Brief

Researchers have unified two families of training-free guided generation techniques for flow and diffusion models. They demonstrate that posterior guidance can be viewed as a greedy approach to end-to-end guidance. This theoretical unification allows for an interpolation between the two methods, offering a trade-off between computational cost and accuracy in gradient calculations. The findings were validated on inverse image problems and property-guided molecular generation. AI

Researchers have introduced Rex, a novel family of reversible solvers designed for neural differential equations. These solvers address the limitations of existing methods by enabling near-machine-precision inversion, which is crucial for applications requiring exact reconstruction. Rex achieves this by applying Lawson methods to convert standard explicit Runge-Kutta schemes into algebraically reversible ones, demonstrating improved performance in tasks like Boltzmann sampling and image generation. AI