for graphs, given limited labeled graphs for each class. To tackle the bottleneck of label scarcity, recent works propose to incorporate few-shot learning frameworks for fast adaptations to graph classes with limited labeled graphs. Specifically, these works propose to accumulate meta-knowledge across diverse meta-training tasks, and then generalize such meta-knowledge to the target task with a disjoint label set. However, existing methods generally ignore task correlations among meta-training tasks while treating them independently. Nevertheless, such task correlations can advance the model generalization to the target task for better classification performance. On the other hand, it remains non-trivial to utilize task correlations due to the complex components in a large number of meta-training tasks. To deal with this, we propose a novel few-shot learning framework FAITH that captures task correlations via constructing a hierarchical task graph at different granularities. Then we further design a loss-based sampling strategy to select tasks with more correlated classes. Moreover, a task-specific classifier is proposed to utilize the learned task correlations for few-shot classification. Extensive experiments on four prevalent few-shot graph classification datasets demonstrate the superiority of FAITH over other state-of-the-art baselines.

Graph Neural Networks (GNNs) have shown satisfying performance in various graph analytical problems. Hence, they have become the de facto solution in a variety of decision-making scenarios. However, GNNs could yield biased results against certain demographic subgroups. Some recent works have empirically shown that the biased structure of the input network is a significant source of bias for GNNs. Nevertheless, no studies have systematically scrutinized which part of the input network structure leads to biased predictions for any given node. The low transparency on how the structure of the input network influences the bias in GNN outcome largely limits the safe adoption of GNNs in various decision-critical scenarios. In this paper, we study a novel research problem of structural explanation of bias in GNNs. Specifically, we propose a novel post-hoc explanation framework to identify two edge sets that can maximally account for the exhibited bias and maximally contribute to the fairness level of the GNN prediction for any given node, respectively. Such explanations not only provide a comprehensive understanding of bias/fairness of GNN predictions but also have practical significance in building an effective yet fair GNN model. Extensive experiments on real-world datasets validate the effectiveness of the proposed framework towards delivering effective structural explanations for the bias of GNNs. Open-source code can be found at https://github.com/yushundong/REFEREE.

Node classification is of great importance among various graph mining tasks. In practice, real-world graphs generally follow the long-tail distribution, where a large number of classes only consist of limited labeled nodes. Although Graph Neural Networks (GNNs) have achieved significant improvements in node classification, their performance decreases substantially in such a few-shot scenario. The main reason can be attributed to the vast generalization gap between meta-training and meta-test due to the task variance caused by different node/class distributions in meta-tasks (i.e., node-level and class-level variance). Therefore, to effectively alleviate the impact of task variance, we propose a task-adaptive node classification framework under the few-shot learning setting. Specifically, we first accumulate meta-knowledge across classes with abundant labeled nodes. Then we transfer such knowledge to the classes with limited labeled nodes via our proposed task-adaptive modules. In particular, to accommodate the different node/class distributions among meta-tasks, we propose three essential modules to perform node-level, class-level, and task-level adaptations in each meta-task, respectively. In this way, our framework can conduct adaptations to different meta-tasks and thus advance the model generalization performance on meta-test tasks. Extensive experiments on four prevalent node classification datasets demonstrate the superiority of our framework over the state-of-the-art baselines. Our code is provided at https://github.com/SongW-SW/TENT https://github.com/SongW-SW/TENT.

Graph few-shot learning is of great importance among various graph learning tasks. Under the few-shot scenario, models are often required to conduct classification given limited labeled samples. Existing graph few-shot learning methods typically leverage Graph Neural Networks (GNNs) and perform classification across a series of meta-tasks. Nevertheless, these methods generally rely on the original graph (i.e., the graph that the meta-task is sampled from) to learn node representations. Consequently, the learned representations for the same nodes are identical in all meta-tasks. Since the class sets are different across meta-tasks, node representations should be task-specific to promote classification performance. Therefore, to adaptively learn node representations across meta-tasks, we propose a novel framework that learns a task-specific structure for each meta-task. To handle the variety of nodes across meta-tasks, we extract relevant nodes and learn task-specific structures based on node influence and mutual information. In this way, we can learn node representations with the task-specific structure tailored for each meta-task. We further conduct extensive experiments on five node classification datasets under both single- and multiple-graph settings to validate the superiority of our framework over the state-of-the-art baselines.

Knowledge graphs (KGs) are of great importance in various artificial intelligence systems, such as question answering, relation extraction, and recommendation. Nevertheless, most real-world KGs are highly incomplete, with many missing relations between entities. To discover new triples (i.e., head entity, relation, tail entity), many KG completion algorithms have been proposed in recent years. However, a vast majority of existing studies often require a large number of training triples for each relation, which contradicts the fact that the frequency distribution of relations in KGs often follows a long tail distribution, meaning a majority of relations have only very few triples. Meanwhile, since most existing large-scale KGs are constructed automatically by extracting information from crowd-sourcing data using heuristic algorithms, plenty of errors could be inevitably incorporated due to the lack of human verification, which greatly reduces the performance for KG completion. To tackle the aforementioned issues, in this paper, we study a novel problem of error-aware few-shot KG completion and present a principled KG completion framework REFORM. Specifically, we formulate the problem under the few-shot learning framework, and our goal is to accumulate meta-knowledge across different meta-tasks and generalize the accumulated knowledge to the meta-test task for error-aware few-shot KG completion. To address the associated challenges resulting from insufficient training samples and inevitable errors, we propose three essential modules neighbor encoder, cross-relation aggregation, and error mitigation in each meta-task. Extensive experiments on three widely used KG datasets demonstrate the superiority of the proposed framework REFORM over competitive baseline methods.