Research map / Probabilistic and scientific modeling / Molecules and proteins
VecMol: Vector-Field Representations for 3D Molecule Generation
Yuchen Hua, Xingang Peng, Jianzhu Ma, Muhan Zhang
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 |
|---|---|---|---|
| increase autoencoder MLP hidden width from 384 to 1024 (ablation) | VecMol · GEOM-drugs | validity %: 82.5 -> 89.2 | Improved |
| increase autoencoder MLP hidden width from 384 to 1024 (ablation) | VecMol · GEOM-drugs | bond angle W1: 8.98 -> 4.81 | Improved |
| use VecMol instead of EDM/GeoLDM | VecMol · GEOM-drugs | stable molecules %: 80.2 vs 57.9 (GeoLDM), 40.3 (EDM) | Improved |
| use VecMol instead of EDM/GeoLDM | VecMol · QM9 | stable molecules %: 97.5 vs 97.9 (EDM), 97.5 (GeoLDM) | No clear effect |
| use VecMol instead of VoxMol | VecMol · QM9 | stable molecules %: 89.3 -> 97.5 | Improved |
| use VecMol instead of FuncMol | VecMol · CREMP | backbone KL divergence: 0.1997 -> 0.1516 (phi) | Improved |
| use VecMol instead of VoxMol/EDM for sampling speed | VecMol · GEOM-drugs | sampling time (s): 84.32 (VoxMol) -> 31.62 | Improved |
| use VecMol instead of VoxMol for sampling speed | VecMol · QM9 | sampling time (s): 14.18 (VoxMol) -> 14.94 | Hurt |
| use single-step denoised latent codes (VecMol-Diff) instead of encoded codes (VecMol-NF) (ablation) | VecMol · QM9 | stable molecules %: 97.6 -> 97.4 | No clear effect |
| increase latent noise level up to 0.5 (ablation) | VecMol neural field · QM9 and GEOM-drugs | validity, bond length W1, bond angle W1 | Mixed |
Closest papers by meaning
- Score-based 3D molecule generation with neural fields (NeurIPS 2024 · similarity 0.92)
- Learning Disentangled Equivariant Representation for Explicitly Controllable 3D Molecule Generation (AAAI 2025 · similarity 0.89)
- Navigating the Design Space of Equivariant Diffusion-Based Generative Models for De Novo 3D Molecule Generation (ICLR 2024 · similarity 0.89)
- Lift Your Molecules: Molecular Graph Generation in Latent Euclidean Space (ICLR 2025 · similarity 0.88)
- Towards Unified and Lossless Latent Space for 3D Molecular Latent Diffusion Modeling (NeurIPS 2025 · similarity 0.88)
- 3D Molecule Generation from Rigid Motifs via $\mathrm{SE}(3)$ Flows (NeurIPS 2026 · similarity 0.88)
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.
