arXiv Science⌕ Search

arXiv subjects

Haoxuan Cheng

Publications and source records attributed to Haoxuan Cheng.

13 recordsLinked to original sources

Prescribed inverse-weight Lin-Lu-Yau curvature flow on finite trees

We characterize the image of the inverse-weight Lin-Lu-Yau curvature map on finite edge-weighted trees. We prove that the prescribed curvature flow converges if and only if the target curvature is attainable, and that the convergence is exponential whenever it occurs. The key ingredient in the proof is a formulation of the prescribed curvature flow as the gradient flow of a convex functional.

math.DG↗

CrashGym: Measuring the Decision Value of Experimentation for LLM Agents in Counterfactual Market Crashes

Language-model agents are increasingly asked to run experiments, yet benchmarks rarely measure whether those experiments improve the decisions they are meant to inform. We introduce CrashGym, a deterministic, replayable limit-order-book simulator in which an agent must choose between a deleveraging intervention and a liquidity backstop for a market about to suffer a shock-triggered crash. Before committing, the agent may query intervention experiments in a laboratory realization of the market. Its choice is then scored on an independently seeded target realization in which the simulator executes every feasible intervention, so regret relative to the best intervention is computed by directly simulating every alternative on the same target repeats, with no human or model-based judging. The simulator implements each amplification channel---persistent price impact, liquidity withdrawal, margin-driven liquidation, and an evolving strategy mix---as an explicit mechanism that can be switched on in stages. In a prospectively specified study on 24 newly sampled worlds, laboratory access reduces the mean regret of the one deployment with a complete record (Luna) from 8.878 to 3.785, a paired reduction of 5.092 utility units (nominal 95% world-bootstrap interval: 1.028--9.507), or 57.4% of the OBSERVE-ONLY mean. Prospective comparisons for two further deployments are incomplete; in a separate exploratory study, mean regret is lower with laboratory access for all four deployments evaluated. A laboratory whose strategy mix evolves endogenously yields no established advantage over one with a fixed mix. CrashGym thus measures the value of experimentation as avoided decision loss and lets simulator detail be judged by the decisions it supports.

cs.AI↗

K-area bounds for smooth sphere bundles

Let $E\to B$ be a smooth $S^n$ bundle with oriented Riemannian total space $(E,g)$, where $n\ge2$. We prove K-area$(E,g)\le4πn(n-1)/σ$ whenever every fibre has intrinsic scalar curvature at least $σ>0$, independently of the dimension of $B$. In particular, unit round fibres give K-area$(E,g)\le4π$. For arbitrary metrics on two-sphere fibres, K-area is bounded by the supremum of their areas. We also obtain stronger bounds from fibrewise scalar spectra.

math.DG↗

Umbilic slopes and cubic Weingarten surfaces

We study umbilic slopes and the global classification of cubic Weingarten surfaces. For a smooth surface in Euclidean three-space, a nonconstant principal curvature germ with a unique secant tangent has slope zero, minus one, infinity, an odd integer at least three, or its reciprocal. At a nonconstant umbilic germ, a smooth regular relation between mean and Gaussian curvature supplies such a tangent after a continuous labeling of the principal curvatures; the ordered curvature image has at most two limiting directions, both in the same set. We also classify compact immersed surfaces without boundary of class $C^4$ satisfying $κ_2=cκ_1^3$ in some order at every point, for a fixed real $c$: each connected component is embedded as an ellipsoid of revolution. The arguments share a classification of homogeneous Hessian identities at an umbilic. Scalar unique continuation treats infinite-order contact in the smooth slope problem, whereas an exact fourth-order growth estimate and angular energy estimates yield the cubic classification.

math.DG↗

Bounded curvature manifolds without Euclidean isometric immersions of bounded mean curvature

For every integer $n\ge2$, we construct a smooth complete Riemannian metric $G_n$ on $\mathbb{R}^n$ with full curvature norm at most one and injectivity radius at least one for which no $C^2$ isometric immersion into a finite-dimensional Euclidean space has bounded mean curvature. In dimension two, bounded second fundamental form would give uniformly controlled finite Jacobian representations of the Laplacian of the conformal factor. We construct disjoint conformal blocks for which the curvature remains bounded while the finite Jacobian representation cost tends to infinity. Taking Euclidean products gives all higher dimensions. The result answers Yau's Problem~52 negatively under the stronger assumption of bounded full curvature.

math.DG↗

A counterexample to Yau's bounded-mean-curvature embedding question

Yau asked whether every complete Riemannian manifold with bounded Ricci curvature and positive injectivity radius admits an isometric embedding into some Euclidean space with bounded mean curvature. We construct a counterexample: a smooth complete metric on $\mathbb{R}^4$ with bounded Ricci curvature, positive injectivity radius, and unbounded full curvature. The Gauss equation then forces every $C^2$ isometric immersion of this metric into a finite-dimensional Euclidean space to have unbounded mean curvature. The metric is assembled from disjoint curvature peaks whose Ricci contractions stay uniformly bounded. A short-geodesic curvature estimate and the Jacobi equation give the uniform injectivity-radius bound. The construction works in every dimension at least four, and in dimension four the Ricci curvature may be chosen to tend to zero at infinity.

math.DG↗

Unbounded normalized scalar curvature integrals in dimension four

We give counterexamples to Yau's question on normalized scalar curvature integrals in every dimension $n\geq4$. For each such $n$, there exists a smooth complete Riemannian metric $g$ on $\mathbb{R}^n$ with nonnegative Ricci curvature and a pole such that \[ \lim_{R\to\infty}R^{2-n}\int_{B_q(R)}\mathrm{Scal}_g\,\mathrm dV_g=+\infty \] for every fixed $q\in\mathbb{R}^n$. Here $B_q(R)$ denotes the geodesic ball of radius $R$ centered at $q$, and $\mathrm{Scal}_g$ is the scalar curvature of $g$.

math.DG↗

Singular Rotational Self-Similar Tori for Odd $σ_k$-Curvature Flows

For every pair of integers $3\leq k<n$ with $k$ odd, we construct a compact embedded rotational torus in $\mathbb{R}^{n+1}$ whose homothetic dilations satisfy the unnormalised $σ_k$-curvature flow in a Sobolev almost-everywhere sense. Its profile curve has Hölder regularity $C^{1,1/k}$ and Sobolev regularity $W^{2,p}$ for every $1\leq p<k/(k-1)$. Away from two singular latitudes the torus is smooth; globally, the flow equation is interpreted using the weak shape operator of the associated Lipschitz boundary. Under rotational symmetry, the self-similar equation $\langle X,ν\rangle=-σ_k$, where $X$ is the position vector and $ν$ is the unit normal, reduces to a degenerate profile system. We solve this system by combining an odd-power desingularisation, a shooting argument, uniform radial and axial bounds, and a strict gap between the shooting parameters and the cylindrical radius. No classical $C^2$ rotational torus can satisfy the soliton equation, so the loss of regularity is unavoidable within the rotational toroidal class.

math.DG↗

Volume Growth and Recurrence of Fractional Powers of the Laplace--Beltrami Operator

Let $M$ be a connected geodesically complete Riemannian manifold without boundary, write $μ$ for Riemannian volume, and set $V(o,r)=μ(B(o,r))$ for geodesic balls centered at $o$. For $0<α<2$, let $X^{(α)}$ be the process obtained by subordinating Brownian motion with an independent $α/2$-stable subordinator; its $L^2$-generator is $-(-Δ)^{α/2}$. We prove that \[ \int^\infty\frac{dt}{V(o,t^{1/α})}=\infty \] implies that \(X^{(α)}\) is recurrent. The proof uses a spectral trace estimate and radial cutoffs on $M\times(0,\infty)$, where the auxiliary measure is $y^{1-α}\,dμ\,dy$. It proves the sufficient implication in Grigor'yan's Problem~26.

math.PR↗

Discrete Einstein metrics on trees

We establish the existence and uniqueness of discrete Einstein metrics on trees under Lin-Lu-Yau Ricci curvature using Perron-Frobenius theory. We establish a sharp upper bound for the largest eigenvalue of the associated Ricci matrix in terms of the maximum degree. Turning to structural properties, notably, the existence of a positive-curvature Einstein metric implies the tree must be a caterpillar. Furthermore, these metrics exhibit radial monotonicity, with edge weights decreasing strictly away from the maximal edge.

math.DG↗

Edge Subdivision and the Perron Eigenvalue of Tree Ricci Matrices

The Ricci matrix $R_T$ of a finite tree encodes its discrete Einstein metrics via the Perron eigenvector, with Lin-Lu-Yau's Ollivier Ricci curvature: $κ= -λ_{\max}(R_T)$. We show that edge subdivision, the natural operation of lengthening a tree, can decrease, preserve, or increase $λ_{\max}$. Compressing each branch into a scalar feedback function via the Schur complement reduces the spectral problem to a one-dimensional Chebyshev equation. We obtain an exact one-step trichotomy, a scalar transmission equation for arbitrary length, and the long-chain limit. Examples on double stars, including an asymmetric case where subdivision strictly increases $λ_{\max}$, illustrate the theory.

math.DG↗

Positive-Curvature Discrete Einstein Metrics on Trees

For a weighted tree, the Lin--Lu--Yau Ricci curvature admits an explicit formula in terms of the edge weights. Consequently, the constant-curvature equation is equivalent to an eigenvalue problem for an edge-indexed Ricci matrix $R_T$. Building on the spectral characterization of discrete Einstein metrics on trees, we classify all finite trees whose discrete Einstein metric has positive curvature, equivalently all trees satisfying $λ_{\max}(R_T)<0$. For caterpillars with spine order $m\ge 12$, this occurs precisely for the endpoint families $T_m(a,0,\ldots,0,b)$ with $1\le a,b\le 3$ and $(a,b)\ne(3,3)$. The remaining cases $3\le m\le 11$ are settled by an exact finite verification using rational characteristic polynomials and Sturm root counts. We also determine the zero level set $λ_{\max}(R_T)=0$: among caterpillars, it consists of the stable family $(3,0,\ldots,0,3)$ together with nine exceptional short-spine caterpillars, while $S_3^2$ is the unique non-caterpillar zero example.

math.DG↗

Spectral Monotonicity under Leaf Attachment and Limiting Behavior in Discrete Einstein Trees

Let $R_T$ be the Ricci matrix of a finite tree $T$ introduced in \cite{BaiChengHua2026}, the largest eigenvalue $λ_{\max}(R_T)$ determines the sign of a discrete Einstein metric curvature on the tree. This paper investigates the asymptotic behavior of the sequence $λ_k = λ_{\max}(R_{T_k})$ obtained by repeatedly adding pendant edges at a fixed vertex. We prove that $λ_k$ converges to a limit $λ_\infty$ that depends only on the local branch data of $T$, and establish a first-order asymptotic expansion: \[ λ_k = λ_\infty + \fracα{d+k} + O\!\left(\frac{1}{(d+k)^2}\right), \] where $d$ is the degree of the original vertex, and the coefficient $α$ is given by a spectral projection. As a corollary, when $α\neq 0$, $λ_k$ is eventually strictly monotonic (increasing or decreasing). This theory reveals the fine influence of local leaf addition on the global spectrum.

math.DG↗