Christopher Rowley shared the benchmark result on his biaryl test set, which showed 1.2.0 is worse than 1.0. in reproducing QM (CCSD(T)*/CBS) torsion barriers
difference in calculating MM torsion barriers
1.0 training set has a larger proportion of biaryl compounds than the new training set
energy cutoff might be too small (5 kcal/mol)
All byaryls in Roche set:
Chris’s set: