Rex mathematical framework significantly improves fidelity, speed, accuracy of AI models applied to complex scientific simulations

https://techxplore.com/news/2026-06-ai-math-tool-sharpen-image.html

"generative AI relies on diffusion/ flow-matching synthesizing data by progressively transforming random noise into structured outputs, but reversing this process accumulates inaccuracies... overcome using Reversible Exponential Stochastic Runge-Kutta Solvers, aligning forward/ backward computational steps reducing inversion errors by orders magnitude... integrates into established architectures, more robust round-trip digital image editing while preserving fine details... molecular simulations"