Research map / Image generation / Autoregressive and latent generation
Distributional Priors Guided Diffusion for Generating 3D Molecules in Low Data Regimes
Haokai Hong, Wanyu Lin, Ming Yang, Kay Chen Tan
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 |
|---|---|---|---|
| use asymmetric autoencoder (reconstruct full molecule from substructure) instead of symmetric (ablation) | GODD · QM9 OOD-II scaffolds | validity (%): 35.0 -> 83.3 | Improved |
| use asymmetric autoencoder (ablation) | GODD · QM9 OOD-I scaffolds | validity (%): 58.3 -> 80.0 | Improved |
| use asymmetric autoencoder (ablation) | GODD · QM9 OOD-I scaffolds | molecule stability (%): 12.8 -> 45.6 | Improved |
| use asymmetric autoencoder (ablation) | GODD · QM9 OOD-II scaffolds | success rate (%): 31.0 -> 65.8 | Improved |
| use asymmetric autoencoder (ablation) | GODD · QM9 in-distribution scaffolds | scaffold coverage (%): 98.5 -> 92.5 | Hurt |
| use asymmetric autoencoder (ablation) | GODD · QM9 in-distribution scaffolds | success rate (%): 52.1 -> 52.4 | No clear effect |
| add structural priors vs scalar ring-count conditioning (C-EDM etc.) | GODD · QM9 OOD ring-structure | success rate (%): 25.0 -> 40.5 | Improved |
| use GODD instead of fragment-based LinkerNet | GODD · QM9 OOD-II scaffolds | success rate (%): 53.2 -> 65.8 | Improved |
| use GODD instead of LinkerNet | GODD · GEOM-LINKER OOD linker design | validity (%): 48.5 -> 65.2 | Improved |
| use unconditional diffusion models (EDM, GeoLDM) for rare scaffolds | EDM/GeoLDM · QM9 OOD-II scaffolds | success rate (%): 1.5 vs 65.8 (GODD) | Hurt |
Closest papers by meaning
- Navigating the Design Space of Equivariant Diffusion-Based Generative Models for De Novo 3D Molecule Generation (ICLR 2024 · similarity 0.89)
- Equivariant Blurring Diffusion for Hierarchical Molecular Conformer Generation (NeurIPS 2024 · similarity 0.88)
- Towards Unified and Lossless Latent Space for 3D Molecular Latent Diffusion Modeling (NeurIPS 2025 · similarity 0.88)
- NEAT: Neighborhood-Guided, Efficient, Autoregressive Set Transformer for 3D Molecular Generation (NeurIPS 2026 · similarity 0.88)
- Active Flow Expansion for Out-of-Distribution Discovery: from Theory to Molecules (NeurIPS 2026 · similarity 0.87)
- Enhancing Generalizability in Molecular Conformation Generation with METRIZATION-Informed Geometric Diffusion Pretraining (AAAI 2025 · 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.
