Search papers, labs, and topics across Lattice.
The paper introduces Equivariant MeanFlow (EQUIMF), a novel SE(3)-equivariant generative framework that jointly models discrete molecular graph structure and continuous geometry using synchronized MeanFlow dynamics. EQUIMF employs a unified time bridge and mutual conditioning between structure and geometry, enabling efficient few-step generation. Experiments show EQUIMF outperforms existing diffusion and flow-matching methods in generation quality, physical validity, and sampling efficiency for molecular graph generation.
Molecular graph generation gets a major speed boost and quality upgrade with Equivariant MeanFlow, which synchronizes discrete and continuous dynamics for physically realistic structures in fewer steps.
Graph-structured data jointly contain discrete topology and continuous geometry, which poses fundamental challenges for generative modeling due to heterogeneous distributions, incompatible noise dynamics, and the need for equivariant inductive biases. Existing flow-matching approaches for graph generation typically decouple structure from geometry, lack synchronized cross-domain dynamics, and rely on iterative sampling, often resulting in physically inconsistent molecular conformations and slow sampling. To address these limitations, we propose Equivariant MeanFlow (EQUIMF), a unified SE(3)-equivariant generative framework that jointly models discrete and continuous components through synchronized MeanFlow dynamics. EQUIMF introduces a unified time bridge and average-velocity updates with mutual conditioning between structure and geometry, enabling efficient few-step generation while preserving physical consistency. Moreover, we develop a novel discrete MeanFlow formulation with a simple yet effective parameterization to support efficient generation over discrete graph structures. Extensive experiments demonstrate that EQUIMF consistently outperforms prior diffusion and flow-matching methods in generation quality, physical validity, and sampling efficiency.
