# Claim 5

**Result: falsified as literally registered.** The authors write the
ν=2 formula with `∼` as a formal large-size expression. At the literal finite
cell p=H=1, σ²=1/2, t=0, the Gaussian moment calculation gives exact expected
loss **3/8**, whereas the registered universal
reading of Eqs. (19)-(20) gives **1/4**. The exact
gap is 1/8.

This does not challenge the source's asymptotic/formal use; it rejects only the
registered extension to every parameter scaling.

## Registered wording

For the canonical polyadic (symmetric, nu=2) identity-tensor target, Section 9 derives a complete closed-form solution to the gradient flow (Equations 19-20), valid across all parameter scalings.