Research map / Image generation / Autoregressive and latent generation
GAGA: Gaussianity-Aware Gaussian Approximation for Efficient 3D Molecular Generation
Jingxiang Qu, Wenhan Gao, Ruichen Xu, Yi Liu
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 |
|---|---|---|---|
| truncate trajectory with Gaussian approximation (GAGA) | EDM · QM9 | molecule stability / sampling time: 82.0 -> 85.6 mol stability; 1000 -> 550 steps; 0.65 -> 0.36 s | Improved |
| truncate trajectory with GAGA | GeoLDM · QM9 | molecule stability: 89.4 -> 92.3 | Improved |
| truncate trajectory with GAGA | EquiFM · QM9 | validity: 94.7 -> 96.2 | Improved |
| truncate trajectory with GAGA | EquiFM · QM9 | Valid*Uniq: 93.5 -> 93.7 | No clear effect |
| truncate trajectory with GAGA | GeoLDM · GEOM-Drugs | validity: 99.3 -> 99.3 | No clear effect |
| truncate trajectory with GAGA | EDM · GEOM-Drugs | atom stability: 81.3 -> 84.3 | Improved |
| apply GA too early (T*=450) (ablation) | EDM + GAGA · QM9 | Valid*Uniq: 92.0 -> 91.4 | Hurt |
| apply GA too late (T*=650) (ablation) | EDM + GAGA · QM9 | Valid*Uniq: 92.0 -> 91.6 | Hurt |
| apply GA too late (T*=750) (ablation) | GeoLDM + GAGA · QM9 | Valid*Uniq: 94.4 -> 93.5 | Hurt |
| add GAGA on top of DDIM | EDM + DDIM · QM9 (DDIM 2x) | Valid*Uniq: 83.7 -> 83.9 | No clear effect |
| add GAGA on top of DDIM | GeoLDM + DDIM · QM9 (DDIM 2x) | Valid*Uniq: 85.8 -> 87.5 | Improved |
| use GAGA instead of DDIM for acceleration | EDM · QM9 | Valid*Uniq: 83.7 (DDIM, 500 steps) -> 92.0 (GAGA, 550 steps) | Improved |
Closest papers by meaning
- Unified Generative Modeling of 3D Molecules with Bayesian Flow Networks (ICLR 2024 · similarity 0.88)
- Accelerating 3D Molecule Generation via Jointly Geometric Optimal Transport (ICLR 2025 · similarity 0.87)
- Generative Modeling of Molecular Dynamics Trajectories (NeurIPS 2024 · similarity 0.87)
- Geometric Representation Condition Improves Equivariant Molecule Generation (ICML 2025 · similarity 0.87)
- Score-based 3D molecule generation with neural fields (NeurIPS 2024 · similarity 0.87)
- Equivariant Neural Diffusion for Molecule Generation (NeurIPS 2024 · similarity 0.86)
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.
