The field of cellular modeling and multi-omics analysis is moving towards a more integrated and hierarchical approach, recognizing the complex relationships between perturbations, transcriptional responses, and phenotypic changes. Recent developments prioritize the modeling of causal links from RNA to morphology, as well as the alignment of heterogeneous multi-omics data into a shared latent space. This shift enables the development of more robust and generalizable models, such as those employing trimodal cascade generative frameworks or batch-robust multi-omics representations. Noteworthy papers include TRIDENT, which achieves state-of-the-art performance in cellular morphology synthesis, and MoRE, which provides a practical step towards general-purpose omics foundation models. Additionally, CHMR offers a robust framework for learning cell-aware hierarchical multi-modal representations, while a decoupled alignment kernel approach shows promise for peptide membrane permeability predictions.