Published June 29, 2026 | Version v3
Dataset Open

Reproducibility package for "Attention Is Not All You Need for Diffraction"

Authors/Creators

  • 1. NIST Center for Neutron Research, National Institute of Standards and Technology, Gaithersburg, MD 20899, USA; Department of Physics, Boston College, Chestnut Hill, MA 02467, USA
  • 2. NIST Center for Neutron Research, National Institute of Standards and Technology, Gaithersburg, MD 20899, USA; School of Computer Science, Carnegie Mellon University, Pittsburgh, PA 15213, USA
  • 3. School of Information, University of California, Berkeley, CA 94720, USA
  • 4. NIST Center for Neutron Research, National Institute of Standards and Technology, Gaithersburg, MD 20899, USA
  • 5. NIST Center for Neutron Research, National Institute of Standards and Technology, Gaithersburg, MD 20899, USA; Department of Materials Science and Engineering, University of Maryland, College Park, MD 20742, USA

Description

Paper reproducibility package for the revised manuscript "Attention Is Not All You Need for Diffraction". The package contains the current flat-37 Vision Transformer checkpoint set, the matched regular-transformer architecture-control checkpoint, generated run configs and metadata, training/evaluation wrappers, frozen extinction-group-to-rule mapping assets, compact calibration and information-theory result summaries, compact SG-to-EG and XRDBench comparison outputs, and RRUFF-325/RRUFF-473 benchmark case-ID CSV files. The archive is intended as the paper artifact layer; generated training corpora are not included.

Notes

The archive contains MANIFEST.csv and CHECKSUMS.sha256. SG/EG categorical-control training checkpoints and posterior NPZ audit bundles are outside the main package scope; compact JSON/CSV outputs are included.

Files

Files (65.7 MB)

Name Size Download all
md5:f62a305aa02b0c77bcd64d750c1e55ff
65.7 MB Download

Additional details

Related works

Is supplement to
Software: https://github.com/scattering/paper-ai-diffraction (URL)