Skip to main content
. 2024 Aug 1;128(32):7888–7902. doi: 10.1021/acs.jpcb.4c03167

Table 4. Comparison of SQM, DFT-3c, and ML Methods on This Benchmark Seta.

  RMSE (kcal/mol)
specification whole set neutral subset charged subset
GFN2-XTB
graphic file with name jp4c03167_m160.jpg
graphic file with name jp4c03167_m161.jpg
graphic file with name jp4c03167_m162.jpg
AIMNET2
graphic file with name jp4c03167_m163.jpg
graphic file with name jp4c03167_m164.jpg
graphic file with name jp4c03167_m165.jpg
MACE-OFF23
graphic file with name jp4c03167_m166.jpg
graphic file with name jp4c03167_m167.jpg
graphic file with name jp4c03167_m168.jpg
HF-3c
graphic file with name jp4c03167_m169.jpg
graphic file with name jp4c03167_m170.jpg
graphic file with name jp4c03167_m171.jpg
a

The overall RMSE in torsion profile energies of the molecule set with respect to the reference level CCSD(T)/[haTQZ; δ:haDZ]//MP2/heavy-aug-cc-pVTZ level of theory for the whole benchmark set and subsets of neutral and charged molecules. The 95% confidence intervals, calculated with cinnabar, are presented as subscripts and superscripts.