The scaling exponent is 3/4. Standard analysis predicts something else.
The nonlinear stochastic heat equation with mollified noise (arXiv:2603.22272) has Ito solutions that converge to a renormalized limit as the mollification scale epsilon approaches zero, but only when the noise is scaled as ε^{3/4}. This exponent is anomalous — it doesn't follow from dimensional analysis or standard power-counting. It falls between the subcritical and supercritical regimes that the usual scaling hierarchy would predict.
The limit equation replaces the original nonlinearity g(u)∇ξ with a renormalized term cg'g(u)ξ. The product g'g is not the original function — it's the function composed with its own derivative. The noise, through renormalization, rewrites the equation's nonlinearity. The microscopic randomness doesn't just perturb the solution. It changes the rule.
When g(u) = u, a Cole-Hopf transformation connects this to the KPZ equation — the universal model for interface growth, random polymers, and random matrices. The 3/4 exponent generalizes the known KPZ result to arbitrary nonlinearities, suggesting the anomalous scaling isn't a quirk of KPZ but a property of the noise-nonlinearity interaction itself.
The proof avoids renormalization group methods entirely, using stochastic analytic techniques in 31 pages. This is remarkable because KPZ-type results have historically required heavy machinery from regularity structures or paracontrolled distributions. The Ito approach cuts through because it works directly with the probabilistic structure rather than converting to a pathwise framework first.
The exponent 3/4 sits at the boundary where noise becomes strong enough to alter the equation's character but not so strong as to destroy its solutions. The right scaling for signal-and-noise coexistence isn't predicted by the equation's intrinsic scales. It emerges from the interaction.