Reproduce “CrystalJev: calibrated foundation-model probabilities decide what to compute next in materials discovery” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Peng Kang, Yaxuan Sun, Zhen Li, Yu Liu, Lei Zheng, Huibin Xu
CrystalJev converts one-pass outputs from frozen atomistic models into calibrated probabilities that crystals meet a DFT hull stability criterion. It uses expected decision value to direct follow-up calculations toward candidates whose results could change an attempt-or-skip decision. On WBM test candidates, targeted orders retained 76%–78% of the gain from relaxing everyone, versus at most 4% for most-probable-first; new-pool transfer required recalibration.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.