Meng Xu

K-means++ systematically struggles in large-cluster regimes, but reformulating cardinality constraints into a continuous Riemannian difference-of-convex minimax game yields provably exact penalization and superior cluster assignments.

SNIPTEST reveals that targeted fuzzing of code slices can significantly enhance the validation of static analysis warnings, achieving over 54% success in identifying true vulnerabilities.

Decoupling generation from selection leads to substantial gains in factual accuracy and relevance in summarization, even under strict budget constraints.

Spotting AI-generated harms requires more than just image analysis: this system links images to their source even when paired with harmful text, enabling accountability.