> Leaderboard
Rankings based on weighted MAE across benchmark datasets. Lower is better.
B
R2SCAN (Baseline)
The functional to beat
1.0000
normalized score
rankings
Loading leaderboard...
$ scoring_methodology
Each functional is evaluated on multiple benchmark datasets covering atomization energies (W4-11), barrier heights (BH76), non-covalent interactions (S22), and bond lengths. The score is a weighted mean absolute error (WMAE) normalized so that R2SCAN = 1.0000.
A score below 1.0 means your functional outperforms R2SCAN on our benchmark suite. The weights balance accuracy across different chemical properties.