Precipitation Nowcasting, which aims to predict precipitation within the next 0 to 6 hours, is critical for disaster mitigation and real-time response planning. However, most time series forecasting benchmarks in meteorology are evaluated on variables with strong periodicity, such as temperature and humidity, which fail to reflect model capabilities in more complex and practically meteorology scenarios like precipitation nowcasting. To address this gap, we propose RainfallBench, a benchmark designed for precipitation nowcasting, a highly challenging and practically relevant task characterized by zero inflation, temporal decay, and non-stationarity, focusing on predicting precipitation within the next 0 to 6 hours. The dataset is derived from five years of meteorological observations, recorded at hourly intervals across six essential variables, and collected from more than 140 Global Navigation Satellite System (GNSS) stations globally. In particular, it incorporates precipitable water vapor (PWV), a crucial indicator of precipitation that is absent in other datasets. We further design specialized evaluation protocols to assess model performance on key meteorological challenges, including multi-scale prediction, multi-resolution forecasting, and extreme precipitation events, benchmarking 17 state-of-the-art models across six major architectures on RainfallBench. Additionally, to address the zero-inflation and temporal decay issues overlooked by existing models, we introduce Bi-Focus Precipitation Forecaster (BFPF), a plug-and-play module that incorporates domain-specific priors to enhance precipitation nowcasting. Statistical analysis and ablation studies validate the comprehensiveness of our dataset as well as the effectiveness of our methodology. Code and datasets are available at https://github.com/yfangZhang/RainfallBench.
Currently, the main approach for Large Language Models (LLMs) to tackle the hallucination issue is incorporating Knowledge Graphs(KGs).However, LLMs typically treat KGs as plain text, extracting only semantic information and limiting their use of the crucial structural aspects of KGs. Another challenge is the gap between the embedding spaces of KGs encoders and LLMs text embeddings, which hinders the effective integration of structured knowledge. To overcome these obstacles, we put forward the SSKG-LLM, an innovative model architecture that is designed to efficiently integrate both the Structural and Semantic information of KGs into the reasoning processes of LLMs. SSKG-LLM incorporates the Knowledge Graph Retrieval (KGR) module and the Knowledge Graph Encoding (KGE) module to preserve semantics while utilizing structure. Then, the Knowledge Graph Adaptation (KGA) module is incorporated to enable LLMs to understand KGs embeddings. We conduct extensive experiments and provide a detailed analysis to explore how incorporating the structural information of KGs can enhance the factual reasoning abilities of LLMs. Our code are available at https://github.com/yfangZhang/SSKG-LLM.
Existing few-shot classification methods ignore the local spatial image features information, as image embeddings extracted globally may not optimally discriminative for the target task. In this paper, we argue that support embeddings supposed to represent image features adaptively for different tasks. Specifically, we extract local features at each spatial locations of images, and introduce an embeddings generator to produce task-specific embeddings of support samples adapt to the task at hand by reweighting local feature weights. The weights optimization is guided by capturing the correlations between intra and inter-class support samples. Then, query images are classified based on the feature maps reconstructed by support embeddings of different classes. We conduct extensive experiments to validate our method and the results indicate that our method improves the performance of few-shot classification on three benchmark datasets.
Traffic safety accidents and events are easily happened on campus. A multiply information fusion method for supporting emergency management on campus is proposed. This study analyzes the transportation characteristics on campus road, the difference between campus transportation and urban transportation, and the cause of traffic accident taking place frequently on campus road. Moreover, this paper proposes a relative strategy for preventing traffic accident on campus. The abnormal conflict situations are defined as the abnormal events, and these events could be detected and alarmed automatically by a real-time anomaly detection system. This information is further processed and stored in a big-data center, which could enable the campus administrators to take prompt and propitiate measures to deal with the situations. This project could improve the ability of the campus administrators to deal with the mega-events such as the sports meeting and the international conferences. Scientific measures and plans can be taken to ensure the campus safety in daily management, which could improve the efficiency of campus safety administration as well.