Tianle Liu, Zhiliang Tian, Zhen Huang, Tianlun Liu, Jingyuan Huang, Zhaoning Zhang, Chengcheng Shao, Dongsheng Li. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Rumor detection is essential for building a responsible web and internet ecosystem, which has attracted significant attention from the research community. However,emerging topic rumor detection, i.e., identify rumors at the early stages of a topic's emergence where only limited discussions can be observed, still remains a challenge. Technically, this scenario is accompanied by the issues ofdata scarcityon emerging topics and thedata distribution discrepancybetween old topics and emerging new topic. In this paper, we propose a new framework termedLLM-drivenADversarialExampleSynthesis (LADES) for emerging topic rumor detection. LADES utilizes Large Language Models (LLMs) for generating readable and contextually coherent adversarial examples. The generated adversarial examples not only expand the training set to tackle the data scarcity issue, but also act as a bridge to connect the data distribution of old and new topics. To overcome training instability in adversarial example generation, LADES introduces a gradient-free Markov Chain Monte Carlo (MCMC) sampling method. This method ensures adversarial examples are readable and contextually coherent by harnessing LLMs, while promoting effective attacks through entropy-based sampling that targets model uncertainty. To mitigate the impact of potential mislabeling in synthetic data, LADES implements a meta-mixed-learning mechanism. This mechanism dynamically adjusts the weights of synthetic adversarial examples, guided by limited labeled data from emerging topics, thereby alleviating the data noise.

Dissolved Oxygen (DO) is an important indicator of water quality and sea-air interactions, and the formation mechanism and dramatic change of Oxygen Minimum Zones (OMZs) are closely related to the activities of marine organisms and marine ecological events. However, due to the sparsity and scarcity of DO data, the studies on OMZs get a lot of limits. This paper proposes a deep learning method called SE-LeNet to reconstruct underwater multilayer DO, which integrates the channel attention module called SE-Block into the LeNet model. Simultaneously, the relationship between sea surface elements (e.g. spCO2, PhyC) and DO are created based on CEMES reanalysis data, which has never been considered in the reconstructed model before. Extensive experiments are conducted, demonstrating that the deep learning methods notably outperform the traditional statistical and other machine learning methods when processing high-resolution multi-dimensional data with spatial information. Moreover, the result accuracy is improved after adding the SEBlock module, and the R2and RMSE of SE-LeNet are as low as 0.97 and 12.00 μmol/kg, respectively. SE-LeNet can be successfully used in reconstructing DO data and then promoting the studies on OMZs in the ocean field.

The extreme fast charging (XFC) technology helps to reduce refueling time, alleviate mile anxiety, extend driving range and finally promote the popularity of electric vehicles (EVs). However, it would also pose great challenges on the power grid infrastructure especially distribution networks, due to the large-scale and intermittent power demand. This paper proposes a coordinated planning method for power distribution networks and XFC EV charging stations, with the on-site batteries considered. Firstly, considering the traffic flow pattern, the operation of XFC stations is analyzed on both energy and power demand. Secondly, the coordinated planning model is developed to satisfy the time-varying XFC load, with both transportation and electricity constraints considered. In addition, the on-site batteries are introduced to flatten the XFC energy used and supplement its power supply. The case studies have verified the effectiveness of the proposed method. The influence of XFC on the distribution networks and the effects of the on-site storage are also studied.

Summary Hundreds of millions of Chinese people have become social network users in recent years, and aligning the accounts of common Chinese users across multiple social networks is valuable to many inter‐network applications, for example, cross‐network recommendation and cross‐network link prediction. Many methods have explored the proper ways of utilizing account name information into aligning the common English users' accounts. However, how to properly utilize the account name information when aligning the Chinese user accounts remains to be detailedly studied. In this article, we first discuss the available naming behavioral models as well as the related features for different types of Chinese account name matchings. Second, we propose the framework of Multi‐View Cross‐Network User Alignment (MCUA) method, which uses a multi‐view framework to creatively integrate different models to deal with different types of Chinese account name matchings, and can consider all of the studied features when aligning the Chinese user accounts. Finally, we conduct experiments to prove that MCUA can outperform many existing methods on aligning Chinese user accounts between Sina Weibo and Twitter. Besides, we also study the best learning models and the top‐k valuable features of different types of name matchings for MCUA over our experimental datasets.

The popularity of online social network (OSN) services has given rise to a variety of social network applications. However these applications lack a common platform for information sharing and people interoperating. In this paper, we propose SocAware, a middleware designed for OSN services. SocAware extracts social relation from heterogeneous networks, and builds a uniform knowledge base to manage the social information. SocAware distinguishes itself from other OSN middlewares by analyzing the OSN activities to classify the social relations, and by calculating the strength of social relations to provide reusability between social applications. We also provide a set of API to facilitate third-party application development and the effective utilization of these relations. In order to validate SocAware, we developed two prototype applications above the middleware. The experimental results demonstrate the usability and expansibility of the middleware.