Research map / Graph learning / Graph self-supervised learning
Modular Flows: Differential Molecular Generation
Yogesh Verma, Samuel Kaski, Markus Heinonen, Vikas K Garg
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace E(3)-equivariant EGNN with non-equivariant 3-layer GCN (ablation) | ModFlow (3D) · ZINC250K | validity %: 95.4 -> 90.3 | Hurt |
| replace E(3)-equivariant EGNN with non-equivariant 3-layer GCN (ablation) | ModFlow (3D) · QM9 | validity %: 98.3 -> 93.3 | Hurt |
| use 3D instead of 2D coordinates (ablation) | ModFlow EGNN · QM9 | validity %: 96.2 -> 98.3 | Improved |
| use 3D instead of 2D coordinates (ablation) | ModFlow EGNN · ZINC250K | validity %: 94.8 -> 95.4 | Improved |
| add junction-tree representation to 3D coordinates | ModFlow (3D-EGNN) · QM9 | validity %: 98.3 -> 99.1 | Improved |
| add junction-tree representation to 3D coordinates | ModFlow (3D-EGNN) · ZINC250K | validity %: 95.4 -> 98.1 | Improved |
| add junction-tree representation to 2D coordinates | ModFlow (2D-EGNN) · ZINC250K | validity %: 94.8 -> 97.4 | Improved |
| add junction-tree representation to 2D coordinates | ModFlow (2D-EGNN) · QM9 | validity %: 96.2 -> 97.9 | Improved |
| use 3D instead of 2D coordinates (ablation) | ModFlow EGNN · QM9 | uniqueness %: 99.5 -> 99.1 | No clear effect |
| use ModFlow vs MoFlow baseline (one-shot, no validity correction) | ModFlow (JT-3D-EGNN) · ZINC250K | validity %: 50.3 -> 98.1 | Improved |
| use ModFlow vs GraphDF baseline on generation speed | ModFlow (2D-EGNN) · ZINC250K | seconds/molecule: 3.12 -> 0.46 | Improved |
| add junction-tree representation (slower decoding) | ModFlow (2D-EGNN) · QM9 | seconds/molecule: 0.16 -> 0.21 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- GraphDF — GraphDF: A Discrete Flow Model for Molecular Graph Generation (ICML 2021) · better on 5 of 6 shared results
- GraphAF — GraphAF: a Flow-based Autoregressive Model for Molecular Graph Generation (ICLR 2020) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Controlled Generation with Equivariant Variational Flow Matching (ICML 2025 · similarity 0.88)
- ET-Flow: Equivariant Flow-Matching for Molecular Conformer Generation (NeurIPS 2024 · similarity 0.88)
- Navigating Chemical Space with Latent Flows (NeurIPS 2024 · similarity 0.88)
- Efficient Molecular Conformer Generation with SO(3)-Averaged Flow Matching and Reflow (ICML 2025 · similarity 0.87)
- Equivariant Blurring Diffusion for Hierarchical Molecular Conformer Generation (NeurIPS 2024 · similarity 0.87)
- NEAT: Neighborhood-Guided, Efficient, Autoregressive Set Transformer for 3D Molecular Generation (NeurIPS 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
