Research map / Probabilistic and scientific modeling / Molecules and proteins
Symmetric Space Learning for Combinatorial Generalization
Jaehyoung Jeong, Hee-Jun Jung, Kangil Kim
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 |
|---|---|---|---|
| add flow matching to Cartan-loss model (replace VAE) | Cartan + VAE · Sphere manifold reconstruction | Chamfer distance: 0.5763 -> 0.0068 | Improved |
| add GSC loss to Cartan + flow matching (ablation) | Cartan + FM · Sphere manifold reconstruction | Chamfer distance: 0.0068 -> 0.0061 | Improved |
| add Cartan loss to VAE (ablation) | Vanilla VAE · Sphere manifold reconstruction | Chamfer distance: 0.0601 -> 0.5763 | Hurt |
| use CartanFM instead of VAE | VAE · dSprites R2R Case1 | MSE: 28.81 -> 7.02 | Improved |
| use CartanFM instead of VAE | VAE · MPI3D R2R Case1 | MSE: 6.29 -> 0.51 | Improved |
| use CartanFM instead of VAE | VAE · 3D Shapes R2R Case2 | MSE: 37.65 -> 33.87 | Improved |
| add Cartan loss to flow-matching backbone (ablation) | FM backbone · dSprites R2R Case1 | MSE: 29.06 -> 57.38 | Hurt |
| add Cartan loss to flow-matching backbone (ablation) | FM backbone · 3D Shapes R2R Case1 | MSE: 20.12 -> 9.42 | Improved |
| add GSC loss to flow-matching backbone (ablation) | FM backbone · dSprites R2R Case1 | MSE: 29.06 -> 13.11 | Improved |
| add GSC loss to flow-matching backbone (ablation) | FM backbone · dSprites R2R Case2 | MSE: 23.70 -> 47.43 | Hurt |
| add GSC loss on top of Cartan loss (ablation) | FM + Cartan · dSprites R2R Case1 | MSE: 57.38 -> 7.02 | Improved |
| add both Cartan and GSC losses to backbone (ablation) | FM backbone · 3D Shapes R2R Case2 | MSE: 33.09 -> 33.87 | No clear effect |
Closest papers by meaning
- Learning Infinitesimal Generators of Continuous Symmetries from Data (NeurIPS 2024 · similarity 0.86)
- Permutation Equivariant Neural Networks for Symmetric Tensors (ICML 2025 · similarity 0.85)
- Symmetry Discovery Beyond Affine Transformations (NeurIPS 2024 · similarity 0.85)
- Transport of Algebraic Structure to Latent Embeddings (ICML 2024 · similarity 0.84)
- A Generative Model of Symmetry Transformations (NeurIPS 2024 · similarity 0.84)
- Improving Equivariant Networks with Probabilistic Symmetry Breaking (ICLR 2025 · similarity 0.84)
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.
