We prove the record upper bound 1.8263483657186… for the Katz–Tao sum–difference exponent, valid in every abelian group without 2-torsion. Using only Shannon inequalities, conditional copies and Katz–Tao’s original directions, whose classical single-grid argument gives 11/6, we go below 11/6, obtaining 2762/1509 and 152/83. The classical proof can therefore be improved using its own tools. For every abelian group, we prove the universal bound 1.8264161855212…, below the previous record for the real Katz–Tao constant. The same constant is featured in Problem 6.30 of the AlphaEvolve study, where extremal-distribution search improved its lower bound while leaving the upper bound at 11/6. CASPAC, the computational mathematics engine of CASP, discovered the proofs through AI-directed search, and they are verified by independent exact checkers. This upload includes the paper and a reproducible source package containing complete rational certificates, proofs of every auxiliary inequality, verification code and exact verification receipts.
No takes yet. Share an insight, caveat, or question.
Atrin Farnamfar (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: