While digital games cultivate rich shared ideologies, how players materialize these digital identities and communal values within physical, paratextual spaces remains underexplored. To address this, we present an ethnographic study investigating how gamers perform and negotiate their in-game identities through collaborative guestbook practices at the official Monster Hunter Bar in Tokyo. Through 4.5 hours of in-situ observation and analysis of eight physical "Hunter’s Notes," we examined visitors’ situated social behaviors and multi-modal inscriptions. Our findings reveal that players utilize multilingual handwriting and drawing as a social ritual, communicating through a shared visual vocabulary of game lore, such as weapon memes, NPC roleplay, and character illustrations, to bridge cultural barriers. By conceptualizing gamer inscription as a community-building performance, this work illustrates how digital foundations create powerful, asynchronous rituals, contributing novel insights into how game-centric societal norms reflect and challenge real-world communication conventions.

Large language models (LLMs) training or inference across multiple nodes introduces significant pressure on GPU memory and interconnect bandwidth. The Compute Express Link (CXL) shared memory pool offers a scalable solution by enabling memory sharing across nodes, reducing over-provisioning and improving resource utilization. We propose CXL-CCL, a collective communication library, leveraging the CXL shared memory pool to support cross-node GPU operations without relying on traditional RDMA-based networking. Our design addresses the challenges in synchronization, data interleaving, and communication parallelization faced by using the CXL shared memory pool for collective communications. Evaluating on multiple nodes with a TITAN-II CXL switch and six Micron CZ120 memory cards, we show that CXL-CCL achieves highly efficient collective operations across hosts, demonstrating CXL’s potential for scalable, memory-centric GPU communication. Our evaluation demonstrates that CXL-CCL achieves average performance improvements of 1.34 × for AllGather, 1.84 × for Broadcast, 1.94 × for Gather, and 1.07 × for Scatter, compared to the original RDMA-based implementation over 200 Gbps InfiniBand. In addition, an LLM training case study shows 1.11 × speedup compared with the InfiniBand while reducing interconnect hardware cost by 2.75 ×.

Qualitative coding is a demanding yet crucial research method in the field of Human–Computer Interaction (HCI). While recent studies have shown the capability of Large Language Models (LLMs) to perform qualitative coding within theoretical frameworks, their potential for collaborative human-LLM discovery and generation of new insights beyond initial theory remains underexplored. To bridge this gap, we proposed CHALET , a novel approach that harnesses the power of human-LLM partnership to advance theory-driven qualitative analysis by facilitating iterative coding, disagreement analysis, and conceptualization of qualitative data. We demonstrated CHALET ’s utility by applying it to the qualitative analysis of conversations related to mental-illness stigma, using the attribution model as the theoretical framework. Results highlighted the unique contribution of human-LLM collaboration in uncovering latent themes of stigma across the cognitive, emotional, and behavioral dimensions. We discuss the methodological implications of the human-LLM collaborative approach to theory-based qualitative analysis for the HCI community and beyond.

Diffusion models have shown promising ability in generating high-quality time series (TS) data. Despite the initial success, existing works mostly focus on the authenticity of data at the individual level, but pay less attention to preserving the population-level properties on the entire dataset. Such population-level properties include value distributions for each dimension and distributions of certain functional dependencies (e.g., cross-correlation, CC) between different dimensions. For instance, when generating house energy consumption TS data, the value distributions of the outside temperature and the kitchen temperature should be preserved, as well as the distribution of CC between them. Preserving such TS population-level properties is critical in maintaining the statistical insights of the datasets, mitigating model bias, and augmenting downstream tasks like TS prediction. Yet, it is often overlooked by existing models. Hence, data generated by existing models often bear distribution shifts from the original data. We propose Population-aware Diffusion for Time Series (PaD-TS), a new TS generation model that better preserves the population-level properties. The key novelties of PaD-TS include 1) a new training method explicitly incorporating TS population-level property preservation, and 2) a new dual-channel encoder model architecture that better captures the TS data structure. Empirical results in major benchmark datasets show that PaD-TS can improve the average CC distribution shift score between real and synthetic data by 5.9x while maintaining a performance comparable to state-of-the-art models on individual-level authenticity.

Mental-health stigma remains a pervasive social problem that hampers treatment-seeking and recovery. Existing resources for training neural models to finely classify such stigma are limited, relying primarily on social-media or synthetic data without theoretical underpinnings. To remedy this gap, we present an expert-annotated, theory-informed corpus of human-chatbot interviews, comprising 4,141 snippets from 684 participants with documented socio-cultural backgrounds. Our experiments benchmark state-of-the-art neural models and empirically unpack the challenges of stigma detection. This dataset can facilitate research on computationally detecting, neutralizing, and counteracting mental-health stigma. Our corpus is openly available at https://github.com/HanMeng2004/Mental-Health-Stigma-Interview-Corpus.

Large pre-trained models have been extensively explored for numerous biomedical tasks. However, the diversity and complexity of biological systems often make zero-shot learning in a new context challenging. In many instances, the budget allows for the acquisition of a small number of labeled data through experiments for few-shot learning. Yet, the methodology for selecting the optimal set of samples for these experiments remains underexplored. In this work, we present an application focused on drug-induced gene expression prediction to demonstrate a data-driven approach for facilitating sample selection. We developed a system named AnchorDrug, which predicts drug-induced gene expression changes in new cell lines after fine-tuning with experimental data from a limited number of drugs. Initially, we built a pre-trained model with a large dataset of drug-induced gene expressions. We then adopted active learning to identify an optimal set of drugs (i.e. anchor drugs) for experiments, aiming to ensure that the experimental data used for subsequent fine-tuning would maximize model performance. Several acquisition functions are customized and incorporated into our pipeline. Compared with knowledge-based drug selection, our customized active learning methods proved more effective in selecting anchor drugs. A model trained using data from anchor drugs can even perform better than that trained using all available data in certain scenarios. We further provided insights into the reasons behind its superior performance. Our system is designed to mimic real-world scenarios, enabling its easy application to real biomedical research projects.

Local Interpretability Model-agnostic Explanations (LIME) is a well-known post-hoc technique for explaining black-box models. While very useful, recent research highlights challenges around the explanations generated. In particular, there is a potential lack of stability, where the explanations provided vary over repeated runs of the algorithm, casting doubt on their reliability. This paper investigates the stability of LIME when applied to multivariate time series classification. We demonstrate that the traditional methods for generating neighbours used in LIME carry a high risk of creating 'fake' neighbours, which are out-of-distribution in respect to the trained model and far away from the input to be explained. This risk is particularly pronounced for time series data because of their substantial temporal dependencies. We discuss how these out-of-distribution neighbours contribute to unstable explanations. Furthermore, LIME weights neighbours based on user-defined hyperparameters which are problem-dependent and hard to tune. We show how unsuitable hyperparameters can impact the stability of explanations. We propose a two-fold approach to address these issues. First, a generative model is employed to approximate the distribution of the training data set, from which within-distribution samples and thus meaningful neighbours can be created for LIME. Second, an adaptive weighting method is designed in which the hyperparameters are easier to tune than those of the traditional method. Experiments on real-world data sets demonstrate the effectiveness of the proposed method in providing more stable explanations using the LIME framework. In addition, in-depth discussions are provided on the reasons behind these results.

Machine learning approaches have enabled increasingly powerful time series classifiers. While performance has improved drastically, the resulting classifiers generally suffer from poor explainability, limiting their applicability in critical areas. Saliency-based methods designed to highlight the critical features are one of the most promising approaches to improving this explainability. Here, current techniques commonly rely on artificially perturbing the features, using, for example, random noise or ‘zeroing’ these features. We first demonstrate that an important drawback of these methods is that the perturbations used can result in unrealistic assessments of the classifier, since the perturbations force the data outside their original distribution. We articulate how this can result in poor identification of critical features, and hence misleading explanations. In order to address this issue and identify the most important features for the output of a black-box model, we propose a dual approach through meaningful perturbation and optimisation. First, leveraging a mechanism originally proposed in image analysis, a generative model is trained to create within-distribution perturbations of the input. These are then used to reliably evaluate whether a set of features is critical. Second, a greedy-based segmentation and identification strategy is proposed to search for the smallest set of critical features. Experiments show that the proposed approach addresses the out-of-distribution problem and identifies fewer critical features than existing methods. In combination, both aspects of the proposed approach offer a qualitative advance towards generating meaningful and robust explanations in the context of time series classification.

With the rapid development of big data analytics and the Internet of Things (IoT) technology, online learning becomes more and more prevalent. To improve the quality of online learning and help teachers better understand the students’ learning state, this paper proposes a Facial Emotion Recognition (FER) Enabled Education Aids IoT System, named FEAIS. By deploying a well-trained FER model on the remote server, FEAIS can provide instant feedback on students’ learning state by collecting their facial pictures through IoT devices like cameras or vision sensors and making recognization on their emotion. Utilizing this information, teachers can adjust their teaching progress to ensure that everyone could catch up. The experimental results demonstrate that the FEAIS can accurately recognize students’ emotion and give teachers precise feedback automatically, even if the images captured by IoT devices are occluded or deformed.

Leveraging computational methods to generate small molecules with desired properties has been an active research area in the drug discovery field. Towards real-world applications, however, efficient generation of molecules that satisfy multiple property requirements simultaneously remains a key challenge. In this paper, we tackle this challenge using a search-based approach and propose a simple yet effective framework called MolSearch for multi-objective molecular generation (optimization).We show that given proper design and sufficient domain information, search-based methods can achieve performance comparable or even better than deep learning methods while being computationally efficient. Such efficiency enables massive exploration of chemical space given constrained computational resources. In particular, MolSearch starts with existing molecules and uses a two-stage search strategy to gradually modify them into new ones, based on transformation rules derived systematically and exhaustively from large compound libraries. We evaluate MolSearch in multiple benchmark generation settings and demonstrate its effectiveness and efficiency.

Short-term load forecasting (STLF) is an important foundation for the electrical network automation and intelligentization. Classic time series methods such as autoregressive integrated moving average (ARIMA)-based methods, and fashionable data mining technologies such as deep learning are not suitable for STLF on the small datasets with a number of attributes. Therefore, in order to tackle this problem, a novel ensemble formulation is proposed to model the load series in this paper. Specifically, the ensemble formulation decomposes the load consumption into two parts, i.e., the extrinsic-variational component (E-com) described by the external factors, and the intrinsic-stationary component (I-com) representing the internal structure of the series itself. Thereafter, a boosting regression learner consists of several simple learners, is further developed to model the E-com and I-com. By localizing the global attributes, the extraction of E-com is achieved by mining the relationships between load patterns and the corresponding local attributes, while I-com is naturally learned by the classic time series methods. Experimental results show that the proposed method improves the accuracy of STLF on small datasets, as compared to the existing approaches in the literature.

Time series classification is a challenging research area where machine learning techniques such as deep learning perform well, yet lack interpretability. Identifying the most important features for such classifiers provides a pathway to improving their interpretability. Several Feature Importance (FI) identification methods remove the contributions of features, i.e. observations at certain time steps of, from the input and evaluate the change in the classification result to measure the importance of features. As time series features cannot simply be deleted, current techniques generally rely on replacing features with constant or random values. While effective, this approach risks unexpected results in the classification and thus feature importance estimation-as the replacements used may be different to what the classifier encountered in the training phase. This is referred to as the Out-Of-Distribution problem. The OOD problem has been recognised in image and language models but have not received much attention in the context of time series classification. This work addresses the OOD problem in FI identification for time series classifiers. Specifically, we propose a method based on Conditional Variational Autoencoder to generate possible sets of within-distribution inputs, which are used to evaluate feature importance through marginalisation. Experiments on publicly accessible datasets are carried out showing that the method identifies the most important features with higher accuracy than existing methods, providing the basis for improved explainability of time series classifiers.