Bin Liu

Lévy noise can render the 3D torus energy-critical NLS globally well-posed, even for arbitrary initial data, by introducing a cubic lower bound on expected energy and excluding free scattering.

A unified framework for global well-posedness of the stochastic Landau--Lifshitz--Baryakhtar equation is established for initial data down to $\mathbb{L}^2$ in $\mathbb{R}^d$, $d \leq 3$.

Giving agent harness optimizers unconstrained edit scope reliably degrades performance; enforcing graduated, periodic permission scheduling ($1\to2\to3$) instead yields an immediate 20% jump in held-out task completion.