Research map / Point clouds / Point cloud classification
SE(3)-Transformers: 3D Roto-Translation Equivariant Attention Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 SE(3)-equivariant attention instead of Set Transformer attention | N-body prediction · N-body simulations | MSE x: 0.0139 -> 0.0076 | Helped |
| Use SE(3)-equivariant attention instead of Set Transformer attention | N-body prediction · N-body simulations | MSE v: 0.101 -> 0.075 | Helped |
| Add symmetry-breaking z-coordinate input (ablation) | SE(3)-Transformer · ScanObjectNN object only | accuracy: 72.8 % -> 85.0% | Helped |
| Use SE(3)-Transformer rather than the strongest listed baseline | QM9 · QM9 test | MAE α: .084 -> .142$\pm$.002 | Hurt |
| Use SE(3)-Transformer rather than the strongest listed baseline | QM9 · QM9 test | MAE Δε: 49 -> 53.0$\pm$0.3 | Hurt |
| Use SE(3)-Transformer rather than the strongest listed baseline | QM9 · QM9 test | MAE ε_HOMO: 30 -> 35.0$\pm$.9 | Hurt |
| Use SE(3)-Transformer rather than the strongest listed baseline | QM9 · QM9 test | MAE ε_LUMO: 25 -> 33.0$\pm$.7 | Hurt |
| Use SE(3)-Transformer rather than the strongest listed baseline | QM9 · QM9 test | MAE μ: .030 -> .051$\pm$.001 | Hurt |
| Use SE(3)-Transformer rather than the strongest listed baseline | QM9 · QM9 test | MAE Cν: .026 -> .054$\pm$.002 | Hurt |
Closest papers by meaning
- SE(3)-bi-equivariant Transformers for Point Cloud Assembly (NeurIPS 2024 · similarity 0.89)
- Approximately Piecewise E(3) Equivariant Point Networks (ICLR 2024 · similarity 0.87)
- Fast, Expressive $\mathrm{SE}(n)$ Equivariant Networks through Weight-Sharing in Position-Orientation Space (ICLR 2024 · similarity 0.86)
- Attention on the Sphere (NeurIPS 2025 · similarity 0.86)
- Pose-Transformed Equivariant Network for 3D Point Trajectory Prediction (CVPR 2024 · similarity 0.85)
- On the Fourier analysis in the SO(3) space : the EquiLoPO Network (ICLR 2025 · similarity 0.85)
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.