Post-CCSD(T) corrections in the S66 noncovalent interactions benchmark

For noncovalent interactions, it is generally assumed that CCSD(T) approaches the exact solution within the basis set. For most of the S66 benchmark, we present CCSDT and CCSDT(Q) corrections with a DZP basis set. For hydrogen bonds, pure London, and mixed-influence complexes, CCSD(T) benefits from error cancellation between (repulsive) higher-order triples, T₃−(T), and (attractive) connected quadruples, (Q). For π-stacking complexes, this cancellation starts breaking down and CCSD(T) overbinds; CCSD(T)_Λ corrects the problem at the expense of London complexes. Simple two- or three-parameter models predict CCSDT(Q)–CCSD(T) differences to 0.01 kcal mol⁻¹RMS, requiring no calculations with steeper scaling than O(N⁷).