arXiv subjects

Guantao Chen

Publications and source records attributed to Guantao Chen.

At least 19 recordsLinked to original sources

arXiv2026

Cayley Tournaments Simultaneously Critical for the Clique and Dichromatic Numbers

For a tournament $T$, let $\omega(T)$ be the minimum clique number among the backedge graphs of $T$, and let $\chi(T)$ be its dichromatic number. We give a template-lifting construction. It turns a $k$-template into a regular, vertex-transitive Cayley tournament that is simultaneously $(k+1)$-$\omega$-critical and $(k+1)$-$\chi$-critical. The output is also a $(k+1)$-template. Iterating the construction, we prove that for every $k\geq3$, there is a positive even integer $m_k$ with the following property. Every $N>1$ with $N\equiv1\pmod{m_k}$ is the order of a regular, vertex-transitive Cayley tournament that is simultaneously $k$-$\omega$-critical and $k$-$\chi$-critical. This proves a conjecture of Aboulker, Aubian, Charbit, and Lopes and gives a negative answer to their bounded-certificate question when the hypothesis is $\omega(T)\geq k$. We also find the clique number of a cyclic substitution when each block satisfies $\omega=\chi$. We then describe exactly when this substitution is $\omega$-critical if the blocks are $\chi$-critical and satisfy $\omega=\chi$.

math.CO↗

arXiv2026

On the Spectra of Chromatic Number and Chromatic Index of Cyclic Covers

For a fixed integer $\ell \ge 2$, we study what values of chromatic index and chromatic number can be attained by some $\ell$-fold cyclic cover of a loopless multigraph. For edge-coloring, we first investigate the density, a fundamental lower bound for the chromatic index, and show that the density of every $\ell$-fold cyclic cover of a graph $G$ is at most that of $G$. We further prove that if $\ell$ is even, then the spectrum of chromatic indices over all $\ell$-fold cyclic covers of $G$ contains every integer between $\Delta(G)$ and $\chi'(G)$. When $\ell$ is odd, the chromatic-index spectrum need not be complete in general; for edge-chromatic critical graphs, we determine exactly which values are attainable. For vertex-coloring, we prove that if $\chi(G)\ge 3$, then the spectrum of chromatic numbers over all $\ell$-fold cyclic covers of $G$ contains every integer between $3$ and $\chi(G)$. Moreover, this spectrum contains $2$ if and only if $G$ is bipartite or $\ell$ is even.

math.CO↗

arXiv2026

Odd-Girth Bounds for Defective Edge Coloring

A $(k,d)$-edge coloring of a loopless multigraph $G$ is an edge coloring using at most $k$ colors such that the subgraph formed by each color class has maximum degree at most $d$. The least such $k$ is denoted by $\chi'_d(G)$. Let $G$ be a loopless non-bipartite multigraph with maximum degree $\Delta(G)$ and odd girth $g_0(G)$, and let $d\ge1$ be odd. We prove that \[ \chi'_d(G)\le\left\lceil\frac{g_0(G)\Delta(G)-1}{dg_0(G)-1}\right\rceil. \] For $d=1$, this is Goldberg's odd-girth refinement of Shannon's theorem, while for $g_0(G)=3$ it is the defective Shannon bound of Aboulker, Aubian, and Huang. For every odd $d>1$, every odd $g_0\ge3$, and every $\Delta>d$, an almost full ring multigraph $R(\Delta,g_0)$, an odd cycle with edge multiplicities alternating between $\lfloor\Delta/2\rfloor$ and $\lceil\Delta/2\rceil$, except that two consecutive edges have multiplicity $\lfloor\Delta/2\rfloor$, attains equality. We also derive a range in which the defective Goldberg--Seymour conjecture holds.

math.CO↗

arXiv2026

Towards the Overfull Conjecture II

Let $G$ be a simple graph with maximum degree $\Delta(G)$. A subgraph $H\subseteq G$ is $\Delta(G)$-overfull if $|E(H)|>\Delta(G)\left\lfloor |V(H)|/2\right\rfloor$. In any edge coloring of $G$, each color class restricted to $H$ is a matching of size at most $\left\lfloor |V(H)|/2\right\rfloor$. Thus, if $G$ contains a $\Delta(G)$-overfull subgraph, then $G$ cannot be edge-colored with only $\Delta(G)$ colors. By Vizing's Theorem, $\chi'(G)\le \Delta(G)+1$, and hence $G$ is class $2$. In 1986, Chetwynd and Hilton conjectured that whenever $\Delta(G)>|V(G)|/3$, the converse also holds: every class $2$ graph $G$ contains a $\Delta(G)$-overfull subgraph. This statement, commonly known as the Overfull Conjecture, is one of the most influential conjectures in graph edge coloring. It would imply a polynomial-time algorithm for determining the chromatic index of graphs $G$ with $\Delta(G)>|V(G)|/3$, and would also imply several other longstanding conjectures in the area, including the Just-overfull Conjecture and the Vertex-splitting Conjecture. In previous work, the third author verified the conjecture for large graphs $G$ with maximum degree at least $13|V(G)|/14$. In this paper, we confirm the conjecture for robust expanders satisfying certain density constraints. As a consequence, for every $0<\varepsilon<1$, the conjecture holds for all sufficiently large graphs $G$ with maximum degree at least $(1+\varepsilon)|V(G)|/2$.

math.CO↗

arXiv2026

Average degrees of edge-$\Delta$-critical multigraphs

Let $G$ be a loopless multigraph with maximum degree $\Delta(G)$, average degree $\overline{d}(G)$, density $\Gamma(G)$, and chromatic index $\chi'(G)$. A multigraph $G$ is called edge-$\Delta$-critical if $\Delta(G)=\Delta$, $\chi'(G)=\Delta(G)+1$ and $\chi'(H) \le \Delta(G)$ for every proper subgraph $H\subset G$. Vizing conjectured that if $G$ is an edge-$\Delta$-critical simple graph on $n$ vertices, then $\overline{d}(G) \ge \Delta-1+\tfrac{3}{n}$. Motivated by this, we conjecture that every edge-$\Delta$-critical multigraph $G$ satisfies $\overline{d}(G) \ge \tfrac{2\Delta+2}{3}$, which is best possible. We first give a general lower bound in this direction. For any such graph $G$, \[ \overline{d}(G) \ge \begin{cases} \frac{\sqrt{17}-3}{2}(\Delta+1) & \text{if } \Delta \le 112;\\[4pt] \frac{\Delta+\sqrt{2\Delta-1}}{2} & \text{if } \Delta \ge 113. \end{cases} \] This bound can be further improved under an additional condition on the multiplicity $\mu$. In this case, \[ \overline{d}(G)\ge \min\left\{ \frac{2\mu\Delta+2\mu(2\mu-1)}{4\mu-1},\; \frac{\sqrt{17}-3}{2}(\Delta+1) \right\}. \] We also confirm the conjecture for $\Delta \in \{2,3,4,5,6,7,8\}$. As a consequence, Goldberg's conjecture~\cite{Goldberg1984} holds for $\Delta(G)\in\{2,3,4,5\}$, that is, every multigraph $G$ with $\chi'(G)\ge \Delta(G)+1$ satisfies $\Gamma(G)\ge \Delta(G)$.

math.CO↗

arXiv2026

Dynamic Video Generation: Shaping Video Generation Across Time and Space

Diffusion models have achieved impressive performance in video generation, but their iterative denoising process remains computationally expensive due to the large number of tokens processed at each timestep. Recently, progressive resolution sampling has emerged as a promising acceleration approach by reducing latent resolution in early stages. However, scaling this idea to video generation remains challenging, as the additional temporal dimension introduces diverse spatio-temporal demands across different videos, and compressing only a single dimension often leads to limited acceleration or degraded quality. Therefore, we propose DVG, a Dynamic Video Generation framework that jointly allocates computation across time and space, automatically selecting content-aware acceleration strategies without manual tuning or retraining. DVG achieves near-lossless acceleration across models and tasks, reaching up to 7 times speedup on HunyuanVideo and HunyuanVideo-1.5, and 18 times when combined with distillation, demonstrating its potential as a key component in today's large-scale efficient video generation systems. Our code is in supplementary material and will be released on Github.

cs.CV↗

arXiv2026

On graphs with girth at least five achieving Steffen's edge coloring bound

Vizing and Gupta showed that the chromatic index $\chi'(G)$ of a graph $G$ is bounded above by $\Delta(G) + \mu(G)$, where $\Delta(G)$ and $\mu(G)$ denote the maximum degree and the maximum multiplicity of $G$, respectively. Steffen refined this bound, proving that $\chi'(G) \leq \Delta(G) + \left\lceil \mu(G)/\left\lfloor g(G)/2 \right\rfloor \right\rceil$, where $g(G)$ is the girth of the graph $G$. A {\it ring graph} is a graph obtained from a cycle by duplicating some edges. The equality in Steffen's bound is achieved by ring graphs of the form $\mu C_g$, obtained from an odd cycle $C_g$ by duplicating each edge $\mu$ times. We answer two questions posed by Stiebitz et al. regarding the characterization of graphs which achieve Steffen's bound. In particular, we show that if $G$ is a critical graph which achieves Steffen's bound with $g(G)\geq 5$ and $\chi'(G)\geq \Delta+2$, then $G$ must be a ring graph of odd girth.

math.CO↗

arXiv2026

Forecast the Principal, Stabilize the Residual: Subspace-Aware Feature Caching for Efficient Diffusion Transformers

Diffusion Transformer (DiT) models have achieved unprecedented quality in image and video generation, yet their iterative sampling process remains computationally prohibitive. To accelerate inference, feature caching methods have emerged by reusing intermediate representations across timesteps. However, existing caching approaches treat all feature components uniformly. We reveal that DiT feature spaces contain distinct principal and residual subspaces with divergent temporal behavior: the principal subspace evolves smoothly and predictably, while the residual subspace exhibits volatile, low-energy oscillations that resist accurate prediction. Building on this insight, we propose SVD-Cache, a subspace-aware caching framework that decomposes diffusion features via Singular Value Decomposition (SVD), applies exponential moving average (EMA) prediction to the dominant low-rank components, and directly reuses the residual subspace. Extensive experiments demonstrate that SVD-Cache achieves near-lossless across diverse models and methods, including 5.55$\times$ speedup on FLUX and HunyuanVideo, and compatibility with model acceleration techniques including distillation, quantization and sparse attention. Our code is in supplementary material and will be released on Github.

cs.CV↗

arXiv2026

From Sketch to Fresco: Efficient Diffusion Transformer with Progressive Resolution

Diffusion Transformers achieve impressive generative quality but remain computationally expensive due to iterative sampling. Recently, dynamic resolution sampling has emerged as a promising acceleration technique by reducing the resolution of early sampling steps. However, existing methods rely on heuristic re-noising at every resolution transition, injecting noise that breaks cross-stage consistency and forces the model to relearn global structure. In addition, these methods indiscriminately upsample the entire latent space at once without checking which regions have actually converged, causing accumulated errors, and visible artifacts. Therefore, we propose \textbf{Fresco}, a dynamic resolution framework that unifies re-noise and global structure across stages with progressive upsampling, preserving both the efficiency of low-resolution drafting and the fidelity of high-resolution refinement, with all stages aligned toward the same final target. Fresco achieves near-lossless acceleration across diverse domains and models, including 10$\times$ speedup on FLUX, and 5$\times$ on HunyuanVideo, while remaining orthogonal to distillation, quantization and feature caching, reaching 22$\times$ speedup when combined with distilled models. Our code is in supplementary material and will be released on Github.

cs.CV↗

arXiv2025

Let Features Decide Their Own Solvers: Hybrid Feature Caching for Diffusion Transformers

Diffusion Transformers offer state-of-the-art fidelity in image and video synthesis, but their iterative sampling process remains a major bottleneck due to the high cost of transformer forward passes at each timestep. To mitigate this, feature caching has emerged as a training-free acceleration technique that reuses or forecasts hidden representations. However, existing methods often apply a uniform caching strategy across all feature dimensions, ignoring their heterogeneous dynamic behaviors. Therefore, we adopt a new perspective by modeling hidden feature evolution as a mixture of ODEs across dimensions, and introduce HyCa, a Hybrid ODE solver inspired caching framework that applies dimension-wise caching strategies. HyCa achieves near-lossless acceleration across diverse domains and models, including 5.55 times speedup on FLUX, 5.56 times speedup on HunyuanVideo, 6.24 times speedup on Qwen-Image and Qwen-Image-Edit without retraining.

cs.CV↗

arXiv2025

Strong binding numbers and factors

Let $G$ be a simple graph. The $k$-th neighborhood of a vertex subset $S \subseteq V(G)$, denoted $\Lambda^k(S)$, is the set of vertices that are adjacent to at least $k$ vertices in $S$. The $k$-th binding number $\beta^k(G)$ is defined as the minimum ratio $|\Lambda^k(S)|/|S|$ over all subsets $S \subseteq V(G)$ with $|S| \ge k$ and $\Lambda^k(S) \ne V(G)$. This parameter generalizes the classical binding number introduced by Woodall. Andersen showed that the condition $\beta^1(G) \ge 1$ does not guarantee the existence of a $1$-factor in $G$, while Bar\'at et al. proved that $\beta^2(G) \ge 1$ suffices for the existence of a $2$-factor. In this paper, we extend this result to general $k \ge 2$ by showing that any graph $G$ with even $k|V(G)|$ and $\beta^k(G) \ge 1$ contains a $k$-factor. Moreover, if $G$ is additionally a split graph of even order, then it admits a $(k+1)$-factor. We also prove that any graph $G$ with $\beta^k(G) \ge 1$ contains at least $k-1$ disjoint perfect or near-perfect matchings. Finally, for any bipartite graph $G$ with bipartition $(X, Y)$, we introduce an analogue of the $k$-th binding number and show that, under the condition $\beta^k(G, X) \ge 1$, the graph admits $k$ disjoint matchings, each covering $X$.

math.CO↗

arXiv2025

Bipartite graphs with the double Hall property

The super-neighborhood of a vertex set $A$ in a graph $G$, denoted by $\Lambda^2(A)$, is the set of vertices adjacent to at least two vertices in $A$. We say that a bipartite graph $G=(X, Y)$ with $|X| \geq 2$ satisfies the double Hall property (with respect to $X$) if $|\Lambda^2(A)| \geq |A|$ for any subset $A \subseteq X$ with $|A| \geq 2$. Kostochka et al. first conjectured that if a bipartite graph $G=(X, Y)$ satisfies a slightly weaker version of the double Hall property, then $G$ contains a cycle that covers all vertices of $X$. They verified their conjecture for $|X| \leq 6$. In this paper, we extend their result to $|X| = 7$. Later, Salia conjectured that every bipartite graph satisfying the double Hall property has a cycle covering all vertices of $X$. We show that Salia's conjecture is almost equivalent to a much weaker conjecture requiring vertices in $Y$ to have high degrees. By extending a result of Bar\'at et al., we also show that Salia's conjecture holds for some graphs where the vertices of $Y$ have degree either $2$ or very high. Finally, we establish a lower bound for the maximum degree of graphs satisfying the double Hall property and present deterministic and probabilistic constructions of such graphs that approach this bound.

math.CO↗

arXiv2024

A short proof of the Goldberg-Seymour conjecture

For a multigraph $G$, $\chi'(G)$ denotes the chromatic index of $G$, $\Delta(G)$ the maximum degree of $G$, and $\Gamma(G) = \max\left\{\left\lceil \frac{2|E(H)|}{|V(H)|-1} \right\rceil: H \subseteq G \text{ and } |V(H)| \text{ odd}\right\}$. As a generalization of Vizing's classical coloring result for simple graphs, the Goldberg-Seymour conjecture, posed in the 1970s, states that $\chi'(G)=\max\{\Delta(G), \Gamma(G)\}$ or $\chi'(G)=\max\{\Delta(G) + 1, \Gamma(G)\}$. Hochbaum, Nishizeki, and Shmoys further conjectured in 1986 that such a coloring can be found in polynomial time. A long proof of the Goldberg-Seymour conjecture was announced in 2019 by Chen, Jing, and Zang, and one case in that proof was eliminated recently by Jing (but the proof is still long); and neither proof has been verified. In this paper, we give a proof of the Goldberg-Seymour conjecture that is significantly shorter and confirm the Hochbaum-Nishizeki-Shmoys conjecture by providing an $O(|V|^5|E|^3)$ time algorithm for finding a $\max\{\Delta(G) + 1, \Gamma(G)\}$-edge-coloring of $G$.

math.CO↗

arXiv2023

Decreasing the mean subtree order by adding $k$ edges

The mean subtree order of a given graph $G$, denoted $\mu(G)$, is the average number of vertices in a subtree of $G$. Let $G$ be a connected graph. Chin, Gordon, MacPhee, and Vincent [J. Graph Theory, 89(4): 413-438, 2018] conjectured that if $H$ is a proper spanning supergraph of $G$, then $\mu(H) > \mu(G)$. Cameron and Mol [J. Graph Theory, 96(3): 403-413, 2021] disproved this conjecture by showing that there are infinitely many pairs of graphs $H$ and $G$ with $H\supset G$, $V(H)=V(G)$ and $|E(H)|= |E(G)|+1$ such that $\mu(H) < \mu(G)$. They also conjectured that for every positive integer $k$, there exists a pair of graphs $G$ and $H$ with $H\supset G$, $V(H)=V(G)$ and $|E(H)| = |E(G)| +k$ such that $\mu(H) < \mu(G)$. Furthermore, they proposed that $\mu(K_m+nK_1) < \mu(K_{m, n})$ provided $n\gg m$. In this note, we confirm these two conjectures.

math.CO↗

arXiv2023

A linear algorithm for obtaining the Laplacian eigenvalues of a cograph

In this paper, we give a linear algorithm for obtaining the Laplacian eigenvalues of a cograph. This approach is more efficient as there is no need to directly compute the eigenvalues of Laplacian matrix related to this class of graph. As an application, we use this algorithm as a tool for obtaining a closed formula for the number of spanning trees of a cograph.

math.CO↗

arXiv2023

A note on Gupta's co-density conjecture

Let $G$ be a multigraph. A subset $F$ of $E(G)$ is an edge cover of $G$ if every vertex of $G$ is incident to an edge of $F$. The cover index, $\xi(G)$, is the largest number of edge covers into which the edges of $G$ can be partitioned. Clearly $\xi(G) \le \delta(G)$, the minimum degree of $G$. For $U\subseteq V(G)$, denote by $E^+(U)$ the set of edges incident to a vertex of $U$. When $|U|$ is odd, to cover all the vertices of $U$, any edge cover needs to contain at least $(|U|+1)/2$ edges from $E^+(U)$, indicating $ \xi(G) \le |E^+(U)|/ (|U|+1)/2$. Let $\rho_c(G)$, the co-density of $G$, be defined as the minimum of $|E^+(U)|/((|U|+1)/2)$ ranging over all $U\subseteq V(G)$ with $|U| $ odd and at least 3. Then $\rho_c(G)$ provides another upper bound on $\xi(G)$. Thus $\xi(G) \le \min\{\delta(G), \lfloor \rho_c(G) \rfloor \}$. For a lower bound on $\xi(G)$, in 1967, Gupta conjectured that $\xi(G) \ge \min\{\delta(G)-1, \lfloor \rho_c(G) \rfloor \}$. Gupta showed that the conjecture is true when $G$ is simple, and Cao et al. verified this conjecture when $\rho_c(G)$ is not an integer. In this note, we confirm the conjecture when the maximum multiplicity of $G$ is at most two or $ \min\{\delta(G)-1, \lfloor \rho_c(G) \rfloor \} \le 6$.

math.CO↗

arXiv2022

Overfullness of edge-critical graphs with small minimal core degree

Let $G$ be a simple graph. Denote by $n$, $Δ(G)$ and $χ' (G)$ be the order, the maximum degree and the chromatic index of $G$, respectively. We call $G$ \emph{overfull} if $|E(G)|/\lfloor n/2\rfloor > Δ(G)$, and {\it critical} if $χ'(H) < χ'(G)$ for every proper subgraph $H$ of $G$. Clearly, if $G$ is overfull then $χ'(G) = Δ(G)+1$. The \emph{core} of $G$, denoted by $G_Δ$, is the subgraph of $G$ induced by all its maximum degree vertices. We believe that utilizing the core degree condition could be considered as an approach to attacking the overfull conjecture. Along this direction, we in this paper show that for any integer $k\geq 2$, if $G$ is critical with $Δ(G)\geq \frac{2}{3}n+\frac{3k}{2}$ and $δ(G_Δ)\leq k$, then $G$ is overfull.

math.CO↗

arXiv2022

Linear arboricity of degenerate graphs

A linear forest is a union of vertex-disjoint paths, and the linear arboricity of a graph $G$, denoted by $\operatorname{la}(G)$, is the minimum number of linear forests needed to partition the edge set of $G$. Clearly, $\operatorname{la}(G) \ge \lceil\Delta(G)/2\rceil$ for a graph $G$ with maximum degree $\Delta(G)$. On the other hand, the Linear Arboricity Conjecture due to Akiyama, Exoo, and Harary from 1981 asserts that $\operatorname{la}(G) \leq \lceil(\Delta(G)+1) / 2\rceil$ for every graph $ G $. This conjecture has been verified for planar graphs and graphs whose maximum degree is at most $ 6 $, or is equal to $ 8 $ or $ 10 $. Given a positive integer $k$, a graph $G$ is $k$-degenerate if it can be reduced to a trivial graph by successive removal of vertices with degree at most $k$. We prove that for any $k$-degenerate graph $G$, $\operatorname{la}(G) = \lceil\Delta(G)/2 \rceil$ provided $\Delta(G) \ge 2k^2 -k$.

math.CO↗