Proposes a Bayesian Hypergraph Pathway Inference framework for multi-disease EHR modeling, with structured variational inference and a repulsion prior for interpretable, identifiable, and uncertainty-aware pathway discovery.
A large-scale open-source benchmark for topic classification in 200+ languages and dialects, designed to improve evaluation quality for multilingual NLU.
Introduces a human-annotated simplification dataset across English, Sinhala, Tamil, Thai, and Pashto, and benchmarks open-weight multilingual LLMs under zero-shot/few-shot settings to expose high-resource vs low-resource performance gaps.
A multi-way parallel corpus of math word problems in nine languages, including low-resource settings, with modeling strategies that integrate mathematical constraints into generation.
A refined IRT-style framework with enhanced discrimination modeling and group-wise diagnostics to better assess question quality and learner ability.
An educational Android app for water conservation and purification awareness, combining interactive quizzes and virtual experiments for community learning.
A stable hyperelastic simulation method using closed-form spectral analysis and Hessian projection, improving robustness under large deformation, inversion, and rotation.
Explores ridge, LASSO, and elastic-net methods under summary-data settings for neurobiological and behavioral analysis without relying on conventional cross-validation.