CatFlow is a flow-matching generative model that produces slab-adsorbate complexes with higher structural fidelity than autoregressive baselines on the Open Catalyst 2020 dataset.
This is achieved by filtering atoms based on their tags, where atoms with the tag 2 (representing the adsorbate) are excluded
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cond-mat.mtrl-sci 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
CatFlow: Co-generation of Slab-Adsorbate Systems via Flow Matching
CatFlow is a flow-matching generative model that produces slab-adsorbate complexes with higher structural fidelity than autoregressive baselines on the Open Catalyst 2020 dataset.