Search papers, labs, and topics across Lattice.
University of Science, VNU-HCM, Vietnam National University, Thong Nhat Hospital
2
0
5
Vision-language models can achieve over 96% accuracy in identifying dangerous actions by children, far surpassing traditional deep learning methods.
Naive nearest neighbor retrieval cripples tabular in-context learning for clinical risk prediction, but task-aligned retrieval unlocks significant gains, especially when data is messy.
