[WIP] Updated Matcalc Benchmark PBE with new references

[rul048]

Summary

In our previous study, we used the relaxed geometries and formation energies of randomly sampled WBM structures—computed with PBE PAW dataset version 52—as our equilibrium benchmarking data. Since the MatPES dataset employs the newer PBE PAW dataset version 64 and the perturbation logic of pymatgen recently changes , we have now re-relaxed all geometries and recomputed formation energies using the MatPESStaticSetting together with the MatPES-PBE elemental references. We then reran the equilibrium benchmarking and updated the resulting statistics to ensure full consistency across training datasets.

The basic idea here is:

Fingerprint distance:

1. For MPF, MPtrj and OMat24 models: We took the relaxed geometries produced by WBM, applied small perturbations to them, and used these perturbed structures as inputs. We then computed the fingerprint distance between the original WBM geometries and the geometries re-optimized by each model.
2. For MPF and OMat24 models: We took the re-relaxed geometries produced by ourselves, applied small perturbations to those structures, and used them as inputs. We then calculated the fingerprint distance between our re-relaxed geometries and the geometries optimized by each model.
   All perturbations were applied using the latest version of pymatgen.

Formation energy:

1. For MPF, MPtrj and OMat24 models: ...
2. For MPtrj models: ...
3. For MatPES models: ...

The new EquilibriumBenchmark method in MatCalc will be updated soon.

Major changes:

• feature 1: ...
• fix 1: ...

Todos

If this is work in progress, what else needs to be done?

• feature 2: ...
• fix 2:

Checklist

• Google format doc strings added. Check with ruff.
• Type annotations included. Check with mypy.
• Tests added for new features/fixes.
• If applicable, new classes/functions/modules have duecredit @due.dcite decorators to reference relevant papers by DOI (example)

Tip: Install pre-commit hooks to auto-check types and linting before every commit:

pip install -U pre-commit
pre-commit install