Research map / Probabilistic and scientific modeling / Molecules and proteins
CatFlow: Co-generation of Slab-Adsorbate Systems via Flow Matching
Minkyu Kim, Nayoung Kim, Honghui Kim, Sungsoo Ahn
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 |
|---|---|---|---|
| Co-generate slab and adsorbate with flow matching instead of autoregressive CatGPT | CatFlow · OC20 de novo generation | uniqueness (%): 79.91 -> 94.69 | Improved |
| Co-generate slab and adsorbate with flow matching instead of CatGPT | CatFlow · OC20 de novo generation | validity (%): 92.67 -> 97.33 | Improved |
| Co-generate with flow matching instead of CatGPT | CatFlow · OC20 de novo generation | compositional diversity: 15.0655 -> 15.0724 | No clear effect |
| Co-generate with flow matching instead of CatGPT | CatFlow · OC20 de novo generation | system energy change on relaxation (eV): 33.1267 -> 28.0060 | Improved |
| End-to-end co-generation instead of two-step DiffCSP+AdsorbDiff | CatFlow · OC20 structure prediction | validity (%): 64.95 -> 98.16 | Improved |
| End-to-end co-generation instead of two-step DiffCSP+AdsorbDiff | CatFlow · OC20 structure prediction | match rate (%): 7.71 -> 11.09 | Improved |
| End-to-end co-generation instead of two-step DiffCSP+AdsorbDiff | CatFlow · OC20 structure prediction | adsorption energy success rate (%): 10.57 -> 16.46 | Improved |
| Use CatFlow instead of random plus heuristic adsorbate placement | CatFlow · OC20-Dense global minima | success rate (%): 10.5 -> 15.2 | Improved |
| Use CatFlow instead of random plus heuristic placement | CatFlow · OC20-Dense global minima | mean abs adsorption energy difference (eV): 1.9472 -> 0.3798 | Improved |
| Use primitive-cell factorized representation instead of full slab atoms (ablation) | CatFlow · OC20 slab atom count | learnable variables: reduced by an average of 9.2x | Improved |
Closest papers by meaning
- AdsorbDiff: Adsorbate Placement via Conditional Denoising Diffusion (ICML 2024 · similarity 0.88)
- EP-Flow: Disordered Crystal Structure Prediction without Site-level Annotation (NeurIPS 2026 · similarity 0.85)
- DR-Label: Label Deconstruction and Reconstruction of GNN Models for Catalysis Systems (AAAI 2024 · similarity 0.84)
- Open Materials Generation with Stochastic Interpolants (ICML 2025 · similarity 0.83)
- MOFFlow: Flow Matching for Structure Prediction of Metal-Organic Frameworks (ICLR 2025 · similarity 0.83)
- FlowLLM: Flow Matching for Material Generation with Large Language Models as Base Distributions (NeurIPS 2024 · similarity 0.83)
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.
