Zero-shot stance detection (ZSSD) seeks to determine the stance of text toward previously unseen targets, a task critical for analyzing dynamic and polarized online discourse with limited labeled data. While large language models (LLMs) offer zero-shot capabilities, prompting-based approaches often fall short in handling complex reasoning and lack robust generalization to novel targets. Meanwhile, LLM-enhanced methods still require substantial labeled data and struggle to move beyond instance-level patterns, limiting their interpretability and adaptability. Inspired by cognitive science, we propose the Cognitive Inductive Reasoning Framework (CIRF), a schema-driven method that bridges linguistic inputs and abstract reasoning via automatic induction and application of cognitive reasoning schemas. CIRF abstracts first-order logic patterns from raw text into multi-relational schema graphs in an unsupervised manner, and leverages a schema-enhanced graph kernel model to align input structures with schema templates for robust, interpretable zero-shot inference. Extensive experiments on SemEval-2016, VAST, and COVID-19-Stance benchmarks demonstrate that CIRF not only establishes new state-of-the-art results, but also achieves comparable performance with just 30% of the labeled data, demonstrating its strong generalization and efficiency in low-resource settings.

Fuqiang Niu, Zini Chen, Zhiyu Xie, Hu Huang, Qing Liao, Qianlong Wang, Genan Dai, Bowen Zhang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Graph classification is a pivotal challenge in machine learning, especially within the realm of graph-based data, given its importance in numerous real-world applications such as social network analysis, recommendation systems, and bioinformatics. Despite its significance, graph classification faces several hurdles, including adapting to diverse prediction tasks, training across multiple target domains, and handling small-sample prediction scenarios. Current methods often tackle these challenges individually, leading to fragmented solutions that lack a holistic approach to the overarching problem. In this paper, we propose an algorithm aimed at addressing the aforementioned challenges. By incorporating insights from various types of tasks, our method aims to enhance adaptability, scalability, and generalizability in graph classification. Motivated by the recognition that the underlying subgraph plays a crucial role in GNN prediction, while the remainder is task-irrelevant, we introduce the Core Knowledge Learning (CKL) framework for graph adaptation and scalability learning. CKL comprises several key modules, including the core subgraph knowledge submodule, graph domain adaptation module, and few-shot learning module for downstream tasks. Each module is tailored to tackle specific challenges in graph classification, such as domain shift, label inconsistencies, and data scarcity. By learning the core subgraph of the entire graph, we focus on the most pertinent features for task relevance. Consequently, our method offers benefits such as improved model performance, increased domain adaptability, and enhanced robustness to domain variations. Experimental results demonstrate significant performance enhancements achieved by our method compared to state-of-the-art approaches. Specifically, our method achieves notable improvements in accuracy and generalization across various datasets and evaluation metrics, underscoring its effectiveness in addressing the challenges of graph classification.

Topic evolution and stance dynamics are deeply intertwined in online social media, shaping the fragmentation and polarization of public discourse.Yet existing dynamic topic models and stance analysis approaches usually consider these processes in isolation, relying on abstractions that lack interpretability and agent-level behavioral fidelity.We present stance and topic evolution reasoning framework (SPARK), the first LLM-based multi-agent simulation framework for jointly modeling the co-evolution of topics and stances through natural language interactions.In SPARK, each agent is instantiated as an LLM persona with unique demographic and psychological traits, equipped with memory and reflective reasoning.Agents engage in daily conversations, adapt their stances, and organically introduce emergent subtopics, enabling interpretable, fine-grained simulation of discourse dynamics at scale.Experiments across five real-world domains show that SPARK captures key empirical patterns-such as rapid topic innovation in technology, domain-specific stance polarization, and the influence of personality on stance shifts and topic emergence.Our framework quantitatively reveals the bidirectional mechanisms by which stance shifts and topic evolution reinforce each other, a phenomenon rarely addressed in prior work.SPARK provides actionable insights and a scalable tool for understanding and mitigating polarization in online discourse.Code and simulation resources will be released after acceptance.

Zero-shot stance detection (ZSSD) is a challenging task that requires classifying stances towards unseen targets without large, well-curated training datasets. Existing data augmentation methods for ZSSD often suffer from semantic inconsistencies, hindering their effectiveness. To address these limitations, we propose a Logically Consistent Data Augmentation (LCDA) framework that enhances data quality by maintaining logical coherence. LCDA comprises two key components: data augmentation and logical graph convolutional network. Our data augmentation approach leverages chain-of-thought (CoT) prompting to enable large language models (LLMs) to distill training text into first-order logic (FOL) expressions, which provide a clear and concise representation of the underlying logic for stance prediction. These FOL expressions are then used to generate high-quality augmented samples. Furthermore, we introduce a novel logical graph convolutional network that effectively exploits FOL knowledge and combines it with augmented data samples to train a more accurate stance detection model. Our experiments on benchmark datasets demonstrate that the LCDA framework significantly outperforms existing ZSSD techniques, highlighting the effectiveness of integrating FOL into data augmentation for improving stance detection accuracy.

Human mobility nowcasting is crucial for public safety, especially during disasters when human mobility significantly differs from normal patterns, posing unique challenges. Recent studies have shown a correlation between disaster-related social media information and abnormal patterns in human mobility. However, these studies mainly focus on text counts while neglecting semantic text, which limits the effective use of social media data and reduces model prediction performance. The social text semantics reveal inherent non-pairwise relationships between regions in human mobility, posing a challenge to traditional graph neural network approaches. Thus, we propose a Semantics-Guided Dynamic Hypergraph Convolutional Network (SG-DyHGCN) for human mobility nowcasting in disaster. The model leverages semantic information to guide dynamic hyper-graph construction, enabling flexible adjustments to the hyper-graph structure, effectively capturing non-pairwise relationships between regions, and enhancing prediction performance. Experimental results validate the effectiveness of our method.

Next location prediction is a critical task in human mobility modeling, with broad applications in personalized recommendation, urban planning, and location-based services. Recently, researchers have used prompt-based large language models (LLMs) to improve next location prediction with pre-trained knowledge. However, they face inherent challenges in bridging the gap between textual prompts for semantic contextual understanding and human mobility data for transition pattern modeling. In this paper, we introduce SILO, a framework designed for Semantic Integration in LOcation prediction via LLMs. We first construct a hybrid semantic space that seamlessly integrates ID-based embeddings, text-derived semantics, and auxiliary contextual information, enabling comprehensive modeling of sequential mobility patterns alongside contextual nuances. We then propose user-centric prompts that specify the prediction task for LLMs while embedding user context within a special token. Further, we utilize LLMs as the prediction backbone to process both user-specific prompts and hybrid ID-context embeddings of location sequences. To enhance predictive performance, we finally introduce a dual-logits strategy, combining sequential transition logits with user profile-guided semantic preference logits. Extensive experiments on two large-scale real-world mobility datasets demonstrate that SILO significantly outperforms state-of-the-art baselines, validating its effectiveness in modeling complex mobility patterns through semantic integration using LLMs.

Crowd flows prediction is an important problem for traffic management and public safety. Graph Convolutional Network (GCN), known for its ability to effectively capture and utilize topological information, has demonstrated significant advancements in addressing this problem. However, GCN-based models were often based on predefined crowd-flow graphs via historical movement behaviors of human beings and traffic vehicles, which ignored the abnormal changes in crowd flows. In this study, we propose a multi-scale fusion GCN-based framework with Tucker decomposition named mTDNet to enhance dynamic GCN for crowd flows prediction. Following the paradigm of extant methods, we also employ the predefined crowd-flow graphs as a part of mTDNet to effectively capture the historical movement behaviors of crowd flows. To capture the abnormal changes, we propose a Tucker decomposition-based network with the product of the adjacency matrix of historical movement pattern graphs and an Adaptive Learning Tensor ( ALT ) by reconstructing the crowd flows. Particularly, we utilize the Tucker decomposition scheme to decompose ALT , which enhances the dynamic learning of graph structures, allowing for effective capturing of the dynamic changes in crowd flow, including abnormal changes. Furthermore, a multi-scale 3DGCN is utilized to mine and fuse the multi-scale spatio-temporal information from crowd flows, to further boost the mTDNet prediction performance. Experiments conducted on two real-world datasets showed that the proposed mTDNet surpasses other crowd flow prediction methods.

Sleep staging is imperative for evaluating sleep quality and diagnosing sleep disorders. Extant sleep staging methods with fusing multiple data-views of physiological signals have achieved promising results. However, they remain neglectful of the relationship among different data-views at different feature scales with view position-alignment. To address this, we propose a novel cross-view alignment network, termed cVAN, utilising scale-aware attention for sleep stages classification. Specifically, cVAN principally incorporates two sub-networks of a residual-like network which learn spectral information from time-frequency images and a transformer-like network which learns corresponding temporal information. The prime advantage of cVAN is to adaptively align the learned feature scales among the different data-views of physiological signals with a scale-aware attention by reorganizing feature maps. Extensive experiments on three public sleep datasets demonstrate that cVAN can achieve a new state-of-the-art result, which is superior to existing counterparts.

Stance detection, which aims to identify public opinion towards specific targets using social media data, is an important yet challenging task. With the proliferation of diverse multimodal social media content including text, and images multimodal stance detection (MSD) has become a crucial research area. However, existing MSD studies have focused on modeling stance within individual text-image pairs, overlooking the multi-party conversational contexts that naturally occur on social media. This limitation stems from a lack of datasets that authentically capture such conversational scenarios, hindering progress in conversational MSD. To address this, we introduce a new multimodal multi-turn conversational stance detection dataset (called MmMtCSD). To derive stances from this challenging dataset, we propose a novel multimodal large language model stance detection framework (MLLM-SD), that learns joint stance representations from textual and visual modalities. Experiments on MmMtCSD show state-of-the-art performance of our proposed MLLM-SD approach for multimodal stance detection. We believe that MmMtCSD will contribute to advancing real-world applications of stance detection research.

Traffic prediction is a classical spaial-temporal prediction problem with many real-world applications.In general, existing traffic prediction methods capture the complex spatial-­temporal features by iterative mechanism or non-iterative mechanism. However, the iterative mechanism often causes the prediction error accumulation and the non-­iterative mechanism is hard to capture the dynamic propagation information. The shortcomings of both mechanisms lead to their poor performance in long-­term prediction tasks. Target at the shortcomings of existing methods, in this paper, we propose a novel deep learning framework called Long-term Traffic Prediction based on Hybrid Model (LTPHM), which is designed to simulate the dynamic transmission process of traffic information on the road network by connecting the prediction values of the current step with the next step. Each spatial-­temporal module uses graph convolution (GCN) with an adaptive matrix to capture spatial dependence. Besides, we use Gated Dilated Convolution Networks (GDCN) and Gated Linear Unit convolution networks (GLU) to capture temporal dependence. Since LTPHM integrates the advantages of both iterative and non-iterative prediction, it can efficiently capture the complex and dynamic spatial-temporal features, especially the long-range temporal sequences. Experiments with three real-world traffic datasets demonstrate the effectiveness of our proposed model.

Traffic prediction is a classical spatial-temporal prediction problem with many real-world applications such as intelligent route planning, dynamic traffic management, and smart location-based applications. Due to the high nonlinearity and complexity of traffic data, deep learning approaches have attracted much interest in recent years. However, few methods are satisfied with both long and short-term prediction tasks. Target at the shortcomings of existing studies, in this paper, we propose a novel deep learning framework called Long Short-term Graph Convolutional Networks (LSGCN) to tackle both traffic prediction tasks. In our framework, we propose a new graph attention network called cosAtt, and integrate both cosAtt and graph convolution networks (GCN) into a spatial gated block. By the spatial gated block and gated linear units convolution (GLU), LSGCN can efficiently capture complex spatial-temporal features and obtain stable prediction results. Experiments with three real-world traffic datasets verify the effectiveness of LSGCN.

Optimal location querying (OLQ) in road networks is important for various applications. Existing work assumes no labels for servers and that a client only visits the nearest server. These assumptions are not realistic and it renders the existing work not useful in many cases. In this paper, we introduce the KOLQ problem which considers the k nearest servers of clients and labeled servers. We also proposed algorithms for the problem. Extensive experiments on the real road networks illustrate the efficiency of our proposed solutions.

Crowd flows prediction is an important problem of urban computing. The existing method adopts three deep residual networks to model spatio-temporal properties and achieves good prediction performance. However, since three separated network structures are used to model the properties, the time cost is often expensive for the existing method. In this paper, we propose an improved method to reduce the running time of the existing method by simplifying its architecture. In addition, we apply attention mechanism to make better use of temporal information. As shown in experiments, compared with the existing method, the improved method has significantly reduced running time and achieved better prediction performance.