arXiv Science⌕ Search

arXiv subjects

Xiao Yao

Publications and source records attributed to Xiao Yao.

13 recordsLinked to original sources
arXiv2026

Tree-of-Ideas: Automated Research Ideation via Cross-Trajectory Reasoning over Scholarly Evolution

Effective research ideation requires moving beyond a static understanding of prior work to trace how research problems and solutions evolve across the literature. Existing methods either treat papers as unstructured context or model scholarly evolution as isolated citation chains, overlooking interactions among research trajectories. We propose Tree-of-Ideas (ToI), a two-stage framework. EvoTrace reconstructs branching scholarly trajectories from citations, tracking evolving methods, resolved problems, and gaps. EvoAgent then reasons across trajectories to identify convergent problems and complementary solutions, generating grounded research ideas. Across six AI research topics, ToI achieves the highest score among automatic methods (6.27 vs. 5.36 for the strongest baseline on a 10-point scale), with strong Novelty (6.36) and Groundedness (7.00). Also, its score approaches that of human-paper references (6.29), demonstrating the value of cross-path evolutionary reasoning.

cs.AI↗
arXiv2026

CanvasAgent: Enabling Complex Image Creation and Editing via Visual Tool Orchestration

Complex image creation and editing often require more than a single generation or editing model. A user request may involve synthesizing images, localizing objects, segmenting regions, editing selected content, compositing intermediate assets, reading text, and enhancing the final result. Such tasks shift multimodal agents from perception-augmented reasoning to manipulation-centered visual creation, where tools must actively transform visual states rather than merely inspect them. However, existing multimodal tool-use agents are mostly optimized for perception, search, or domain-specific editing, and lack large-scale supervision for executable image-creation trajectories. In this paper, we introduce CanvasCraft, a large-scale multimodal tool-use dataset for complex image creation and editing, and \textbf{CanvasAgent}, a tool-augmented multimodal agent that learns to orchestrate heterogeneous visual tools through multi-turn interaction. CanvasCraft contains 140K fully annotated executable trajectories and 10K RL task specifications. CanvasAgent is first trained with SFT to learn executable reasoning-action trajectories, and is then optimized with GRPO using a hybrid reward that combines outcome- and process-level signals. During rollout, CanvasAgent inspects intermediate results, tracks visual assets, and adapts tool decisions to the evolving visual state. Experiments evaluate both final image quality and trajectory behavior, demonstrating the effectiveness of CanvasAgent and the proposed dataset for complex multi-tool image creation workflows.

cs.CV↗
arXiv2025

On uniqueness of functions in the extended Selberg class with moving targets

We study the question of when two functions L_1,L_2 in the extended Selberg class are identical in terms of the zeros of L_i-h(i=1,2). Here, the meromorphic function h is called moving target. With the assumption on the growth order of h, we prove that L_1\equiv L_2 if L_1-h and L_2-h have the same zeros counting multiplicities. Moreover, we also construct some examples to show that the assumption is necessary. Compared with the known methods in the literature of this area, we developed a new strategy which is based on the transcendental directions first proposed in the study of distribution of Julia set in complex dynamical system. This may be of independent interest.

math.CV↗
arXiv2024

Bayesian Variable Selection via Hierarchical Gaussian Process Model in Computer Experiments

Identifying the active factors that have significant impacts on the output of the complex system is an important but challenging variable selection problem in computer experiments. In this paper, a Bayesian hierarchical Gaussian process model is developed and some latent indicator variables are embedded into this setting for the sake of labelling the important variables. The parameter estimation and variable selection can be processed simultaneously in a full Bayesian framework through an efficient Markov Chain Monte Carlo (MCMC) method -- Metropolis-within-Gibbs sampler. The much better performances of the proposed method compared with the related competitors are evaluated by the analysis of simulated examples and a practical application.

stat.ME↗
arXiv2023

Difference independence of the Euler gamma function

In this paper, we established a sharp version of the difference analogue of the celebrated H\"{o}lder's theorem concerning the differential independence of the Euler gamma function $\Gamma$. More precisely, if $P$ is a polynomial of $n+1$ variables in $\mathbb{C}[X, Y_0,\dots, Y_{n-1}]$ such that \begin{equation*} P(s, \Gamma(s+a_0), \dots, \Gamma(s+a_{n-1}))\equiv 0 \end{equation*} for some $(a_0, \dots, a_{n-1})\in \mathbb{C}^{n}$ and $a_i-a_j\notin \mathbb{Z}$ for any $0\leq i<j\leq n-1$, then we have $$P\equiv 0.$$ Our result complements a classical result of algebraic differential independence of the Euler gamma function proved by H\"{o}lder in 1886, and also a result of algebraic difference independence of the Riemann zeta function proved by Chiang and Feng in 2006.

math.NT↗
arXiv2022

Rationality of meromorphic functions between real algebraic sets in the plane

We study one variable meromorphic functions mapping a planar real algebraic set $A$ to another real algebraic set in the complex plane. By using the theory of Schwarz reflection functions, we show that for certain $A$, these meromorphic functions must be rational. In particular, when $A$ is the standard unit circle, we obtain an one dimensional analog of Poincar\'e(1907), Tanaka(1962) and Alexander(1974)'s rationality results for $2m-1$ dimensional sphere in $\mathbb{C}^m$ when $m\ge 2$.

math.CV↗
arXiv2021

The spectrum of period-doubling Hamiltonian

In this paper, we show the following: the Hausdorff dimension of the spectrum of period-doubling Hamiltonian is bigger than $\log \alpha/\log 4$, where $\alpha$ is the Golden number; there exists a dense uncountable subset of the spectrum such that for each energy in this set, the related trace orbit is unbounded, which is in contrast with a recent result of Carvalho (Nonlinearity 33, 2020); we give a complete characterization for the structure of gaps and the gap labelling of the spectrum. All of these results are consequences of an intrinsic coding of the spectrum we construct in this paper.

math.SP↗
arXiv2020

Inequalities Concerning Maximum Modulus and Zeros of Random Entire Functions

Let $f_\omega(z)=\sum\limits_{j=0}^{\infty}\chi_j(\omega) a_j z^j$ be a random entire function, where $\chi_j(\omega)$ are independent and identically distributed random variables defined on a probability space $(\Omega, \mathcal{F}, \mu)$. In this paper, we first define a family of random entire functions, which includes Gaussian, Rademacher, Steinhaus entire functions. Then, we prove that, for almost all functions in the family and for any constant $C>1$, there exist a constant $r_0=r_0(\omega)$ and a set $E\subset [e, \infty)$ of finite logarithmic measure such that, for $r>r_0$ and $r\notin E$, $$ |\log M(r, f)- N(r,0, f_\omega)|\le (C/A)^{\frac1{B}}\log^{\frac1{B}}\log M(r,f) +\log\log M(r, f), \qquad a.s. $$ where $A, B$ are constants, $M(r, f)$ is the maximum modulus, and $N(r, 0, f)$ is the weighted counting-zero function of $f$. As a by-product of our main results, we prove Nevanlinna's second main theorem for random entire functions. Thus, the characteristic function of almost all functions in the family is bounded above by a weighed counting function, rather than by two weighted counting functions in the classical Nevanlinna theory. For instance, we show that, for almost all Gaussian entire functions $f_\omega$ and for any $\epsilon>0$, there is $r_0$ such that, for $r>r_0$, $$ T(r, f) \le N(r,0, f_\omega)+(\frac12+\epsilon) \log T(r, f). $$

math.CV↗
arXiv2018

Large scaled geometry of Julia sets of entire and meromorphic functions

In this paper, we study the large scaled geometric structure of Julia sets of entire and meromorphic functions. Roughly speaking, the structure gives us some asymptotic information about the Julia set near the essential singularity. We will show that part of this structure are determined by the transcendental directions coming from function theoretic point of view.

math.DS↗
arXiv2016

Lyapunov exponents and related concepts for entire functions

Let $f$ be an entire function and denote by $f^\#$ be the spherical derivative of $f$ and by $f^n$ the $n$-th iterate of $f$. For an open set $U$ intersecting the Julia set $J(f)$, we consider how fast $\sup_{z\in U} (f^n)^\#(z)$ and $\int_U (f^n)^\#(z)^2 dx\:dy$ tend to $\infty$. We also study the growth rate of the sequence $(f^n)^\#(z)$ for $z\in J(f)$.

math.DS↗
arXiv2015

"Mixed spectral nature" of Thue--Morse Hamiltonian

We find three dense subsets $\Sigma_I,\Sigma_{II}$ and $\Sigma_{III}$ of the spectrum of the Thue-Morse Hamiltonian, such that each energy in $\Sigma_I$ is extended, each energy in $\Sigma_{II}$ is pseudo-localized and each energy in $\Sigma_{III}$ is one-sided pseudo-localized. We also obtain exact estimations on the norm of the transfer matrices and the norm of the formal solutions for these energies. Especially, for $E\in \Sigma_{II}\cup\Sigma_{III}, $ the norms of the transfer matrices behave like $$ e^{c_1\gamma\sqrt{n}}\le\|T_{n}(E)\|\le e^{c_2\gamma\sqrt{n}}. $$ The local dimensions of the spectral measure on these subsets are also studied. The local dimension is $0$ for energy in $\Sigma_{II}$ and is larger than $1$ for energy in $\Sigma_{I}\cup\Sigma_{III}.$ In summary, the Thue-Morse Hamiltonian exhibits "mixed spectral nature".

math-ph↗
arXiv2015

A new geometric characterization of the Julia set

This article concerns a new geometric characterization of the Julia set. By using Ahlfors-Shimizu's characteristic, we establish some growth results which indicates the characterization of the Julia set. The main technique is to estimate the lower bound of $S(f^n,U)$, where $U$ is an open neighbourhood of some point in $\mathcal{J}(f)$.

math.DS↗