8 papers indexed
Lei Li, Zhenhui Pan, Yaqin Ren, Cheng Xiao
Abstract General teacher Digital Competence (DC) frameworks, which lack a subject-specific pedagogical perspective, are limited in their ability to effectively guide technology integration practices in physical education contexts. Using an exploratory sequential mixed-methods des…
Lichao Mou, Shilan Zhang, Chunlei Li, Bingcong Yan, Jingliang Hu, Yilei Shi, et al.
Large vision-language models (LVLMs) can be adapted to specialized medical imaging tasks via parameter-efficient fine-tuning approaches such as low-rank adaptation (LoRA), leading to a growing ecosystem of expert models tailored to specific imaging modalities and clinical scenari…
Guang Yang, Wentian Xu, Siyu Wang, Betty Raman, Lei Li, Vicente Grau
Myocardial infarction (MI) remains a leading cause of mortality worldwide. Echocardiography (Echo) is a widely available modality for MI assessment, where regional wall motion abnormality is a key indicator. Prior learning based methods for myocardial motion analysis often use ha…
Tianfang Zhang, Fengyi Wu, Lei Li, Chang Liu, Zhenming Peng, Huaping Zhang, et al.
Infrared small target detection (IRSTD) aims to identify long distance small targets from complex infrared backgrounds, and is a fundamental task in remote sensing. Deep learning methods have improved IRSTD by learning discriminative image-to-mask mappings, but such feed-forward…
Haochen Luo, Yi Huang, Sichun Luo, Fengyuan Liu, Lei Li, Zefa Hu, et al.
Self-evolving frameworks usually optimize task solutions while treating the surrounding harness as fixed. We introduce Harness-Aware Self-Evolving (HASE), an agentic reinforcement-learning framework in which a single model can generate task solutions or edit selected harness comp…
Chengyuan Yang, Yifan Wang, Chun Kwang Tan, Sherwin Stephen Chan, Youlong Wang, Xiaoyue Yan, et al.
The aging global population drives demand for assistive robots, yet the safety risks and costs of physical testing make Human-in-the-Loop (HITL) simulation an attractive alternative. Its fidelity for coupled systems, however, is limited by interaction models whose impedance param…
Xiaoyue Liu, Dongcheng Cang, Xiaohan Yuan, Mark YY Chan, Ching-Hui Sia, Lei Li
Accurate 4D whole-heart mesh reconstruction from sparse cine MRI is critical for creating cardiac digital twins, but remains challenging due to limited 2D slice coverage and the complex coupling between cardiac shape and motion. Existing methods often rely on intermediate contour…
Wenhan Ma, Jianyu Wei, Liang Zhao, Hailin Zhang, Bangjun Xiao, Lei Li, et al.
Modern large language models (LLMs) rely on reinforcement learning during post-training to push specific capabilities, yet integrating multiple capabilities into one model remains hard. Existing methods, such as Off-Policy Finetune and Mix-RL, are either inefficient or lose perfo…