Multi-domain service platforms increasingly require atomic workflows that span multiple autonomous domains with shared mutable states (e.g., blockchains, databases, and stateful microservices). Such settings face three practical challenges: delayed state visibility across domains, black-box internal scheduling inside each domain, and complex dependency cycles that span domains. To address these challenges, we propose a practical framework for multi-domain service transactions in deployments where participating domains accept ordered batches endorsed by a Byzantine fault-tolerant validator set (e.g., permissioned domains governed by a shared BFT group). First, we formalize the Optimal Cross-domain Serializable Schedule (OCS2) problem and prove its NP-hardness. Second, we introduce a unified transaction footprint model that integrates Account-based and UTXO-based semantics for universal conflict detection across heterogeneous domains. Third, we propose the Hierarchical Conflict-aware Scheduling (HCS) algorithm that transforms the NP-hard (OCS2) problem into parallelizable sub-problems through hierarchical decomposition. Finally, we design a digest-based batch commit protocol (HOBC) that shifts agreement from individual transactions to batch-level result digests. To handle delayed visibility and state drift between analysis and execution, the protocol validates execution-time preconditions and safely aborts transactions whose prerequisites are no longer satisfied, without compromising correctness. Our prototype evaluation shows improved concurrency efficiency under contention while preserving end-to-end correctness for cross-domain service transactions.

Graph learning has been increasingly deployed in critical and sensitive domains, raising pressing demands for trustworthiness-robustness, fairness, and beyond. However, these properties are often undermined by various perturbations, which induce distributional uncertainty and compromise the trustworthiness of graph learning. To address this, we propose DICT, a novel framework that models distributional uncertainty to achieve trustworthy graph learning. Specifically, DICT formulates a unified optimization objective that captures perturbation-induced distributional shifts in graph topology, node features, and labels, and minimizes the worst-case risk over the uncertainty set. However, directly optimizing this objective in its primal form leads to an infinite-dimensional problem. To make this problem tractable, we integrate strong duality and local Lipschitz continuity of the loss, reformulating the objective as a finite-dimensional min-max problem. We focus on robustness and fairness as primary instantiations of DICT because they are not only critical in real-world applications, but also provide transferable modeling principles for broader trustworthiness objectives. By formulating fairness in the form of an uncertainty set, DICT pioneers unified robustness and fairness within a single optimization framework. Extensive experiments across diverse benchmarks and backbones demonstrate that DICT consistently improves both robustness and fairness, validating the effectiveness and adaptability of the DICT framework.

Agent planning faces dynamic heterogeneity—nonstationary observations, dynamics, and objectives with sparse, delayed rewards—which dominant methods largely ignore, leading to poor generalization under environment shifts. We propose Flow-Matching for Agent Planning (FlowMAP), which formulates planning as a continuous-time flow matching by learning a planning-time velocity field that transports an initial meta-state distribution toward a task-conditioned target. FlowMAP introduces Value-Transport Flow Matching to provide distribution-level planning objective that steers transport toward high-value regions in meta-state distribution, mitigating error accumulation under environmental shifts. To enforce alignment between meta-state distributions transport and action-environment interaction, FlowMAP further propose Flow-Policy Co-Training, which jointly optimizes the planning flow and policy so that the flow transport directly regularizes the policy-induced meta-distribution dynamics. Across diverse agent planning benchmarks, FlowMAP consistently outperforms strong baselines, yielding improvement in planning generalization.

Building self-explaining NLP models is a powerful approach in Explainable Artificial Intelligence (XAI). Selective Rationalization (SR) and Multi-Hop Question Answering (MHQA) reasoning tasks have received increasing research attention currently. SR models usually select text segments related to the downstream prediction task from the input text as the rationale for their predictions. MHQA reasoning model first retrieves relevant context matching the question from multiple documents, then combines multiple evidence documents for logical reasoning, and finally forms a reasoning path and provides the correct answer. However, SR models often suffer from feature spurious correlations and degeneration problem. MHQA reasoning tasks are easily disrupted by multiple disjoint text fragments or entities, causing the reasoning chain to break and ultimately failing to arrive at the correct answer. To address these challenges, we propose an NLP Self-Explaining framework based on cooperative rationalization and multi-hop evidence reasoning (S-Explainer). S-Explainer integrates the SR and MHQA reasoning tasks into a single framework for study, effectively improving task performance and enhancing model robustness through a cooperative game and two-stage refinement selection method. A series of experiments conducted on three real datasets also verified the effectiveness of our proposed method.

Despite the widespread success of graph neural networks (GNNs) in various graph learning tasks, their performance is often hampered by the over-squashing issue, which impedes the propagation of messages from distant nodes and limits the expressive power of the model. Existing solutions to mitigate over-squashing primarily rely on altering graph topology to improve message flow. However, such modification of structure can disrupt the graph’s intrinsic topology and corresponding inductive bias, while the introduction of extra edges may increase the risk of over-smoothing. To address these limitations, we propose the Virtual Path Fusion (VPF) framework, an enhanced GNN that tackles over-squashing by facilitating message flow through virtual paths, offering a topology-preserving solution that sidesteps the inherent risks of structural distortion and over-smoothing. Specifically, our method leverages effective resistance, a universal measure that captures both sensitivity and spectral properties, to guide the construction of virtual paths that target structurally susceptible bottlenecks. These paths are encoded via sequence models to capture long-range dependencies, thereby adaptively strengthening interactions between distant nodes. VPF is designed as a model-agnostic and plug-and-play module, making it compatible with a variety of message-passing GNN architectures, while also contributing to the mitigation of over-smoothing. Extensive experiments demonstrate that VPF consistently and significantly outperforms baseline methods across multiple benchmarks, validating virtual path augmentation as an effective and versatile strategy for tackling over-squashing. The code is available at: https://github.com/BHuiwen/VPF.

Biplex, as a relaxation of the biclique model, has emerged as an important cohesive subgraph model for bipartite graph analysis. The maximum k-biplex search problem aims to identify the k-biplex with maximum number of edges and has been widely applied in various real-world applications, including community detection, online recommendation, and fraud detection. However, the problem is NP-hard, and existing exact algorithms remain inefficient on large-scale bipartite graphs with large values of k (e.g., k≥ 3). In this paper, we revisit the maximum k-biplex search problem from a complementary perspective. We reveal a novel structural duality: finding a maximum k-biplex in a bipartite graph is equivalent to finding a minimal k-bounded-degree deletion in its complement graph. Based on this observation, we propose a novel deletion-based algorithm for the maximum k-biplex search problem. We theoretically prove that the proposed algorithm achieves a worst-case time complexity of O^*(γ_k^n), where γ_k<2. Specifically, γ_1=1.725, γ_2=1.856, and γ_3=1.928. To further enhance practical efficiency, we develop several effective upper-bounding techniques and a heuristic strategy for obtaining high-quality initial solutions, which substantially reduce the search space. Extensive experiments on eight real-world bipartite graphs demonstrate the efficiency of our approach, which achieves up to four orders of magnitude speedups over state-of-the-art algorithms.

Multimodal-attributed graphs (MAGs), whose nodes carry modalities such as images and text alongside topological structure, now pervade applications including social platforms, e-commerce, and biomedical networks, offering richer semantic signals than single-modality graphs. In practice, such graphs are fragmented across privacy-restricted silos owned by different platforms and institutions, so learning a broadly transferable model over them demands collaborative training that never exposes raw data. This places the task at the intersection of multimodal graph learning and federated learning, yet existing methods cover only one side of it. To address the challenges from these two perspectives, we propose FedGAMMA, casting federated multimodal graph foundation learning as a two-stage semantic-structural alignment problem of federated pre-training and prompt-based fine-tuning. During pre-training, a shared-private semantic enhancer disentangles cross-modal commonality from modality-specific information, aligning it through optimal transport, a topology-aware graph fusion module decouples semantic and structural views via semantic residual graphs and dual positional encodings, and a dual-channel affinity-aware aggregation mechanism estimates client similarity from feature and graph centroids without exposing raw data. During fine-tuning, FedGAMMA adapts the pretrained encoder through lightweight graph-aware prompts, a shared prompt pool with controlled exploration, and channel-wise prompt synchronization. Experiments on twelve multimodal graph datasets show FedGAMMA consistently surpassing a broad range of baselines across downstream tasks, with gains of up to 12.96

Graph Convolutional Networks (GCNs) are widely used for large graph-structured data, including social, citation, and e-commerce networks, but their deployment is constrained by irregular memory access and severe GPU workload imbalance. These challenges arise in two dimensions: width imbalance from power-law degree distributions and depth imbalance from heterogeneous neighborhood connectivity.We present DualGCN, a GPU acceleration framework addressing both dimensions through dual-metric graph partitioning and adaptive kernel execution. DualGCN combines node degree, reflecting aggregation width, with neighborhood density estimated by anonymous random walks, capturing multihop connectivity and access depth. This hybrid workload metric enables connectivity-aware partitioning of large graphs into sparse and dense regions while reducing workload imbalance from linear to logarithmic complexity. DualGCN then selects partition-specific execution strategies: sparse partitions use warp-level parallelism and coalesced memory access, whereas dense partitions exploit instruction-level parallelism to hide latency and improve GPU utilization. Experiments on twelve real-world graph datasets show that DualGCN consistently accelerates GCN computation, achieving average speedups of 2.53x, 3.8x, and 2.13x over cuSPARSE, GNNAdvisor, and ACCEL, respectively. These results demonstrate that jointly optimizing graph partitioning and kernel execution provides an effective solution for processing large-scale graph and socialnetwork workloads.

Implicit collaborative filtering (CF) models are widely used in recommender systems due to their cost-effective data collection and broad applicability. Negative sampling methods play a crucial role in training implicit CF models. However, most of these methods ignore the false negative problem, and the remaining methods rely solely on statistical features to identify them. The integration of contextual information to address this issue remains unexplored. To fill this gap, we propose a novel sampling method called Reliable Hard Negative Sampling (RHNS). Our approach consists of two key components: a hard negative sampling module, which constructs hard negative samples through a controllable dimension-level fusion approach, and a reliable transformation module, which incorporates contextual information to transform hard negative samples, including those that may be false negatives, into more reliable hard negative samples. We provide theoretical insights showing that the hard negative sampling module tends to generate harder negative samples than existing hard negative sampling methods, while the reliable transformation module tends to mitigate false negative risk. Extensive experimental results on five real-world datasets demonstrate the superiority of RHNS.

Multimodal Graph Neural Networks (MGNNs) have shown strong potential for learning from multimodal attributed graphs, yet most existing approaches rely on tightly coupled architectures that suffer from prohibitive computational overhead. In this paper, we present a systematic empirical analysis showing that decoupled MGNNs are substantially more efficient and scalable for large-scale graph learning. However, we identify a critical bottleneck in existing decoupled pipelines, namely modal conflict, which arises in both the propagation and aggregation stages. Specifically, independent multi-hop diffusion causes cross-modal semantic divergence during propagation, while naive fusion fails to align multi-hop feature trajectories during aggregation, jointly limiting effective representation learning. To address this challenge, we propose CAMPA, a Cross-modal Aligned Multimodal Propagation Aggregation framework for decoupled multimodal graph learning. Concretely, CAMPA introduces a two-stage alignment mechanism: (1) cross-modal aligned propagation, which injects cross-modal similarity priors into message passing to preserve semantic consistency without additional parameter overhead; (2) trajectory aligned aggregation, which leverages trajectory-level self-attention and cross-attention to capture and align long-range dependencies across modalities and hops. Extensive experiments on diverse benchmark datasets and tasks demonstrate that CAMPA consistently outperforms strong coupled and decoupled baselines while preserving the efficiency advantages of the decoupled paradigm.

Federated graph learning (FGL) enables collaborative training on graph data across multiple clients. With the rise of large language models (LLMs), textual attributes in FGL graphs are gaining attention. Text-attributed graph federated learning (TAG-FGL) improves FGL by explicitly leveraging LLMs to process and integrate these textual features. However, current TAG-FGL methods face three main challenges: \textbf{(1) Overhead.} LLMs for processing long texts incur high token and computation costs. To make TAG-FGL practical, we introduce graph condensation (GC) to reduce computation load, but this choice also brings new issues. \textbf{(2) Suboptimal.} To reduce LLM overhead, we introduce GC into TAG-FGL by compressing multi-hop texts/neighborhoods into a condensed core with fixed LLM surrogates (summaries/embeddings). However, this one-shot condensation is often not client-adaptive, leading to suboptimal performance. \textbf{(3) Interpretability.} LLM-based condensation further introduces a black-box bottleneck: summaries lack faithful attribution and clear grounding to specific source spans, making local inspection and auditing difficult. To address the above issues, we propose \textbf{DANCE}, a new TAG-FGL paradigm with GC. To improve \textbf{suboptimal} performance, DANCE performs round-wise, model-in-the-loop condensation refresh using the latest global model. To enhance \textbf{interpretability}, DANCE preserves provenance by storing locally inspectable evidence packs that trace predictions to selected neighbors and source text spans. Across 8 TAG datasets, DANCE improves accuracy by \textbf{2.33\%} at an \textbf{8\%} condensation ratio, with \textbf{33.42\%} fewer tokens per condensed node than TAG-FGL baselines.

The problem of identifying the maximum edge biclique in bipartite graphs has attracted considerable attention in bipartite graph analysis, with numerous real-world applications such as fraud detection, community detection, and online recommendation systems. However, real-world graphs may contain noise or incomplete information, leading to overly restrictive conditions when employing the biclique model. To mitigate this, we focus on a new relaxed subgraph model, called the k -defective biclique, which allows for up to k missing edges compared to the biclique model. We investigate the problem of finding the maximum edge k -defective biclique in a bipartite graph, and prove that the problem is NP-hard. To tackle this computation challenge, we propose a novel algorithm based on a new branch-and-bound framework, which achieves a worst-case time complexity of O ( mα n k ), where α k < 2. We further enhance this framework by incorporating a novel pivoting technique, reducing the worst-case time complexity to O(mβ n k ) , where β k < α k . To improve the efficiency, we develop a series of optimization techniques, including graph reduction methods, novel upper bounds, and a heuristic approach. Extensive experiments on 11 large real-world datasets validate the efficiency and effectiveness of the proposed approaches. The results indicate that our algorithms consistently outperform state-of-the-art algorithms, offering up to 1000× speedups across various parameter settings.

Directed graphs are ubiquitous across numerous domains, where the directionality of edges encodes critical causal dependencies. However, existing GNNs and graph Transformers tailored for directed graphs face two major challenges: (1) effectively capturing long-range causal dependencies derived from directed edges; (2) balancing accuracy and training efficiency when processing large-scale graph datasets. In recent years, state space models (SSMs) have achieved substantial progress in causal sequence tasks, and their variants designed for graphs have demonstrated state-of-the-art accuracy while maintaining high efficiency across various graph learning benchmarks. However, existing graph state space models are exclusively designed for undirected graphs, which limits their performance in directed graph learning. To this end, we propose an innovative approach DirEgo2Token which sequentializes directed graphs via k-hop ego graphs. This marks the first systematic extension of state space models to the field of directed graph learning. Building upon this, we develop DirGraphSSM, a novel directed graph neural network architecture that implements state space models on directed graphs via the message-passing mechanism. Experimental results demonstrate that DirGraphSSM achieves state-of-the-art performance on three representative directed graph learning tasks while attaining competitive performance on two additional tasks with 1.5× to 2× training speed improvements compared to existing state-of-the-art models.

Most implicit collaborative filtering (CF) models are trained with negative sampling, where existing work designs sophisticated strategies for high-quality negatives while largely overlooking the exploration of positive samples. Although some denoising recommendation methods can be applied to implicit CF for denoising positive samples, they often sparsify positive supervision. Moreover, these approaches generally overlook user activity bias during training, leading to insufficient learning for inactive users. To address these issues, we propose a simple yet effective negative sampling plugin, PSP-NS, from the perspective of enhancing positive supervision signals. It builds a user-item bipartite graph with edge weights indicating interaction confidence inferred from global and local patterns, generates positive sample pairs via replication-based reweighting to strengthen positive signals, and adopts an activity-aware weighting scheme to effectively learn inactive users' preferences. We provide theoretical insights from a margin-improvement perspective, explaining why PSP-NS tends to improve ranking quality (e.g., Precision@k/Recall@k), and conduct extensive experiments on four real-world datasets to demonstrate its superiority. For instance, PSP-NS boosts Recall@30 and Precision@30 by 32.11

A data lake maintains large amounts of heterogeneous data with different data schemas and query interfaces. Efficiently querying and analyzing the heterogeneous data enables users to gain more complete insights. In this article, we study a novel problem of distributed keyword search across heterogeneous data sources. Traditional distributed search algorithms generally require the predefined crossing edges connecting relevant data instances for communication between different sources, which is unpractical for the data lake due to the schema heterogeneity. To effectively perform keyword search over the data lake, we first introduce canonical graphs and then develop a best-first search algorithm called UnifySea, which explores the answers across different sources based on the unified identification of related instances. To further improve the query efficiency, we propose a novel incremental keyword search algorithm called DistSea, which just need to identify the promising relevant data between different sources. DistSea incrementally calculates the optimal answers based on locally partial evaluation. Equipped with several efficient pruning rules, DistSea reduces unpromising tree calculation across different sources. Experimental evaluations on six real-world benchmarks demonstrate the effectiveness, efficiency, and scalability of the proposed algorithms.

Negative sampling strategies are widely used in implicit collaborative filtering to address issues like data sparsity and class imbalance. However, these methods often introduce false negatives, hindering the model's ability to accurately learn users' latent preferences. To mitigate this problem, existing methods adjust the negative sampling distribution based on statistical features from model training or the hardness of negative samples. Nevertheless, these methods face two key limitations: (1) over-reliance on the model's current representation capabilities; (2) failure to leverage the potential of false negatives as latent positive samples to guide model learning of user preferences more accurately. To address the above issues, we propose a Topology-aware Positive Sample Set Construction and Feature Optimization method (TPSC-FO). First, we design a simple topological community-aware false negative identification (FNI) method and observe that topological community structures in interaction networks can effectively identify false negatives. Motivated by this, we develop a topology-aware positive sample set construction module. This module employs a differential community detection strategy to capture topological community structures in implicit feedback, coupled with personalized noise filtration to reliably identify false negatives and convert them into positive samples. Additionally, we introduce a neighborhood-guided feature optimization module that refines positive sample features by incorporating neighborhood features in the embedding space, effectively mitigating noise in the positive samples. Extensive experiments on five real-world datasets and two synthetic datasets validate the effectiveness of TPSC-FO.

Interval-aware Approximate Nearest Neighbor (ANN) search arises in applications where each object is associated with a numeric value or interval, and queries must satisfy both vector-similarity and interval constraints. Existing methods are typically tailored to a single query semantics, such as interval-filtered ANN search, and therefore require multiple specialized indexes to support diverse workloads, leading to substantial indexing and memory overhead. To address this limitation, we propose the Unified Interval-aware Relative Neighborhood Graph (URNG), a unified graph framework for interval-aware ANN search. URNG preserves the monotonic searchability of relative-neighborhood-graph based ANN indexes while additionally ensuring structural heredity over query-induced subgraphs, enabling a single index to support multiple interval-aware query semantics. Building on this framework, we develop UG, a practical graph index that efficiently approximates URNG through unified interval-aware pruning and iterative repair, together with a query algorithm for interval-aware ANN search. Extensive experiments on 5 datasets show that UG consistently achieves a strong accuracy-efficiency trade-off across diverse interval-aware workloads while maintaining competitive index construction cost and memory usage.

Attributed community search (ACS) aims to identify subgraphs satisfying both structural cohesiveness and attribute homogeneity in attributed graphs, given a query consisting of query nodes and query attributes. Previously, algorithmic approaches deal with ACS through a two-stage paradigm, which suffers from structural inflexibility and attribute irrelevance. To overcome these limitations, learning-based approaches have recently been proposed to learn both structures and attributes simultaneously as a one-stage paradigm. However, these approaches train a transductive model that assumes the graph used for inference on unseen queries is the same as the graph used for training. That limits the generalization and adaptation of these approaches to different heterogeneous graphs. In this paper, we propose a new framework, Inductive Attributed Community Search, ^+ , based on inductive learning, which can infer new queries for different communities and graphs. Specifically, ^+ employs an encoder-decoder neural architecture to handle one ACS task at a time, where a task consists of a graph with only a few queries and their corresponding ground-truth. We design a three-phase workflow, ‘training, adaptation, inference refinement’, that learns a shared model to absorb and induce prior effective common knowledge about ACS across different tasks. The shared model can then swiftly adapt to a new task with a small number of ground-truth labels. We conduct substantial experiments on 8 real-world datasets to verify the effectiveness of ^+ . Our approach ^+ achieves average absolute improvements of 29.96% in -score for ACS tasks.

Spatial crowdsourcing platforms have become indispensable in addressing the evolving needs of modern society. These platforms facilitate essential services such as ride-sharing, on-demand food delivery, and efficient parcel distribution. However, the uneven distribution of workers and requests under a single-platform setting may lead to the loss of tasks. To address this issue, we introduce the Cross Online Matching (COM) problem, which facilitates collaboration among multiple platforms. We first propose DemCOM and RamCOM, which adopt deterministic greedy and randomized trade-off strategies, respectively. Furthermore, we develop a Utility-Distribution Aware Cooperative Online Matching (UDACOM) algorithm that leverages supply-demand relationships to optimize decision-making. Theoretical analysis confirms the competitive ratios of our algorithms. Validated on both real and synthetic datasets, our approach significantly outperforms state-of-the-art methods, achieving a 5% increase in total revenue and a 3% improvement in the successful matching rate.

Computing graph-propagation based node similarities is a fundamental operator in many graph mining and graph learning tasks. The state-of-the-art approach to compute the graphpropagation based similarity is based on a push-style iterative framework. The push framework is very efficient when the resulting node similarity vector pi has a smallL1-norm (e.g., personalized PageRank and heat kernel PageRank). However, we find that when pi has a large L1-norm (e.g., Katz scores and exponential communicability), such a framework is inefficient. To overcome this issue, we propose a novel framework, called AdaPush, which is more efficient and flexible than the state-of-the-art (SOTA) framework. Based on the AdaPushframework, we develop two new algorithms with two different carefully-designed randomized acceleration techniques, respectively. We prove that both of our new algorithms can achieve a relative-error guarantee. Additionally, a striking feature of our algorithms is that their time complexity is insensitive to ||pi||(1), thus they are efficient even when ||pi||(1) is large. Extensive experiments on 5 large real-life datasets demonstrate that our algorithms substantially outperform the SOTA algorithms for computing Katz score and exponential communicability in terms of both running time and estimation accuracy.