We propose a hierarchical attention mechanism that improves cross-lingual inference by aligning contextual word representations across languages. The model outperforms traditional translation-based systems, offering enhanced semantic understanding with minimal labeled samples and better generalization in multilingual NLP benchmarks.

Michael Thomas Greene, Olivia Katherine Brooks, Emma Rose Collins, James Alexander Ford, Laura Michelle Hamilton

We develop a graph neural model with temporal attention for forecasting irregular environmental sensor data. The architecture dynamically weights node importance across time, improving multivariate prediction accuracy and resilience to missing data in spatially distributed observation systems.

Nathaniel George Harris, Emily Frances Peterson, Robert Jonathan Scott, Sarah Isabella Young, Rachel Olivia Bennett

This paper presents a transformer-based video understanding model that fuses multi-resolution language and action embeddings. It enhances temporal alignment and contextual reasoning across modalities, outperforming baselines in cross-modal retrieval and action captioning tasks on benchmark video datasets.

Chen Long Wei, Liu Zhi Fang, Xu Hao Ning, Zhang Wen Jun, Gao Ping Zhen

This research introduces a federated meta-learning framework for low-resource image classification. It integrates task-aware initialization and adaptive local updates to achieve high performance with minimal data, promoting efficient training across distributed environments with diverse client characteristics.

Aditya Kiran Deshmukh, Elena Sofia Martinez, Ravi Sandeep Iyer, Jean Pierre Moreau, Maria Clara Navarro

We propose a graph-based relational model for explainable multi-hop question answering. The system constructs dynamic subgraphs from structured knowledge bases and conversational context, supporting accurate, interpretable reasoning in AI agents operating in real-time dialogue environments.

Thomas Edward Harris, Laura Ann Jennings, Michael Brian Walker, Sarah Elizabeth Foster, Olivia Claire Newton

This paper introduces a deep generative model with latent class embeddings to enhance rare event detection. It captures complex temporal dependencies and improves separation between minority and majority sequences, enabling early anomaly identification with high confidence in imbalanced sequential datasets.

Liu Cheng Hao, Zhang Rui Wen, Xu Bo Liang, Huang Ming Fei, Chen Xiao Yu, Gao Xiu Zhen

This work introduces a quantum machine learning model designed for real-time event prediction in high-frequency sensor systems. It utilizes quantum-enhanced embeddings and dynamic updating to improve detection speed, noise resistance, and predictive reliability in time-sensitive monitoring applications.

Liu Fang Rui, Zhang Xiao Cheng, Gao Li Zhen, Xu Ming Long, Huang Yi Tao