EF-TALFM framework in PCQM4Mv2 study generates variable-size 3D molecules; 89.4% unique and valid molecules
EF-TALFM framework in PCQM4Mv2 study generates variable-size 3D molecules; 89.4% unique and valid molecules Fixed-Dimensional Latent Flow for Generating Variable-Size 3D Molecules Fixed-Dimensional Latent Flow for Generating Variable-Size 3D Molecules In molecular discovery, molecule size is coupled to composition, structure, stability, binding affinity, and other target properties. Yet most 3D generators require molecule size to be specified before generation. Here, we introduce Equivariant-Free Transformer-Autoencoded Latent Flow Matching (EF-TALFM), a two-stage generative framework that relies entirely on a single fixed-dimensional molecule-level latent representation to generate variable-size molecules. The second-stage flow matching model samples this latent vector, and an autoregressive Transformer decoder then determines molecule size while generating atom types, coordinates, and chemically informative states. Canonical atom ordering and rigid-pose alignment enable standard Transformers without equivariant layers, while joint decoding of molecular geometry and an enriched chemical state enables reliable, deterministic, chemistry-guided graph recovery without requiring a learned dense pairwise bond decoder. The same fixed-dimensional latent supports unconditional and property-conditioned flow matching, while optional property supervision adds an internal ranking readout, with no separate predictor or reference calculations. On PCQM4Mv2, EF-TALFM achieves the highest fraction of molecules that are unique, absent from the training set, pass sanitization and PoseBusters sanity checks, 89.4%, compared with 75.6% for UAE-3D and 69.8% for FlowMol. Under the reported budgets on the same hardware, EF-TALFM also achieves higher measured computational throughput, with sampling-throughput speedups of 1.26 × \times and 3.62 × \times and approximate end-to-end training-throughput speedups of 3.20 × \times and 314 × \times over UAE-3D and FlowMol, respectively. Across ten target HOMO–LUMO gaps, internal ranking increases the density functional theory (DFT)-verified hit rate within 0.1 eV 0.1,\mathrm{eV} from 25.0% to 52.4%, while preserving 97% novelty among unique verified hits relative to their property-matched training subsets. These results demonstrate that fixed-dimensional molecule-level generation followed by symmetry-resolved autoregressive realization provides a practical architecture for open-ended and property-directed 3D molecular design. ...