Junbin Gao

SEP 9, 2026 · PREPRINT

Characterizing Language Generation in the Limit: Finite Witnesses and a Separation-Width Hierarch

arXiv

This is a theoretical computer science preprint presenting mathematical characterizations of language generation in the limit, with formal proofs but no empirical validation, clinical data, or application to medical or clinical practice.

SEP 8, 2026 · PREPRINT

Nearly Tight Rademacher Bounds for Sparsely Activated Neural Networks

arXiv

This is a theoretical analysis of statistical complexity bounds for sparse neural networks with no empirical validation, experimental data, or clinical application.

AUG 11, 2026 · PREPRINT

Optimistic Rates for Multiclass PAC Learning

arXiv

This is a theoretical computer science paper establishing mathematical bounds for multiclass learning; it advances understanding of learning theory but does not report empirical validation, clinical outcomes, or results applicable to clinical practice.