Optimizing recommender systems for objectives beyond accuracy, such as diversity, novelty, and personalization, is crucial for long-term user satisfaction. To this end, the industry has accumulated vast amounts of structured domain knowledge, which we term human priors (e.g., item taxonomies, temporal patterns). This knowledge is typically applied through post-hoc adjustments during ranking or post-ranking. However, this approach remains decoupled from the core model learning, which is particularly undesirable as the industry shifts to end-to-end generative recommendation foundation models. On the other hand, many methods targeting these beyond-accuracy objectives often require architecture-specific modifications and discard these valuable human priors by learning user intent in a fully unsupervised manner. Instead of discarding the human priors accumulated over years of practice, we introduce a backbone-agnostic framework that seamlessly integrates these human priors directly into the end-to-end training of generative recommenders. With lightweight, prior-conditioned adapter heads inspired by efficient LLM decoding strategies, our approach guides the model to disentangle user intent along human-understandable axes (e.g., interaction types, long- vs. short-term interests). We also introduce a hierarchical composition strategy for modeling complex interactions across different prior types. Extensive experiments on three large-scale datasets demonstrate that our method significantly enhances both accuracy and beyond-accuracy objectives. We also show that human priors allow the backbone model to more effectively leverage longer context lengths and larger model sizes.
论文检索
输入标题、作者或关键词，从 10,462 篇学术成果中精准定位
The surge in intelligent applications driven by large language models (LLMs) has made it increasingly difficult for bandwidth-limited cloud servers to process extensive LLM workloads in real time without compromising user data privacy. To solve these problems, recent research has focused on constructing cloud-edge consortia that integrate server-based LLM with small language models (SLMs) on mobile edge devices. Furthermore, designing collaborative training mechanisms within such consortia to enhance inference performance has emerged as a promising research direction. However, the cross-domain deployment of SLMs, coupled with structural heterogeneity in SLMs architectures, poses significant challenges to enhancing model performance. To this end, we propose Co-PLMs, a novel co-tuning framework for collaborative training of large and small language models, which integrates the process of structure-agnostic mutual learning to realize knowledge exchange between the heterogeneous language models. This framework employs distilled proxy models (DPMs) as bridges to enable collaborative training between the heterogeneous server-based LLM and on-device SLMs, while preserving the domain-specific insights of each device. The experimental results show that Co-PLMs outperforms state-of-the-art methods, achieving average increases of 5.38% in Rouge-L and 4.88% in EM. Our code has been released at https://github.com/papercode-DFL/Co-PLMs.
The rise of Internet of Vehicles (IoV) technologies is transforming traffic management from isolated control to a collective, multi-vehicle process. At the heart of this shift is multi-vehicle dynamic navigation, which requires simultaneously routing large fleets under evolving traffic conditions. Existing path search algorithms and reinforcement learning methods struggle to scale to city-wide networks, often failing to capture the nonlinear, stochastic, and coupled dynamics of urban traffic. To address these challenges, we propose CityNav, a hierarchical, LLM-powered framework for large-scale multi-vehicle navigation. CityNav integrates a global traffic allocation agent, which coordinates strategic traffic flow distribution across regions, with local navigation agents that generate locally adaptive routes aligned with global directives. To enable effective cooperation, we introduce a cooperative reasoning optimization mechanism, in which agents are jointly trained with a dual-reward structure: individual rewards promote per-vehicle efficiency, while shared rewards encourage network-wide coordination and congestion reduction. Extensive experiments on four real-world road networks of varying scales (up to 1.6 million roads and 430,000 intersections) and traffic datasets demonstrate that CityNav consistently outperforms nine classical path search and RL-based baselines in city-scale travel efficiency and congestion mitigation. Our results highlight the potential of LLMs to enable scalable, adaptive, and cooperative city-wide traffic navigation, providing a foundation for intelligent, large-scale vehicle routing in complex urban environments. Our project is available at https://github.com/usail-hkust/CityNav.
Cross-view geo-localization (CVGL) establishes correspondences between ground-level and satellite images of the same geographic location, serving as a fundamental technology for smart city applications, including autonomous navigation, urban planning, and location-based services. Current CVGL approaches fall into two categories: feature-based methods achieve superior performance through 2D representation learning but lack interpretability. Spatial-based methods provide geometric understanding and interpretable matching but suffer from limited spatial modeling and weak cross-view alignment, leading to lower performance. We reformulate CVGL from a spatial perspective and propose an auxiliary task-enhanced network. The network captures spatial semantics and provides explicit alignment processes with visualizable results. We introduce an auxiliary spatial semantic alignment (SSA) task that learns spatial structure via vision foundation models (VFM) and BEV transformation to enhance the primary CVGL task. The primary task captures visual semantics, including texture and appearance. Within this unified framework with shared encoders, the primary task enriches the learned embeddings by fusing spatial structure with visual semantics, yielding spatially complete representations. Extensive experiments on three standard CVGL benchmarks demonstrate that our method significantly surpasses previous spatial-based approaches while maintaining competitive performance with state-of-the-art (SOTA) feature-based methods, achieving 98.48% R@1 on CVUSA and 71.05% R@1 on CVACT\_test. We provide comprehensive analyses through pixel-level activation maps and feature-space UMAP visualizations to validate both effectiveness and interpretability.
Generative modeling offers a promising solution to data scarcity and privacy challenges in time series analysis. However, the structural complexity of time series, characterized by multi-scale temporal patterns and heterogeneous components, remains insufficiently addressed. In this work, we propose a structure-disentangled multiscale generation framework for time series. Our approach encodes sequences into discrete tokens at multiple temporal resolutions and performs autoregressive generation in a coarse-to-fine manner, thereby preserving hierarchical dependencies. To tackle structural heterogeneity, we introduce a dual-path VQ-VAE that disentangles trend and seasonal components, enabling the learning of semantically consistent latent representations. Additionally, we present a guidance-based reconstruction strategy, where coarse seasonal signals are utilized as priors to guide the reconstruction of fine-grained seasonal patterns. Experiments on six datasets show that our approach produces higher-quality time series than existing methods. Notably, our model achieves strong performance with a significantly reduced parameter count and exhibits superior capability in generating high-quality long-term sequences. Our implementation is available at https://github.com/decisionintelligence/TimeMAR.
The current paradigm of training large language models (LLMs) on public available Web data is becoming unsustainable as high-quality data sources in specialized domains near exhaustion. Federated Learning (FL) emerges as a practical solution for the next generation of AI on a decentralized Web, enabling privacy-preserving collaborative fine-tuning on decentralized private data. While Low-Rank Adaptation (LoRA) is standard for efficient fine-tuning, its federated application faces a critical bottleneck: communication overhead under heterogeneous network conditions. Structural redundancy in LoRA parameters increases communication costs and causes aggregation conflicts. To address this, we propose FedSRD, a Sparsify-Reconstruct-Decompose framework for communication-efficient federated LLM fine-tuning. We introduce importance-aware sparsification to reduce the upload parameter count while preserving the structural integrity of LoRA updates. The server aggregates updates in full-rank space to mitigate conflicts, then decomposes the global update into a sparse low-rank format for broadcast, ensuring a symmetrically efficient cycle. We also propose an efficient variant, FedSRD-e, to reduce computational overhead. Experiments on 10 benchmarks show our framework significantly reduces communication costs by up to 90% while improving performance on heterogeneous client data.
Federated Graph Learning (FGL) has emerged as an efficient paradigm to address the pronounced data heterogeneity common in real-world decentralized graph datasets, attracting significant interest from both academia and industry. Existing FGL methods often degrade the global model's performance due to a fundamental mismatch between the model's fixed-geometry embedding space and the diverse geometric structures of client data. To address this issue, we propose a novel framework for Personalized Federated Riemannian Graph Learning, namely FedRGL. FedRGL introduces a personalized mixed-curvature product space for each heterogeneous client, mapping local graph data into a tailored geometric space composed of Euclidean, hyperbolic, and spherical manifolds. Furthermore, it leverages a Ricci-Gated Graph Convolutional Network to dynamically adapt its message-passing mechanism to the local topology of each graph. Extensive experiments across diverse geometric heterogeneity settings demonstrate that FedRGL significantly outperforms state-of-the-art FGL methods in terms of model accuracy and generalization.
The World Wide Web increasingly relies on intelligent services that require accurate time series forecasting, from urban mobility platforms to adaptive web-based decision systems. In practice, building effective forecasting models typically requires abundant high-quality data, which may not always be available in all cities due to sensing limitations or data sparsity. To address this challenge, transfer learning methods aim to transfer knowledge from data-rich source cities to data-scarce target cities. However, source and target data distributions are often not identical: while some patterns from source cities may be beneficial, others can be irrelevant or even misleading. Existing transfer learning methods generally train the target model using all available source data without explicitly distinguishing between useful and non-useful knowledge, which may hinder performance. In this work, we propose xRAG4TS, a novel framework that integrates Retrieval-Augmented Generation (RAG) with Large Language Models (LLMs) for cross-city time series forecasting. xRAG4TS introduces a Cross-City Selective Retriever Module that filters semantically relevant historical patterns and documents from data-rich source cities, and incorporates them as structured prompts in an LLM Inference Module to guide forecasting in data-scarce target cities. By enabling selective, interpretable, and context-aware knowledge transfer, our method enhances robustness and scalability in web-oriented spatio-temporal applications. Extensive experiments on real-world cross-city datasets demonstrate that xRAG4TS significantly outperforms state-of-the-art baselines, highlighting its potential for powering adaptive and trustworthy web services under severe data scarcity.
Embedded AI applications usually require compact on-device models that can continually adapt to new tasks. However, recent studies have revealed that neural networks trained on non-stationary data streams gradually lose their ability to adapt to new tasks, a phenomenon known as plasticity loss. Moreover, to enable neural networks to run on resource-constrained embedded devices, model pruning is commonly applied for model compression, which may further affect their plasticity. To conduct efficient model adaptation on new tasks on embedded devices, we propose Plasticity-aware Continual Pruning (PaCP), a novel framework that operates in two stages. First, a pre-deployment stage uses a plasticity-aware strategy to prune the model while optimizing its initial structure for future adaptability. Second, during continual learning, the model's capacity is temporarily expanded at task boundaries to efficiently learn new information, before plasticity-aware pruning restores its compact form. Extensive experiments on multiple continual learning benchmarks demonstrate that PaCP significantly outperforms existing plasticity-maintenance methods and, remarkably, even surpasses non-pruned models lacking explicit plasticity preservation.
Relational databases (RDBs) underpin the majority of global data management systems, where information is structured into multiple interdependent tables. In social media platforms, for instance, massive user-generated data are organized across related tables such as users, posts, comments, and interactions, enabling large-scale analysis and predictive modeling of social behaviors. To effectively use the knowledge within RDBs for predictive tasks, recent advances leverage graph representation learning to capture complex inter-table relations as multi-hop dependencies. Despite achieving state-of-the-art performance, these methods remain hindered by the prohibitive storage overhead and excessive training time, due to the massive scale of database and the computational burden of intensive message passing across interconnected tables. To alleviate these concerns, we propose and study the problem of Relational Database Distillation (RDD). Specifically, we aim to distill large-scale RDBs into compact heterogeneous graphs while retaining the predictive power (i.e., utility) required for training graph-based models. Multi-modal column information is preserved through node features, and primary–foreign key relations are encoded via heterogeneous edges, thereby maintaining both data fidelity and relational structure. To ensure adaptability across diverse downstream tasks without engaging the traditional, inefficient bi-level distillation framework, we further design a kernel ridge regression-guided objective with pseudo-labels, which produces quality features for the distilled graph. Extensive experiments on multiple real-world RDBs demonstrate that our solution substantially reduces the data size while maintaining competitive performances on classification and regression tasks, creating an effective pathway for scalable learning with RDBs.
They Said Memes Were Harmless - We Found the Ones That Hurt: Decoding Jokes, Symbols, and Cultural References
PDF ↗Meme-based social abuse detection is challenging because harmful intent often relies on implicit cultural symbolism and subtle cross-modal incongruence. Prior approaches, from fusion-based methods to in-context learning with Large Vision-Language Models (LVLMs), have made progress but remain limited by three factors: i) cultural blindness (missing symbolic context), ii) boundary ambiguity (satire vs. abuse confusion), and iii) lack of interpretability (opaque model reasoning). We introduce CROSS-ALIGN+, a three-stage framework that systematically addresses these limitations: (1) Stage I mitigates cultural blindness by enriching multimodal representations with structured knowledge from ConceptNet, Wikidata, and Hatebase; (2) Stage II reduces boundary ambiguity through parameter-efficient LoRA adapters that sharpen decision boundaries; and (3) Stage III enhances interpretability by generating cascaded explanations. Extensive experiments on five benchmarks and eight LVLMs demonstrate that CROSS-ALIGN+ consistently outperforms state-of-the-art methods, achieving up to 17% relative F1 improvement while providing interpretable justifications for each decision.
Detecting social bots is critical to ensuring the security of online discourse and maintaining trust in social networks. Early feature-based and text-based methods often fail against bots that mimic human behavior, and graph-based approaches have emerged to better exploit structural signals. However, most existing Graph Neural Networks (GNNs) still focus on pairwise connections, overlooking higher-order relational patterns, and their multi-relation fusion strategies are typically simplistic, ignoring dependencies between relations and user-specific preferences. To overcome these limitations, we propose MPS-Bot, a model that integrates higher-order structure modeling with user-specific cross-relation dependency learning. MPS-Bot introduces a simplex convolutional layer that leverages simplexes derived from network structures to capture group coordination patterns beyond pairwise connections. In addition, a cross-relation dependency attention mechanism adaptively fuses relation-specific representations according to each user's relational preferences, leading to more discriminative and robust multi-relation representations. Extensive experiments on two widely used Twitter bot detection benchmarks, MGTAB and TwiBot-22, show that MPS-Bot generally outperforms state-of-the-art baselines. These findings highlight the effectiveness of higher-dimensional message passing over simplexes as a powerful approach to unmasking bots in social networks.
Abnormal user detection has been a critical and widely studied research problem in social networks since these users can create significant risks to platform security and privacy leakage. Currently, graph-based models are commonly used for exploring the structured social network data and temporally dynamic user interactions, leading to significant advances in dynamic heterogeneous graph-based abnormal user detection. However, most existing approaches are correlation-driven and lack the ability to separate stable patterns from transient noise. Furthermore, these methods are highly dependent on inherent labels and fail to detect common few-shot anomalies in social networks. To address these issues, we propose CIFAD, a Causal-Invariant Few-shot Anomaly Detection method that improves few-shot anomaly detection with an active annotation strategy. Specifically, CIFAD first integrates a sparse lagged attention encoder to model multi-relational temporal interactions. Furthermore, it introduces causal-invariant subspace decomposition to disentangle stable causal signals from dynamic environmental noise and improve generalization. Finally, it designs an active annotation strategy based on influence functions and coverage optimization to maximize the utility of limited labels in a closed-loop process. Extensive experiments on multiple real-world social network datasets demonstrate that our method achieves higher accuracy than state-of-the-art methods, validating its robustness in abnormal user detection for social networks.
Illicit activities such as financial fraud and fake promotions are increasingly prevalent on social applications. The diverse behavioral structures of different types of illicit accounts pose difficulties in designing unified detection strategies. Unlike existing studies that focus on binary detection and often rely on user static profiles, we introduce a novel task: illicit account detection based solely on user behavior sequences. This task presents two key challenges: 1) illicit accounts often mimic benign users by performing normal-looking behavior subsequences, and 2) behaviors with the same action (e.g., add-friend, initiate-transaction) can serve different purposes, such as illicit or benign. To address these challenges, we propose a Pattern-aware Illicit Accounts Detection (PIAD) framework that consists of three components: 1) a dual-perspective pattern mining module that extracts category-specific self- and interaction-behavior patterns from behavior sequences to capture distinct behavioral regularities across different user types; 2) a contextualized action semantic encoding algorithm that aligns action codings with contextual dependencies among behaviors within user sequences to capture variations in purposes when behaviors with the same actions occur under different contexts; and 3) a pattern-aware fusion model that integrates the mined patterns with the context and interaction in behavior sequences to learn discriminative representations for detection. Extensive experiments on real-world datasets demonstrate that PIAD consistently outperforms state-of-the-art baselines with an average 7.51% improvement on F1 score.
Social event prediction has garnered increasing attention in web-centered society. Most existing studies represent web-based event stream as chronological graph sequences, then leverage RNNs and GNNs to model temporal and relational patterns. However, this paradigm is inherently flawed: (1) RNNs struggle to capture long-term temporal dependencies, ignoring those temporally distant but influential events. (2) Spatio-temporal GNNs exhibit high computational complexity on large-scale real-time event streams, which hinders their web applications. To this end, we explore a novel paradigm called Fourier Graph Learning from the perspective of frequency domain. Specifically, we first define a novel data structure called Fourier Graph (FG). In FG, both nodes and edges are complex vectors, with real part encoding semantics and imaginary part representing semantic-specific temporal patterns. These temporal patterns are obtained by semantic-aware frequency filter, which utilizes semantics as guidance to adaptively incorporates both long-term dependency and short-term dynamic. Based on FG, we further propose Fourier Graph Neural Network (FGNN). It replaces time-domain convolution with frequency-domain multiplication for efficient aggregation. FGNN also includes a complex-valued event decoder, which fully leverages semantics and temporal patterns from complex space to predict future event probabilities. Extensive experiments show our superior performance with higher accuracy, less complexity and better interpretability compared with baselines.
With the growing scale of social media, social event detection and evolution modeling have attracted increasing attention. Graph neural networks (GNNs) and transformer-based pre-trained language models (PLMs) have become mainstream approaches in this area. However, existing methods still face three major challenges. First, the sheer volume of social media messages makes learning resource-intensive. Second, the fragmentation of social media messages often impedes the model's ability to capture a comprehensive view of the events. Third, the lack of structured temporal context has hindered the development of effective models for event evolution, limiting users' access to event information. To address these challenges, we propose a foundation model for unsupervised Social Event Detection and Evolution, namely RagSEDE. Specifically, RagSEDE introduces a representativeness- and diversity-driven sampling strategy to extract key messages from massive social streams, significantly reducing noise and computational overhead. It further establishes a novel paradigm based on Retrieval Augmented Generation (RAG) that enhances PLMs in detecting events while simultaneously constructing and maintaining an evolving event knowledge base. Finally, RagSEDE leverages structural information theory to dynamically model event evolution keywords for the first time. Extensive experiments on two public datasets demonstrate the superiority of RagSEDE in open-world social event detection and evolution.
Graph neural networks (GNNs) have become the cornerstone of graph representation learning in real-world applications, especially in social networks such as Facebook and Twitter, where user interactions naturally form graph-structured data. To safeguard sensitive information, including user attributes and personal connections, locally private graph learning protocols employ local differential privacy (LDP) to provide rigorous user-level guarantees. Despite these protections, we show that such protocols remain highly vulnerable to data poisoning attacks. Our approach targets the full-privacy setting, where both node features and edges are LDP-protected, and executes coordinated manipulations that significantly degrade utility, such as node classification accuracy, across multiple social network benchmarks. To counter these threats, we propose CureNet, a defense framework with four key components: (1) local data perturbation for privacy, (2) trimmed screening to filter abnormal submissions, (3) privacy-aware fake node pruning to remove sophisticated adversaries, and (4) a utility enhancement module to recover graph learning performance under privacy constraints. Extensive experiments on four real-world social network datasets—LastFM, Twitch, Github, and Facebook—demonstrate both the severe impact of our attack and the effectiveness of our defense. This study underscores the critical need to secure privacy-preserving graph learning against sophisticated poisoning threats.
Spamming activities (e.g., fake reviews, click farming, and deceptive content promotion) are increasingly conducted through collusive groups that exploit collective dynamics to manipulate platform metrics and mislead users, posing serious threats to the fairness, credibility, and functionality of online systems. To counteract these harmful behaviors, the task of spam detection has emerged as a critical area of research. However, existing detection methods generally remain limited in three key aspects: (i) They treat detection as a standard classification task, where representation learning and optimization are loosely coupled and suboptimal for capturing complex behaviors; (ii) They rely primarily on individual-level representation modeling, making it difficult to detect collective cheating strategies; (iii) They lack dedicated objective functions explicitly designed to characterize group-level spamming activities. To overcome these limitations, we introduce a collusion-aware Set-level learning framework (SetDet) that redefines the spam Detection task as a unified setwise optimization problem. Our approach offers three core advantages: (i) It enables end-to-end optimization by jointly learning representations and performing detection in a single, integrated process; (ii) It incorporates a model-level design for collusion representation, effectively capturing the temporal and relational patterns of coordinated spam; (iii) It pioneers a dedicated set-level optimization criterion that aligns closely with the structural characteristics of group-based cheating behaviors and accounts for class imbalance in real-world scenarios. Extensive experiments confirm the generalizability and superior performance of our framework across diverse spam scenarios and collusion strategies.
Taxonomy completion is the task of integrating new concepts into an existing taxonomy by determining the appropriate hypernym--hyponym relations. Existing approaches often struggle with the inherent imbalance between leaf and non-leaf edges, which induces bias in representation learning. In this paper, we propose BLEND: Balanced and Leaf-Enhanced Dual Fine-Tuning for Taxonomy Completion, a novel framework designed to mitigate this inductive bias. Our method employs independent fine-tuning of two lightweight large language models (LLMs): one optimized with a leaf-focused objective and the other trained with a balanced focused strategy. To further enhance structural understanding, we apply contrastive learning over structure-encoded paths and introduce a combined loss function, enabling more robust representation of hierarchical relations. Extensive experiments on three real-world benchmark datasets demonstrate that BLEND achieves up to 9.32% improvement in recall or hit metrics compared to state-of-the-art approaches. Moreover, BLEND delivers efficient inference while outperforming the latest baseline COMI, highlighting its effectiveness for taxonomy completion tasks.
As social media and the World Wide Web become hubs for information dissemination, effectively organizing and understanding the vast amounts of dynamically evolving Web content is crucial. Knowledge graphs (KGs) provide a powerful framework for structuring this information. However, the rapid emergence of new hot topics, user relationships, and events in social media renders traditional static knowledge graph embedding (KGE) models rapidly outdated. Continual Knowledge Graph Embedding (CKGE) aims to address this issue, but existing methods commonly suffer from catastrophic forgetting, whereby older, but still valuable, information is lost when learning new knowledge (such as new memes or trending events). This means the model cannot effectively learn the evolution of the data. We propose a novel CKGE framework, BAKE. Unlike existing methods, BAKE formulates CKGE as a sequential Bayesian inference problem and utilizes the Bayesian posterior update principle as a natural continual learning strategy. This principle is insensitive to data order and provides theoretical guarantees to preserve prior knowledge as much as possible. Specifically, we treat each batch of new data as a Bayesian update to the model's prior. By maintaining the posterior distribution, the model effectively preserves earlier knowledge even as it evolves over multiple snapshots. Furthermore, to constrain the evolution of knowledge across snapshots, we introduce a continual clustering method that maintains the compact cluster structure of entity embeddings through a regularization term, ensuring semantic consistency while allowing controlled adaptation to new knowledge. We conduct extensive experiments on multiple CKGE benchmarks, which demonstrate that BAKE achieves the top performance in the vast majority of cases compared to existing approaches.