Developing a universal graph model capable of generalizing across diverse graph domains has consistently been a key objective in graph learning. Recently, many studies have focused on achieving in-context learning (ICL) on graphs, which can generalize to novel tasks without the need for fine-tuning, similar to large language models (LLMs) such as GPT-3. These researches can be primarily divided into graph-based methods and LLM-based methods. However, the generalization performance of the former is limited by the representation capability of GNNs, while the latter faces the challenge of LLMs understanding graph structures. Therefore, we propose CAGML, a context-aware graph meta-learning model, which learns to generalize to cross-domain and cross-granularity graph tasks using a meta-trained Transformer. Firstly, we formulate graph few-shot learning tasks as a structure-aware sequence modeling problem to unify cross-domain and cross-granularity tasks. Then, a structure-aware Transformer (SAT) is introduced as a graph in-context learner to make predictions with a few labels and the task-specific structural context. Finally, we pre-train SAT in a meta-optimization manner on large-scale citation network and knowledge graph. Experiments on 6 cross-domain graph datasets show that, without fine-tuning, CAGML can achieve state-of-the-art (SOTA) performance in terms of average performance across cross-granularity tasks on adopted datasets.

Nowadays, research of Text Classification (TC) based on graph neural networks (GNNs) is on the rise. Both inductive methods and transductive methods have made significant progress. For transductive methods, the semantic interaction between texts plays a crucial role in the learning of effective text representations. However, it is difficult to perform inductive learning while modeling interactions between texts on the graph. To give a universal solution, we propose the graph neural network based on pre-trained semantic interaction called PaSIG. Firstly, we construct a text-word heterogeneity graph and design an asymmetric structure to ensure one-way message passing from words to the test texts. Meanwhile, we use the context representation capability of the pre-trained language model to construct node features that contain classification semantic information. Afterward, we explore the adaptative aggregation methods with a gated fusion mechanism. Extensive experiments on five datasets have shown the effectiveness of PaSIG, with the accuracy exceeding the baseline by 2.7% on average. While achieving state-of-the-art performance, we have also taken measures of subgraph sampling and intermediate state preservation to achieve fast inference.

Mining geographic location of social media users is a crucial technology for realizing the mapping of cyberspace to geographical world, which can provide strong support for wide-ranging location-based services. As a typical approach, user geolocation methods based on relationships rely on the assumption of location homophily between users and their neighbors. However, these methods only utilize the geographic influence between pair-wise relationship, resulting in undesired geolocation performance. In this paper, a social media user geolocation method based on geographically compact social subgraphs (SMUG-GCS) is proposed. Firstly, we analyze the relationship pattern among users in geographic proximity, and find a phenomenon that users who are geographically close tend to have tightly social groups. Based on this finding, a subgraph partitioning algorithm is presented which integrates structure compactness and geographical credibility to identify a set of subgraphs, whose nodes are more tightly connected and geographically proximity. Finally, user locations are inferred using the propagation of user information only based on the geographically compact subgraph. Extensive experiments are conducted on three real-world social media datasets. The results show that, compared with 5 typical relationship-based methods, SMUG-GCS improves the geolocating accuracy while reducing storage costs, leading to a significant reduction in median error distance ranging from 26.7% to 82.9%, as well as decrease in storage requirements by up to 56.5%.

Reasoning over knowledge graphs has attracted considerable attention from researchers and is being widely applied to contribute question answering systems, recommender systems, and other information retrieval systems. However, existing reasoning methods tend to suffer from poor interpretability which is not consistent with human commonsense. The trustworthiness and reliability of the knowledge discover outcomes thus decreased as a result. Inspired by the process of human decision-making, we propose a commonsense-aware hierarchical framework calledHDLH, which incorporates commonsense knowledge into hierarchical knowledge graph reasoning process with deep reinforcement learning.HDLHimplements hierarchical reasoning process through exploration and exploitation sequentially by applying multi-agent reinforcement learning. Multiple agents inHDLHsimulate the multi-level decision-making ability of humans, and reason hierarchically and reasonably to maintain its efficiency and interpretability. Moreover, commonsense knowledge is incorporated by means of the reward-shaping function, ultimately guiding the agent to reason more consistently with human perceptions and reduce the huge search space. We evaluatedHDLHwith various tasks on five real-world datasets. The experimental results reveal thatHDLHachieves better performance compared with state-of-the-art baseline models.

User location discovery from social media is crucial for location-based services such as emergency awareness and event monitoring. Existing approaches generally integrate user-generated text features and social relationships but insufficiently explore location-specific features and geographically proximate relationships, leading to suboptimal accuracy. In this article, we propose a Twitter user geolocation method based on location feature enhancement to better capture the location characteristics in users’ tweets and social relationships. Specifically, a user tweet representation algorithm based on location feature separation (TwLS) is designed. By leveraging words’ location-aware weight matrix and pre-trained embeddings, TwLS calculates a tweet representation for each user in every location, explicitly indicating the relevance between users and various locations. Additionally, we develop the local celebrity discovery method (LocCel) to construct social networks by identifying and preserving geographically concentrated high-degree nodes while filtering noise. Thereby, LocCel enhances local relationships and strengthens location-proximate connections within the user social network. Experiments on two real-world datasets show that our method outperforms seven baselines, improving user geolocation accuracy by 3.1% ∼ 8.1% and 1.8% ∼ 8.8%, while reducing median error by 22.2% ∼ 52.8% and 19.4% ∼ 50.7%, respectively.

In the existing unsupervised domain adaptation (UDA) methods for remote sensing images (RSIs) semantic segmentation, class symmetry is a widely followed ideal assumption, where the source and target RSIs have exactly the same class space. In practice, however, it is often very difficult to find a source RSI with exactly the same classes as the target RSI. More commonly, there are multiple source RSIs available. And there is always an intersection or inclusion relationship between the class spaces of each source–target pair, which can be referred to as class asymmetry. Nevertheless, the class asymmetry domain adaptation segmentation of RSIs with multiple sources has not yet been explored. To this end, a novel class asymmetry RSIs domain adaptation method is proposed for the first time in this article, which consists of four key components. First, a multibranch segmentation network is built to learn an expert for each source RSI. Second, a novel collaborative learning method with the cross-domain mixing strategy is proposed, to supplement the class information for each source while achieving the domain adaptation of each source–target pair. Third, a pseudolabel generation strategy is proposed to effectively combine the strengths of different experts, which can be flexibly applied to two cases where the source class union is equal to or includes the target class set. Fourth, a multiview-enhanced knowledge integration module is developed for high-level knowledge routing and transfer from multiple domains to target predictions. The experimental results of six different class settings on airborne and spaceborne RSIs show that the proposed method can effectively perform the multisource domain adaptation in the case of class asymmetry, and the obtained segmentation performance of target RSIs is significantly better than the existing relevant methods.

The goal of stance detection is to automatically recognize the author's expressed attitude in text towards a given target. However, social media users often express themselves briefly and implicitly, which leads to a significant number of comments lacking explicit reference information to the target, posing a challenge for stance detection. To address the missing relationship between text and target, existing studies primarily focus on incorporating external knowledge, which inevitably introduces noise information. In contrast to their work, we are dedicated to mining implicit relational information within data. Typically, users tend to emphasize their attitudes towards a relevant topic or aspect of the target while concealing others when expressing opinions. Motivated by this phenomenon, we suggest that the potential correlation between text and target can be learned from instances with similar topics. Therefore, we design a pretext task to mine the topic associations between samples and model this topic association as a dynamic weight introduced into contrastive learning. In this way, we can selectively cluster samples that have similar topics and consistent stances, while enlarging the gap between samples with different stances in the feature space. Additionally, we propose a nearest-neighbor prediction mechanism for stance classification to better utilize the features we constructed. Our experiments on two datasets demonstrate the advanced and generalization ability of our method, yielding the state-of-the-art results.

As the intelligent software, the development and application of large language models are extremely hot topics recently, bringing tremendous changes to general AI and software industry. Nonetheless, large language models, especially open source ones, incontrollably suffer from some potential software quality issues such as instability, inaccuracy, and insecurity, making software testing necessary. In this paper, we propose the first solution for functional testing of open large language models to check full-scene availability and conclude empirical principles for better steering large language models, particularly considering their black box and intelligence properties. Specifically, we focus on the model’s causal reasoning ability, which is the core of artificial intelligence but almost ignored by most previous work. First, for comprehensive evaluation, we deconstruct the causal reasoning capability into five dimensions and summary the forms of causal reasoning task as causality identification and causality matching. Then, rich datasets are introduced and further modified to generate test cases along with different ability dimensions and task forms to improve the testing integrity. Moreover, we explore the ability boundary of open large language models in two usage modes: prompting and lightweight fine-tuning. Our work conducts comprehensive functional testing on the causal reasoning ability of open large language models, establishes benchmarks, and derives empirical insights for practical usage. The proposed testing solution can be transferred to other similar evaluation tasks as a general framework for large language models or their derivations.

Automatic source code summarization system aims to generate a valuable natural language description for a program, which can facilitate software development and maintenance, code categorization, and retrieval. However, previous sequence-based research did not consider the long-distance dependence and highly structured characteristics of source code simultaneously. In this article, we present a Transformer-based Graph-Augmented Source Code Summarization (GA-SCS), which can effectively incorporate inherent structural and textual features of source code to generate an effective code description. Specifically, we develop a graph-based structure feature extraction scheme leveraging abstract syntax tree and graph attention networks to mine global syntactic information. And then, to take full advantage of the lexical and syntactic information of code snippets, we extend the original attention to a syntax-informed self-attention mechanism in our encoder. In the training process, we also adopt a reinforcement learning strategy to enhance the readability and informativity of generated code summaries. We utilize the Java dataset and Python dataset to evaluate the performance of different models. Experimental results demonstrate that our GA-SCS model outperforms all competitive methods on BLEU, METEOR, ROUGE, and human evaluations.

Information diffusion prediction aims to predict the tendency of information spreading in the network. Previous methods focus on extracting chronological features from diffusion paths and leverage relations in social graph as side information to facilitate diffusion prediction. However, abundant high-order social relations in information diffusion have not been sufficiently utilized, such as co-repose and co-following which can further mine potential user common preferences. In this paper, we construct a heterogeneous diffusion network (HDN) from the social graph and information cascades to model the high-order social relations in information diffusion. Then, we design a novel model named Multi-Channel Recurrent Graph Convolutional Network (MC-RGCN), which can extract high-order social relation semantics from the channels of HDN to promote prediction performance. In each channel, we depict a specific social relations from the views of global topology, pairwise strength, and local structure. Finally, we conduct extensive experiments on three real-world datasets, and the results show that our proposed method outperforms the state-of-the-art models on diffusion prediction.