Search arXiv⌕ Search

arXiv subjects

Teng Zhang

Publications and source records attributed to Teng Zhang.

At least 55 records · Page 3Linked to original sources
arXiv2026

EnchantDance: Unveiling the Potential of Music-Driven Dance Movement

The task of music-driven dance generation involves creating coherent dance movements that correspond to the given music. While existing methods can produce physically plausible dances, they often struggle to generalize to out-of-set data. The challenge arises from three aspects: 1) the high diversity of dance movements and significant differences in the distribution of music modalities, which make it difficult to generate music-aligned dance movements. 2) the lack of a large-scale music-dance dataset, which hinders the generation of generalized dance movements from music. 3) The protracted nature of dance movements poses a challenge to the maintenance of a consistent dance style. In this work, we introduce the EnchantDance framework, a state-of-the-art method for dance generation. Due to the redundancy of the original dance sequence along the time axis, EnchantDance first constructs a strong dance latent space and then trains a dance diffusion model on the dance latent space. To address the data gap, we construct a large-scale music-dance dataset, ChoreoSpectrum3D Dataset, which includes four dance genres and has a total duration of 70.32 hours, making it the largest reported music-dance dataset to date. To enhance consistency between music genre and dance style, we pre-train a music genre prediction network using transfer learning and incorporate music genre as extra conditional information in the training of the dance diffusion model. Extensive experiments demonstrate that our proposed framework achieves state-of-the-art performance on dance quality, diversity, and consistency.

cs.SD↗
arXiv2026

DC response of an interferometer topology with an L-shaped cavity: a tabletop study

A new interferometer topology for kilohertz gravitational-wave detection was recently proposed in [Zhang et al. Phys. Rev. X 13, 021019 (2023)]. The design is based on an L-shaped optical cavity pumped through a Sagnac-like vortex. We report a tabletop experiment that characterizes the interferometer's optical response near DC. When the laser frequency is locked to the resonance of the L-shaped cavity, we observe that the cavity input coupler becomes effectively transparent, yielding a simple Michelson-like response. Moreover, the Sagnac vortex separates into upper and lower paths, which behave as two independent pumping paths driving the cavity. These observations are in agreement with theoretical predictions. Our results provide an intuitive physical picture of this interferometer topology and offer insight into its lock acquisition strategy.

physics.ins-det↗
arXiv2026

The Sharp Phase Transition of Tyler's M-Estimator for Robust Subspace Recovery

Robust Subspace Recovery (RSR) aims to identify an underlying d-dimensional subspace from a dataset heavily corrupted by outliers. Complexity-theoretic results establish a threshold for the problem's computational hardness based on the dimension-scaled signal-to-noise ratio (DS-SNR): the problem is SSE-hard when the DS-SNR is strictly less than 1, and solvable via practical algorithms when it is greater than 1 under general position assumptions. However, the exact behavior of practical algorithms at the critical boundary DS-SNR = 1 has remained unknown. This work resolves the behavior of Tyler's M-estimator (TME) at this critical boundary, consequently establishing a sharp phase transition. Specifically, we prove that TME converges exactly to the true subspace for DS-SNR \geq 1 under a new stability condition, which is less restrictive than the general position assumptions used in prior literature. Our analysis utilizes a decomposition of the TME iterates within a majorization-minimization framework.

cs.IT↗
arXiv2026

The separable case of Kadison's problem on orthonormal bases of unitaries for type $\mathrm{II}_1$ factors

In 1967, Kadison asked ``does every type $\mathrm{II}_1$ factor have an orthonormal (with respect to the trace) basis consisting of unitaries?'' Using a noncommutative Lyapunov theorem of Akemann and Weaver, we prove that if $M$ is a separable diffuse finite von Neumann algebra with a normal faithful trace $τ$, then $L^2(M,τ)$ admits an orthonormal basis consisting of self-adjoint unitaries in $M$. Consequently, we affirm the separable case of the Kadison problem.

math.OA↗
arXiv2026

Combining Mechanical and Agentic Specification Inference for Move

In this paper, we describe early work on a specification inference tool for the Move Prover that combines a weakest-precondition (WP) analysis over Move bytecode with an agentic coding CLI such as Claude Code. Specification inference reduces the boilerplate of writing specifications in Move: in order to verify a high-level property such as a global state invariant, pre- and post-conditions for the supporting functions typically have to be written by hand, which is tedious. In our setting, a Model Context Protocol (MCP) service exposes the WP analysis and the prover itself to the coding agent. The WP analysis provides a sound, mechanical baseline for inference; the AI is used precisely where WP is weakest -- for loop invariants and high-level idiomatic specifications such as monotonicity, conservation, and structural invariants. The Move Prover serves as the oracle that decides whether the generated specs are valid, and the agent is equipped to generate proof hints and to refine the inferred specification until verification succeeds. The tool has been applied to a corpus of canonical Move code, including code that uses higher-order functions, dynamic dispatch, global state, references, and various forms of loops.

cs.PL↗
arXiv2026

A Bayesian Adaptive Latent Mixture Model for Zero-Inflated Weighted Brain Connectome Analysis

Replicated weighted networks often exhibit many structural zeros alongside heterogeneous non-zero edge strengths. In structural connectomics, this zero-inflation coincides with subjects expressing overlapping, rather than discrete, connectivity patterns. To address these features, we propose a Bayesian adaptive latent mixture model for zero-inflated weighted networks. Our approach represents each subject network as a simplex mixture of shared low-rank latent score matrices, integrated with a hurdle likelihood that separates edge existence from conditional edge strength. A sparsity-coupling parameter enables absent edges to be either independent of, or informative about, the latent connectivity. For computation, we employ transformed Hamiltonian Monte Carlo on unconstrained coordinates, selecting the number of templates via predictive fit, held-out link prediction, and template stability. Theoretically, we establish posterior consistency, local asymptotic normality, a Bernstein--von Mises approximation, and predictive consistency for an identifiable quotient-space estimand under a fixed-template scenario. Simulations demonstrate performance gains over topology-only baselines in settings with mixed memberships or structure-informed sparsity. Applied to Human Connectome Project data, the model recovers stable latent score patterns and heterogeneous subject-level mixtures, with behavioural analyses serving strictly as exploratory annotations rather than confirmatory biomarker claims.

stat.ME↗
arXiv2026

ReflectDrive-2: Reinforcement-Learning-Aligned Self-Editing for Discrete Diffusion Driving

We introduce ReflectDrive-2, a masked discrete diffusion planner with separate action expert for autonomous driving that represents plans as discrete trajectory tokens and generates them through parallel masked decoding. This discrete token space enables in-place trajectory revision: AutoEdit rewrites selected tokens using the same model, without requiring an auxiliary refinement network. To train this capability, we use a two-stage procedure. First, we construct structure-aware perturbations of expert trajectories along longitudinal progress and lateral heading directions and supervise the model to recover the original expert trajectory. We then fine-tune the full decision--draft--reflect rollout with reinforcement learning (RL), assigning terminal driving reward to the final post-edit trajectory and propagating policy-gradient credit through full-rollout transitions. Full-rollout RL proves crucial for coupling drafting and editing: under supervised training alone, inference-time AutoEdit improves PDMS by at most $0.3$, whereas RL increases its gain to $1.9$. We also co-design an efficient reflective decoding stack for the decision--draft--reflect pipeline, combining shared-prefix KV reuse, Alternating Step Decode, and fused on-device unmasking. On NAVSIM, ReflectDrive-2 achieves $91.0$ PDMS with camera-only input and $94.8$ PDMS in a best-of-6 oracle setting, while running at $31.8$ ms average latency on NVIDIA Thor.

cs.RO↗
arXiv2026

Colored Stallings graphs and Counterexamples to Stallings equalizer conjecture

The famous Stallings equalizer conjecture has remained open for more than 40 years, which states that, for any free group \(F_n\) of rank \(n\ge 2\), any free group \(F\), and any two monomorphisms $g,h:F_n\to F,$ the equalizer $\Eq(g,h)=\{w\in F_n\mid g(w)=h(w)\}$ satisfies $\rk \Eq(g,h)\le n.$ The only known case is $n=2$, due to A. D. Logan in 2022. By introducing the notion of colored Stallings graphs, we show that for every integer \(n\ge 2\) there exist monomorphisms $g,h:F_n\longrightarrow F_2$ such that$\rk\Eq(g,h)\ge 2n-2.$ This disproves Stallings equalizer conjecture for $n\ge 3$.

math.GR↗
arXiv2026

Proof of the Agler--McCarthy entropy conjecture

In 2021, J.~Agler and J.~E. McCarthy proposed a two-step programme toward the celebrated Krzyż conjecture. The first step is to prove an entropy conjecture for polynomials whose zeros all lie on the unit circle; the second is to establish a full degree condition for extremal functions in the Krzyż conjecture. The purpose of this paper is to complete the first step. More precisely, we establish the sharp homogeneous entropy inequality for all non-constant polynomials with zeros on the unit circle and determine the equality cases.

math.CV↗
arXiv2026

Multistable energy landscapes for adaptive microscopic machines

The past few years have seen great strides in our ability to build synthetic microscopic machines. However, the function of such machines is often controlled directly by externally applied fields that deterministically specify the instantaneous machine dynamics. A crucial step towards machines that can respond adaptively to changes in their environment is the ability to program multiple functions that actuate under the same external driving field, so that their internal state dictates which function is executed. Here, we demonstrate that energy landscapes with designed multistability enable the same externally applied field to drive multiple configurations and dynamic responses in microscopic machines, enabling increasing levels of autonomy. We show three examples. First, we write a bistable energy landscape into a microscopic device, enabling the device to exhibit two stable mechanical configurations under the same external magnetic field. Next, adding a second degree of freedom enables differing dynamic responses to the same external magnetic field, which we direct into net displacement of the environment. Finally, we demonstrate how a microscopic machine with a continuous symmetry autonomously channels a single degree-of-freedom magnetic actuation into locomotion and adaptively responds to forces induced by other machines.

cond-mat.soft↗
arXiv2026

Sharp condition-number bounds for growth factors of Higham matrices in Gaussian elimination

Higham's conjecture on the growth factor of complex symmetric positive definite matrices is a longstanding problem in the stability theory of Gaussian elimination without pivoting. It asserts that every complex matrix $A=B+iC$ with $B$ and $C$ real symmetric positive definite, is called Higham matrix and has growth factor $ρ_n(A)<2$. In 2013, Drury [Linear Algebra Appl. \textbf{439} (2013), no.~10, 3129--3133] proved that $ρ_n(A)\le 2$. In fact, we will see his sectorial determinant method can be refined to give the strict bound $ρ_n(A)<2$ for each fixed Higham matrix; however, the resulting constant $1+δ_A^2$ depends on the matrix $A$. In this paper, we establish sharp condition-number-dependent lower and upper bounds for the growth factors of Higham matrices, thereby providing a quantitative refinement of Drury's result. The main ingredient is a sharp scalar Schur-complement inequality, proved via a two-dimensional domination.We also obtain corresponding sharp scalar and diagonal estimates for accretive-dissipative matrices, and an improved entrywise growth bound for that broader class.

math.NA↗
arXiv2026

An operator triangle inequality for the quadratic symmetric modulus

50 years after Thompson's famous triangle inequality for the operator right modulus, we establish a triangle inequality for the quadratic symmetric modulus. We also discuss the corresponding equality cases as well as the infinite-dimensional setting. In addition, we obtain Clarkson--McCarthy type inequalities for the quadratic symmetric modulus. Moreover, we answer several questions raised by Bourin and Lee in [\emph{Bull. Lond. Math. Soc.} \textbf{44} (2012), no.~6, 1085--1102] and [\emph{Internat. J. Math.} \textbf{31} (2026), no.~6, 2650018].

math.FA↗
arXiv2026

Plank theorems, Gaussian probabilistic estimates and Rump's 100 Euro conjecture

We prove Rump's 100-euro conjecture by deriving a weighted affine escape theorem from Ball's plank theorem in [Invent. Math. \textbf{104} (1991)]. More precisely, let $\mathbb{K}\in\{\mathbb{R},\mathbb{C}\}$ and let $A\in\mathbb{K}^{n\times n}$. For every $1\le p\le \infty$, we obtain an $\ell_p$-escape principle controlled by the row $\ell_q$-norms of $A$. Its cube case shows that $|A|e=ne$, where $e$ is the all-one vector, implies the existence of a nonzero vector $x$ satisfying $\|x\|_{\infty}\le 1$ and $|Ax|\ge e\ge |x|$, thereby settling the conjecture. As a consequence, we prove the global comparison $ρ_0(|A|)\le n\,ρ_{\mathbb{K}}(A)$,where $ρ_{\mathbb{K}}$ denotes the sign-real or complex spectral radius, respectively. This is the sharp form of Rump's Perron--Frobenius-type estimate, with the factor $3+2\sqrt{2}$ removed. Moreover, our $\ell_\infty$-escape principle sharpens Rump's result in [SIAM Rev. \textbf{41} (1999)] on the relation between the entrywise distance to singularity of a matrix and its entrywise Bauer--Skeel condition number. Finally, we also investigate the weaker Euclidean row condition, including sharp quantitative bounds and counterexamples to possible strengthenings. In particular, we use Gaussian probabilistic estimates to establish a complex analogue of a conjecture of Bünger.

math.FA↗
arXiv2026

Clarkson--McCarthy type inequalities, part I: $\ell_p$--$\ell_p$ and $\ell_q$--$\ell_p$ Schatten $p$-estimates

We characterize the matrices $U=(u_{ij})$ for which the operator square-sum identity $$\sum_{i=1}^m\Big|\sum_{j=1}^n u_{ij}A_j\Big|^2=\sum_{j=1}^n|A_j|^2$$ holds for all Schatten-class operators $A_1,\ldots,A_n$; this happens exactly when $U$ is an isometry.Using this characterization, we establish Clarkson--McCarthy type inequalities for several classes of operator families, including $\ell_p$--$\ell_p$ estimates and mixed $\ell_q$--$\ell_p$ estimates.We also obtain a multivariable extension of the Ball--Carlen--Lieb $2$-uniform convexity inequality and a weaker bound toward Audenaert's norm-compression conjecture.

math.FA↗
arXiv2026

A Solution to a Problem of Rubel on Two-Parameter Normal Families of Entire Functions

We construct an entire function $ F(z,a,b)\in \mathcal{O}(\mathbb{C}^3) $ such that the family $$ \{F(\,\cdot\,,a,b):a,b\in\mathbb{C}\} $$ of entire functions of \(z\) is normal on \(\mathbb{C}\), while \(F\) does not factor through a single entire parameter. This solves a problem of L.~A.~Rubel concerning Liouville-type rigidity. In fact, our example satisfies the stronger condition $$ F_bF_{a,z}-F_aF_{b,z}\neq 0 \qquad\text{on }\mathbb{C}^3. $$ The geometric core of the construction is a Fatou--Bieberbach domain contained in the thin region $$ \{(u,v)\in\mathbb{C}^2:|u-v^2|<1+|v|\}. $$ We obtain this domain from the basin of attraction of an explicit polynomial automorphism of \(\mathbb{C}^2\), together with the theorem of Rosay and Rudin on attracting basins.

math.CV↗
arXiv2026

Global Convergence of Iteratively Reweighted Least Squares for Robust Subspace Recovery

Robust subspace estimation is fundamental to many machine learning and data analysis tasks. Iteratively Reweighted Least Squares (IRLS) is an elegant and empirically effective approach to this problem, yet its theoretical properties remain poorly understood. This paper establishes that, under deterministic conditions, a variant of IRLS with dynamic smoothing regularization converges linearly to the underlying subspace from any initialization. We extend these guarantees to affine subspace estimation, a setting that lacks prior recovery theory. Additionally, we illustrate the practical benefits of IRLS through an application to low-dimensional neural network training. Our results provide the first global convergence guarantees for IRLS in robust subspace recovery and, more broadly, for nonconvex IRLS on a Riemannian manifold.

stat.ML↗
arXiv2026

On a conjecture of $λ$-Aluthge transforms and Hilbert--Schmidt self-commutators

Let $A$ be a complex square matrix, and write its polar decomposition as $A=U|A|$. For $0<λ<1$, the $λ$-Aluthge transform of $A$ is defined by $$ Δ_λ(A)=|A|^λU|A|^{1-λ}. $$ In 2007, Huang and Tam conjectured that the Frobenius norm of the self-commutator is contractive under $Δ_λ$: for every $0<λ<1$, $$ \|A^*A-AA^*\|_{F} \ \ge\ \|Δ_λ(A)^*Δ_λ(A)-Δ_λ(A)Δ_λ(A)^*\|_{F}. $$ If this inequality held, then the iterated self-commutator norms $$ \Bigl\{\bigl\|Δ_λ^{\,m}(A)^*Δ_λ^{\,m}(A) -Δ_λ^{\,m}(A)Δ_λ^{\,m}(A)^*\bigr\|_F\Bigr\}_{m\in\mathbb N} $$ would form a nonincreasing sequence and necessarily converge to $0$. In this paper we provide a counterexample, thereby disproving the conjecture. We also obtain the quantitative bounds $$ \sqrt{\frac32}\ \le\ \sup_{\substack{A\in\mathbb{M}_n(\mathbb{C}),\ A^*A\neq AA^*\\ 0<λ<1}} \frac{\|Δ_λ(A)^*Δ_λ(A)-Δ_λ(A)Δ_λ(A)^*\|_F}{\|A^*A-AA^*\|_F} \ \le\ 2. $$

math.FA↗