Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces Latent Flow Matching (LFM) for molecular graph generation, which operates on a pretrained VAE's latent space rather than the discrete graph space. By learning continuous transformations in this latent space via flow matching and decoding only at the end, LFM achieves competitive validity and FCD scores on molecular benchmarks while offering improved efficiency. A key advantage is the reusability of the learned latent representation for diverse generative objectives, including property-guided and validity-aware generation.
Generating molecular graphs in a pretrained VAE's latent space via flow matching unlocks significant efficiency gains and enables property-guided generation without retraining.
Modern graph generative models typically operate directly in the discrete graph space, explicitly generating node and edge variables, which can become costly as graphs grow. In this paper, we perform generation explicitly on latent representations of entire graphs obtained from a pretrained Variational Autoencoder with high reconstruction fidelity. The generated representations, obtained through flow matching, are then decoded only at the final step. Across molecular benchmarks of increasing size, our approach achieves strong validity and FCD while offering a favorable quality-efficiency trade-off compared with state-of-the-art explicit graph generative models. One of the main advantages of this formulation is that the graph representation only needs to be learned once, after which the same one can be reused across multiple generative objectives without retraining. We demonstrate generation guided by molecular properties and further introduce validity-aware generation though a classifier learned directly in latent space. All code will be made available upon acceptance.