Search papers, labs, and topics across Lattice.
This paper introduces a canonicalization approach to diffusion models for generative tasks with group symmetries, mapping samples to orbit representatives before training a non-equivariant diffusion model. The authors theoretically prove the correctness, universality, and superior expressivity of canonical generative models, demonstrating that canonicalization accelerates training by reducing complexity and conditional variance. Empirically, the proposed framework, CanonDiffusion and CanonFlow, achieves state-of-the-art performance on molecular graph generation tasks, outperforming equivariant baselines with improved efficiency.
Ditch the equivariant constraints: canonicalization lets you train simpler, faster diffusion models that actually *outperform* equivariant architectures for symmetric generative tasks like 3D molecule design.
Many generative tasks in chemistry and science involve distributions invariant to group symmetries (e.g., permutation and rotation). A common strategy enforces invariance and equivariance through architectural constraints such as equivariant denoisers and invariant priors. In this paper, we challenge this tradition through the alternative canonicalization perspective: first map each sample to an orbit representative with a canonical pose or order, train an unconstrained (non-equivariant) diffusion or flow model on the canonical slice, and finally recover the invariant distribution by sampling a random symmetry transform at generation time. Building on a formal quotient-space perspective, our work provides a comprehensive theory of canonical diffusion by proving: (i) the correctness, universality and superior expressivity of canonical generative models over invariant targets; (ii) canonicalization accelerates training by removing diffusion score complexity induced by group mixtures and reducing conditional variance in flow matching. We then show that aligned priors and optimal transport act complementarily with canonicalization and further improves training efficiency. We instantiate the framework for molecular graph generation under $S_n \times SE(3)$ symmetries. By leveraging geometric spectra-based canonicalization and mild positional encodings, canonical diffusion significantly outperforms equivariant baselines in 3D molecule generation tasks, with similar or even less computation. Moreover, with a novel architecture Canon, CanonFlow achieves state-of-the-art performance on the challenging GEOM-DRUG dataset, and the advantage remains large in few-step generation.
MIT CSAIL