🛰️ Daily AI Frontier
‹ back to 2026-08-29

Packora: Systematic Design for Generative Molecular Crystal Structure Prediction

Research Bioinformatics AI

Ranking

Overall 84
Content 100
Popularity 45

Observed public metrics from 1 member.

Representative image for Packora: Systematic Design for Generative Molecular Crystal Structure Prediction

Merged summary

TL;DR - Packora is a flow-based generative model that jointly predicts molecular crystal atomic coordinates and lattices from molecular graphs. It improves structure generation and ranking across six benchmarks while supporting complex crystal types and flexible conditioning.

  • Handles multi-component and organometallic crystals within one model.
  • Conditions on any subset of conformers, stereochemical labels, and space-group information.
  • Separately evaluates generation quality and end-to-end ranking under a shared relaxation and ranking pipeline.
  • Achieves the best matched-budget coverage across all six generation benchmarks, with better experimental-form recovery, lower ranks, and faster ranking convergence than baselines.

Sources (1)

Packora: Systematic Design for Generative Molecular Crystal Structure Prediction

arXiv cs.LG Nayoung Kim, Kiyoung Seong, Sungsoo Ahn 2026-08-27 arXiv:2608.26962
Public signals Hugging Face upvotes 0
Providers: Hugging Face · Upvotes 0 OpenAlex · N/A Publisher · N/A Semantic Scholar · N/A X · N/A Fetched 2026-09-25 14:26:26.769538 UTC

TL;DR - Packora is a flow-based generative model that jointly predicts molecular crystal atomic coordinates and lattices from molecular graphs. It improves structure generation and ranking across six benchmarks while supporting complex crystal types and flexible conditioning.

  • Handles multi-component and organometallic crystals within one model.
  • Conditions on any subset of conformers, stereochemical labels, and space-group information.
  • Separately evaluates generation quality and end-to-end ranking under a shared relaxation and ranking pipeline.
  • Achieves the best matched-budget coverage across all six generation benchmarks, with better experimental-form recovery, lower ranks, and faster ranking convergence than baselines.
item →