Graph diffusion models, dominant in graph generative modeling, remain underexplored for graph-to-graph translation tasks like chemical reaction prediction. We demonstrate that standard permutation equivariant denoisers face fundamental limitations in these tasks due to their inability to break symmetries in noisy inputs. To address this, we propose aligning input and target graphs to break input symmetries while preserving permutation equivariance in non-matching graph portions. Using retrosynthesis (i.e., the task of predicting precursors for synthesis of a given target molecule) as our application domain, we show how alignment dramatically improves discrete diffusion model performance from 5% to a SOTA-matching 54.7% top-1 accuracy. Code is available at https://github.com/Aalto-QuML/DiffAlign.
@article{arxiv.2405.17656,
title = {Equivariant Denoisers Cannot Copy Graphs: Align Your Graph Diffusion Models},
author = {Najwa Laabid and Severi Rissanen and Markus Heinonen and Arno Solin and Vikas Garg},
journal= {arXiv preprint arXiv:2405.17656},
year = {2025}
}
Comments
Accepted at ICLR 2025. Code available at https://github.com/Aalto-QuML/DiffAlign