Intern-Atlas builds a methodological evolution graph from over one million AI papers to model how methods emerge and adapt, enabling automated idea evaluation and generation.
A unified framework combines unsupervised pretraining and supervised neural knowledge graph learning, with a nonasymptotic risk bound showing unlabeled data reduces downstream prediction error.
Neural Structural Reasoner is a brain-inspired network preserving relational structure in neuronal dynamics to achieve interpretable, efficient structural reasoning over knowledge graphs.
NGDB-Zoo improves neural graph database training via operator-level scheduling and semantic augmentation, achieving 1.8, 6.8x throughput without I/O stalls.