Evidence / current public state

Materials Project Elastic Audit

Test whether a conservative descriptor family behaves coherently across shear and bulk targets.

Current stateBOUNDED CASE STUDY
01

Challenge

Test whether a conservative descriptor family behaves coherently across shear and bulk targets.

02

Intervention

Leakage-controlled modelling, null controls, perturbations and locked local repeats on one frozen 2,000-row cache.

03

Evidence

Both target pilots reproduced under locked configuration; bulk showed higher bounded predictive coherence than shear.

04

Limitation

Same-source and same-cache evidence is not independent external validation.

Supports
  • Both target pilots reproduced under locked configuration; bulk showed higher bounded predictive coherence than shear.
Does not prove
  • Same-source and same-cache evidence is not independent external validation.