Reproduce “Deep Scientific Reasoning under Physical Constraints: Structure-Aware Spectrum Prediction” 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

Yingheng Wang, Tao Yu, Shufeng Kong, Francesco Ricci, John M. Gregoire, Carla P. Gomes

DeepSciReasoner predicts electronic density-of-states spectra for crystals from structure. An edge-aware graph Transformer encodes atoms, and energy bins attend to atom representations to form an initial spectrum; a second stage transports normalized spectral mass under a conservation constraint. On CalTech eDOS, refinement lowers MSE from 20.990 ± 0.373 to 19.612 ± 0.225. Evaluation uses simulated, not experimental, spectra.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.