The rise of Large Visual-Language Models (LVLMs) has unlocked new possibilities for seamlessly integrating visual and textual information. However, their ability to interpret cartographic maps remains largely unexplored. In this paper, we introduce CartoMapQA, a benchmark specifically designed to evaluate LVLMs' understanding of cartographic maps through question-answering tasks. The dataset includes over 2000 samples, each composed of a cartographic map, a question (with open-ended or multiple-choice answers), and a ground-truth answer. These tasks span key low-, mid- and high-level map interpretation skills, including symbol recognition, embedded information extraction, scale interpretation, and route-based reasoning. Our evaluation of both open-source and proprietary LVLMs reveals persistent challenges: models frequently struggle with map-specific semantics, exhibit limited geospatial reasoning, and are prone to Optical Character Recognition (OCR)-related errors. By isolating these weaknesses, CartoMapQA offers a valuable tool for guiding future improvements in LVLM architectures. Ultimately, it supports the development of models better equipped for real-world applications that depend on robust and reliable map understanding, such as navigation, geographic search, and urban planning. Our source code and data are openly available to the research community at: https://github.com/ungquanghuy-kddi/CartoMapQA.git

Multivariate long-term time series forecasting (MLTSF), applicable across various domains, has gained increasing research attention. Channel-independent (CI) models, including Linear and Transformer-based architectures, have recently achieved state-of-the-art (SOTA) performance for MLTSF. Notably, Linear models can deliver satisfactory forecasting performance even with just a single linear projection layer. However, we identify a limitation in this architecture: a single linear projection struggles to adequately capture the inter- and intra-variate heterogeneity in temporal patterns. Similarly, any complex models like Transformer-based models that use a single projection layer to generate final predictions, may face capacity bottlenecks. To overcome this, we propose the Mixture of Projection Experts (MoPE), which replaces the single linear projection with multiple projection branches, and employs a gate network to dynamically assign weights to each branch based on the input data. We applied MoPE to multiple SOTA models and evaluated it on nine real-world datasets. Results show that MoPE boosts forecasting accuracy by an average of 9.59%, demonstrating its effectiveness in mitigating the limitations of a single projection layer. Additionally, our experiments demonstrate that integrating our proposal into CI models enhances their generalization to unseen variates. Interpretability analysis also reveals MoPE's ability to disentangle different temporal patterns. Overall, our paper establishes MoPE as an effective solution for MLTSF tasks.

Traffic predictions play a crucial role in intelligent transportation systems. The rapid development of IoT devices allows us to collect different kinds of data with high correlations to traffic predictions, fostering the development of efficient multi-modal traffic prediction models. Until now, there are few studies focusing on utilizing advantages of multi-modal data for traffic predictions. In this paper, we introduce a novel temporal attentive cross-modality transformer model for long-term traffic predictions, namely xMTrans, with capability of exploring the temporal correlations between the data of two modalities: one target modality (for prediction, e.g., traffic congestion) and one support modality (e.g., people flow). We conducted extensive experiments to evaluate our proposed model on traffic congestion and taxi demand predictions using real-world datasets. The results showed the superiority of xMTrans against recent state-of-the-art methods on long-term traffic predictions. In addition, we also conducted a comprehensive ablation study to further analyze the effectiveness of each module in xMTrans.

The paper aims to present a collaboration between the industry and government sectors, focusing on creating a benchmarking platform for predicting urban risk transportation through the utilization of multimodal data. In this collaboration, the industry partner contributes datasets and customer preference surveys obtained from its business operations. On the other hand, government partners curate open datasets sourced from non-profit organizations in both private and public domains. Furthermore, the government provides an accessible platform that allows individuals to conveniently access and leverage resources for the purpose of advancing application development and engaging in research endeavors. Throughout the collaborative effort, a variety of techniques have been under development for forecasting urban risk transportation through the analysis of weather patterns, congestion levels, and people flow data. The core objective of this partnership is to formulate two foundational prediction methods. These methods are intended to serve as benchmarks, offering future users a dependable means to assess the performance of their own approaches in terms of both time-series and datapoints analytics methodologies.

Human activity recognition using ambient sensors has become particularly important due to social demands of applications in smart homes. To address the problem of labeling sensing data for every individual house, cross-house human activity recognition is proposed to use available labeled houses (source domains) to train recognition models for applying to unlabeled houses (target domains). In this paper, we propose a method of source domain selection for cross-house human activity recognition. We first improve the method for representing semantic relationships of sensors. To select the best similar source houses for a target house, we then propose a method for calculating similarity score between two houses. Using 19 houses of the CASAS dataset, we evaluate the recognition performance in target houses using models trained by several similar source houses, randomly selected houses, dissimilar source houses, and all source houses without selection. Experimental results illustrate that the average accuracy of models trained from the small number of the best similar houses achieve the best performance, and thus they confirm the effectiveness of our proposed method.