K-polystable toric Fano varieties with small alpha invariants

Tian's alpha invariant $\alpha(X)$ — the global log canonical threshold of the anti-pluricanonical systems — gives the classical sufficient criterion for K-stability of a Fano variety: $\alpha > \frac{n}{n+1}$ suffices in dimension $n$. In the opposite direction, Fujita–Odaka showed every K-semistable $\mathbb{Q}$-Fano satisfies $\alpha \ge \frac{1}{n+1}$, with $\mathbb{P}^n$ attaining the bound. Liu–Zhuang asked (Nagoya Math. J., 2022, Question 1.5(1)): does there exist a K-semistable $\mathbb{Q}$-Fano of dimension $n$ with $\frac{1}{n+1} < \alpha(X) < \frac{1}{n}$?

Yes, for every $n \ge 2$: an explicit $n$-dimensional K-polystable toric $\mathbb{Q}$-Fano $X_n$ with $\alpha(X_n) = \frac{2}{2n+1}$, squarely inside the window. The examples are singular, so Jiang's conjecture for manifolds remains open — and the paper closes with the sharpened question: can $\alpha$ land in $(\frac{1}{n+1}, \frac{2}{2n+1})$?

Purely combinatorial, and small enough to hold in one hand. Set $\ell = 2n-1$ and work in the hyperplane lattice $M = \{\sum x_i = 0\} \subset \mathbb{Z}^{n+1}$ with cyclic indices. Take $a_i = \frac{e_i - e_{i+1}}{2}$ and $b_i = \frac{e_i - e_{i-1}}{\ell}$, let $P_n$ be their convex hull and $Q_n = P_n^{\vee}$ its polar. In "difference coordinates" $d_i = c_{i+1} - c_i$, the polar $Q_n$ becomes simply a hyperplane slice of the box $[-\ell, 2]^{n+1}$; a two-line gcd pigeonhole classifies its vertices. $X_n$ is the toric variety of the face fan of $Q_n$. Cyclic symmetry forces the barycenter of $P_n$ to be the origin, which is exactly Berman's toric criterion for K-polystability; the maximal pairing between $P_n$ and the vertices of $Q_n$ equals $\frac{2n-1}{2}$, and the Blum–Jonsson toric formula converts it to $\alpha = \frac{2}{2n+1}$.

Remark 1.6 states the pipeline plainly: the proof sketch was obtained by ChatGPT 5.5 pro; the Danus system — a specialized agent built on Rethlas, Peking University's automated conjecture-resolution system with formal verification (arXiv:2604.03789) — consolidated, verified, and wrote it up; human verification and polishing came afterwards. The paper even caveats that automated systems may have missed related references. Toric geometry is the ideal substrate for this: the entire difficulty reduces to finite lattice combinatorics — LLM-sketchable, machine-verifiable, human-auditable.