Non-invertible symmetries — symmetries whose composition is not a simple group operation but a fusion rule — generalize the symmetry concept beyond groups. Fusion rules permit a product of two symmetry operations to yield a sum of outcomes, not a single one. These structures appear in orbifold compactifications of string theory where the orbifold group is non-Abelian.
At tree level, fusion rules impose selection rules on couplings: certain interactions are forbidden by the non-invertible symmetry. At loop level, quantum corrections violate these rules. Forbidden couplings become permitted. The full non-invertible symmetry is broken by loops.
But not all of it. The paper identifies a residual group-like symmetry — a conventional, invertible subgroup — that survives the loop corrections exactly. The procedure, called “groupification,” extracts the largest group structure that remains intact after quantum corrections break the full fusion algebra. Both Abelian and non-Abelian residual symmetries survive.
The practical consequence: most parameters that the full non-invertible symmetry would set to zero are natural in 't Hooft's sense under the residual group symmetry. They're small not because of fine-tuning but because the surviving symmetry protects them. The non-invertible symmetry was too strong — it forbade couplings that quantum mechanics insists on generating — but its group-like core is exactly right. The quantum theory breaks the exotic part of the symmetry while preserving the conventional part, and the conventional part is sufficient to ensure naturalness. The symmetry that survives is the symmetry that matters.