We found that likelihood computations in flow-based Boltzmann generators currently are a pressing problem hindering the approach form being scaled to large systems. Thus, our past efforts focused on overcoming this issues by developing a novel graph flow-based generative model (HollowFlow) that increased the sampling speed for these models by several orders of magnitude. More details about our prior success can be found in activity report of the preceding NAISS project (NAISS 2024/22-688). Most importantly, our new method and the corresponding experimental results required further experimental testing, see activity reports (NAISS 2024/22-688, NAISS 2025/22-841). Additionally, we moved on to learning transition densities of 1) Glassy dynamics. We obtained a proof of concept on small systems (10 particles) and started scaling to larger systems (256 particles). Ultimately, to be physically meaningful, we aim for even larger systems with as much as thousands of atoms. 2) Driven systems. We obtained a proof of concept on small test systems (one- and two-dimensional) and are currently scaling to larger systems.