arXiv subjects

Ce Chen

Publications and source records attributed to Ce Chen.

At least 19 recordsLinked to original sources

arXiv2026

JuZhou 1.0 Technical Report: The First Edge-Native Text-to-Image Foundation Model Trained Entirely on China-Developed AI Accelerators

Text-to-image (T2I) diffusion models typically require substantial computational resources and cloud infrastructure, posing significant challenges for edge deployment in terms of latency, cost, and user privacy. We present JuZhou 1.0, an ultra-lightweight T2I foundation model designed for fully offline, on-device execution. JuZhou 1.0 achieves its efficiency through four key designs: (1) a compact image-generation backbone consisting of a 0.385B-parameter denoising U-Net and a 1.90M-parameter distilled decoder, totaling approximately 0.387B parameters; (2) Rectified Flow training combined with DMD2 distillation, reducing inference to 4 sampling steps; (3) Chinese semantic alignment trained on 9M curated image-text pairs, enabling direct Chinese prompting without external translation at inference time; and (4) a training and distillation pipeline completed on domestically developed Sugon K100 AI accelerators without relying on NVIDIA GPUs for training or distillation. Despite its compact scale, the 28-step base model of JuZhou 1.0 achieves an overall GenEval score of 0.69, outperforming published baselines including SDXL (2.6B, 0.55), SD3-Medium (2B, 0.62), and IF-XL (4.3B, 0.61). We further validate the full poetry-to-image pipeline on Android and the core CLIP-U-Net-VAE generation branch on iOS. On a smartphone powered by the Snapdragon 8 Elite Gen 5 Mobile Platform, the 4-step U-Net denoising branch runs in approximately 1.6 seconds, while the full Android poetry-to-image pipeline takes 4.5 seconds with on-device prompt refinement on Xiaomi 17 Pro Max. These results position JuZhou 1.0 as a practical approach to mobile text-to-image generation and provide a concrete reference for Chinese-native generation, domestic-compute training, and fully offline on-device deployment after one-time installation.

cs.CV↗

arXiv2026

Bipartite cuts in Ramsey-Tur\'an style

We prove that every $K_5$-free $n$-vertex graph with sublinear independence number can be made bipartite by removing at most $n^2(1/18+o(1))$ edges, where the constant $1/18$ is best possible. The proof method is related to extensions of Tur\'an Theorem in edge-weighted settings, and part of the proof uses flag algebra.

math.CO↗

arXiv2026

Avatar V: Scaling Video-Reference Avatar Video Generation

Generating avatar videos that are not merely visually similar to a target individual but behaviorally recognizable, faithfully reproducing their talking rhythm, gestural tendencies, and expression dynamics, remains an open challenge. Existing methods predominantly condition on single static images, which provide insufficient identity information and cannot capture dynamic motion traits, while standard pixel-level objectives underserve the perceptually critical facial regions that determine avatar fidelity. We present Avatar V, a production-scale framework that addresses these limitations through video-reference-conditioned identity modeling. Rather than compressing identity into fixed-size embeddings, the model conditions directly on the full token sequence of a reference video, learning to reproduce both static identity attributes (facial geometry, skin texture) and dynamic behavioral patterns (talking rhythm, micro-expressions) through attention over the reference context. We introduce Sparse Reference Attention, an asymmetric mechanism achieving linear-complexity conditioning on arbitrarily long references; a motion representation stream enabling closed-loop talking style transfer; and an identity-aware super-resolution refiner inheriting the full reference conditioning. These are supported by a data engine curating 100M+ training clips from 50M raw videos, and a five-stage training pipeline with flow matching pre-training, personality fine-tuning, two-phase distillation (>10x acceleration), and RLHF alignment, deployed across thousands of GPUs. Avatar V generates 1080p videos of unlimited duration, achieving state-of-the-art identity preservation, lip synchronization, and generation quality on our cross-scene benchmark, consistently outperforming leading systems including Seedance 2.0, Kling O3 Pro, Veo 3.1, and OmniHuman 1.5 in both automated metrics and human evaluation.

cs.CV↗

arXiv2026

Generate Your Talking Avatar from Video Reference

Existing talking avatar methods typically adopt an image-to-video pipeline conditioned on a static reference image within the same scene as the target generation. This restricted, single-view perspective lacks sufficient temporal and expression cues, limiting the ability to synthesize high-fidelity talking avatars in customized backgrounds. To this end, we introduce Talking Avatar generation from Video Reference (TAVR), a novel framework that shifts the paradigm by leveraging cross-scene video inputs. To effectively process these extended temporal contexts and bridge cross-scene domain gaps, TAVR integrates a token selection module alongside a comprehensive three-stage training scheme. Specifically, same-scene video pretraining establishes foundational appearance copying, which is subsequently expanded by cross-scene reference fine-tuning for robust cross-scene adaptation. Finally, task-specific reinforcement learning aligns the generated outputs with identity-based rewards to maximize identity similarity. To systematically evaluate cross-scene robustness, we construct a new benchmark comprising 158 carefully curated cross-scene video pairs. Extensive experiments show that TAVR benefits from flexible inference-time video referencing and consistently surpasses existing baselines both quantitatively and qualitatively. This work has been deployed to production. For more related research, please visit \href{https://www.heygen.com/research}{HeyGen Research} and \href{https://www.heygen.com/research/avatar-v-model}{HeyGen Avatar-V}.

cs.CV↗

arXiv2026

TransVLM: A Vision-Language Framework and Benchmark for Detecting Any Shot Transitions

Traditional Shot Boundary Detection (SBD) inherently struggles with complex transitions by formulating the task around isolated cut points, frequently yielding corrupted video shots. We address this fundamental limitation by formalizing the Shot Transition Detection (STD) task. Rather than searching for ambiguous points, STD explicitly detects the continuous temporal segments of transitions. To tackle this, we propose TransVLM, a Vision-Language Model (VLM) framework for STD. Unlike regular VLMs that predominantly rely on spatial semantics and struggle with fine-grained inter-shot dynamics, our method explicitly injects optical flow as a critical motion prior at the input stage. Through a simple yet effective feature-fusion strategy, TransVLM directly processes concatenated color and motion representations, significantly enhancing its temporal awareness without incurring any additional visual token overhead on the language backbone. To overcome the severe class imbalance in public data, we design a scalable data engine to synthesize diverse transition videos for robust training, alongside a comprehensive benchmark for STD. Extensive experiments demonstrate that TransVLM achieves superior overall performance, outperforming traditional heuristic methods, specialized spatiotemporal networks, and top-tier VLMs. For more related research, please visit HeyGen Research (https://www.heygen.com/research) and HeyGen Avatar-V (https://www.heygen.com/research/avatar-v-model). Project page: https://chence17.github.io/TransVLM/

cs.CV↗

arXiv2026

Forbidding Exactly One Hamming Distance

Addressing questions raised in recent papers, we study the $r$-distance graph $H_r(n)$ on the Boolean cube $\{0,1\}^n$, where two vertices are adjacent if their Hamming distance is exactly $r$. For fixed integers $s \ge 2$ and even $r \ge 2$, we determine the asymptotic order of the $s$-independence number $\alpha_s(H_r(n))$, showing that \[ \alpha_s\left(H_r(n)\right)=\Theta\left(\frac{2^n}{n^{r/2}}\right). \] The upper bound is derived via a reduction to extremal problems for sunflower-free set systems, while the lower bound is obtained using algebraic constructions based on BCH codes and constant-weight codes.

math.CO↗

arXiv2026

Mobile-VTON: High-Fidelity On-Device Virtual Try-On

Virtual try-on (VTON) has recently achieved impressive visual fidelity, but most existing systems require uploading personal photos to cloud-based GPUs, raising privacy concerns and limiting on-device deployment. To address this, we present Mobile-VTON, a high-quality, privacy-preserving framework that enables fully offline virtual try-on on commodity mobile devices using only a single user image and a garment image. Mobile-VTON introduces a modular TeacherNet-GarmentNet-TryonNet (TGT) architecture that integrates knowledge distillation, garment-conditioned generation, and garment alignment into a unified pipeline optimized for on-device efficiency. Within this framework, we propose a Feature-Guided Adversarial (FGA) Distillation strategy that combines teacher supervision with adversarial learning to better match real-world image distributions. GarmentNet is trained with a trajectory-consistency loss to preserve garment semantics across diffusion steps, while TryonNet uses latent concatenation and lightweight cross-modal conditioning to enable robust garment-to-person alignment without large-scale pretraining. By combining these components, Mobile-VTON achieves high-fidelity generation with low computational overhead. Experiments on VITON-HD and DressCode at 1024 x 768 show that it matches or outperforms strong server-based baselines while running entirely offline. These results demonstrate that high-quality VTON is not only feasible but also practical on-device, offering a secure solution for real-world applications. Code and project page are available at https://zhenchenwan.github.io/Mobile-VTON/.

cs.CV↗

arXiv2026

Condition Matters in Full-head 3D GANs

Conditioning is crucial for stable training of full-head 3D GANs. Without any conditioning signal, the model suffers from severe mode collapse, making it impractical to training. However, a series of previous full-head 3D GANs conventionally choose the view angle as the conditioning input, which leads to a bias in the learned 3D full-head space along the conditional view direction. This is evident in the significant differences in generation quality and diversity between the conditional view and non-conditional views of the generated 3D heads, resulting in global incoherence across different head regions. In this work, we propose to use view-invariant semantic feature as the conditioning input, thereby decoupling the generative capability of 3D heads from the viewing direction. To construct a view-invariant semantic condition for each training image, we create a novel synthesized head image dataset. We leverage FLUX.1 Kontext to extend existing high-quality frontal face datasets to a wide range of view angles. The image clip feature extracted from the frontal view is then used as a shared semantic condition across all views in the extended images, ensuring semantic alignment while eliminating directional bias. This also allows supervision from different views of the same subject to be consolidated under a shared semantic condition, which accelerates training and enhances the global coherence of the generated 3D heads. Moreover, as GANs often experience slower improvements in diversity once the generator learns a few modes that successfully fool the discriminator, our semantic conditioning encourages the generator to follow the true semantic distribution, thereby promoting continuous learning and diverse generation. Extensive experiments on full-head synthesis and single-view GAN inversion demonstrate that our method achieves significantly higher fidelity, diversity, and generalizability.

cs.CV↗

arXiv2025

Reconstruction of C_4-free graphs from the set of closed neighborhoods and digital convexity

Fomin, Kratochv\'il, Lokshtanov, Mancini, and Telle showed that every $C_{4}$-free graph is reconstructible from the \emph{multiset} of closed neighborhoods. We strengthen their result proving that every $C_{4}$-free graph is reconstructible from the \emph{set} of closed neighborhoods. This extends the work of Lafrance et al.\ by showing that all $C_{4}$-free graphs, and hence all graphs of girth at least five, are reconstructible from their digitally convex sets. A subset $S$ of vertices in a graph $G$ is digitally convex if, for every vertex $v \notin S$, there is a private neighbor of $v$. We establish that reconstruction from digitally convex sets is equivalent to reconstruction from the set of closed neighborhoods.

math.CO↗

arXiv2025

Maximal independent sets in the middle two layers of the Boolean lattice

Let $B(2d-1, d)$ be the subgraph of the hypercube $\mathcal{Q}_{2d-1}$ induced by its two largest layers. Duffus, Frankl and R\"odl proposed the problem of finding the asymptotics for the logarithm of the number of maximal independent sets in $B(2d-1, d)$. Ilinca and Kahn determined the logarithmic asymptotics and reiterated the question of what their order of magnitude is. We show that the number of maximal independent sets in $B(2d-1,d)$ is \[ \left(1+o(1)\right)(2d-1)\exp\left(\frac{(d-1)^2}{2^{2d-1}}\binom{2d-2}{d-1}\right)\cdot 2^{\binom{2d-2}{d-1}}, \] and describe their typical structure. The proof uses a new variation of Sapozhenko's Graph Container Lemma, a new isoperimetric lemma, a theorem of Hujter and Tuza on the number of maximal independent sets in triangle-free graphs and a stability version of their result by Kahn and Park, among other tools.

math.CO↗

arXiv2025

Arbitrary control of the flow of light using pseudomagnetic fields in photonic crystals at telecommunication wavelengths

In photonics, the idea of controlling light in a similar way that magnetic fields control electrons has always been attractive. It can be realized by synthesizing pseudomagnetic fields (PMFs) in photonic crystals (PhCs). Previous works mainly focus on the Landau levels and the robust transport of the chiral states. More versatile control over light using complex nonuniform PMFs such as the flexible splitting and routing of light has been elusive, which hinders their application in practical photonic integrated circuits. Here we propose an universal and systematic methodology to design nonuniform PMFs and arbitrarily control the flow of light in silicon PhCs at telecommunication wavelengths. As proofs of concept, a low-loss S-bend and a highly efficient 50:50 power splitter based on PMFs are experimentally demonstrated. A high-speed data transmission experiment is performed on these devices to prove their applicability in real communication systems. The proposed method offers a new paradigm for the exploration of fundamental physics and the development of novel nanophotonic devices.

physics.optics↗

arXiv2024

CT4D: Consistent Text-to-4D Generation with Animatable Meshes

Text-to-4D generation has recently been demonstrated viable by integrating a 2D image diffusion model with a video diffusion model. However, existing models tend to produce results with inconsistent motions and geometric structures over time. To this end, we present a novel framework, coined CT4D, which directly operates on animatable meshes for generating consistent 4D content from arbitrary user-supplied prompts. The primary challenges of our mesh-based framework involve stably generating a mesh with details that align with the text prompt while directly driving it and maintaining surface continuity. Our CT4D framework incorporates a unique Generate-Refine-Animate (GRA) algorithm to enhance the creation of text-aligned meshes. To improve surface continuity, we divide a mesh into several smaller regions and implement a uniform driving function within each area. Additionally, we constrain the animating stage with a rigidity regulation to ensure cross-region continuity. Our experimental results, both qualitative and quantitative, demonstrate that our CT4D framework surpasses existing text-to-4D techniques in maintaining interframe consistency and preserving global geometry. Furthermore, we showcase that this enhanced representation inherently possesses the capability for combinational 4D generation and texture editing.

cs.GR↗

arXiv2024

On the maximum $F$-free induced subgraphs in $K_t$-free graphs

For graphs $F$ and $H$, let $f_{F,H}(n)$ be the minimum possible size of a maximum $F$-free induced subgraph in an $n$-vertex $H$-free graph. This notion generalizes the Ramsey function and the Erd\H{o}s--Rogers function. Establishing a container lemma for the $F$-free subgraphs, we give a general upper bound on $f_{F,H}(n)$, assuming the existence of certain locally dense $H$-free graphs. In particular, we prove that for every graph $F$ with $\mathrm{ex}(m,F) = O(m^{1+\alpha})$, where $\alpha \in [0,1/2)$, we have \[ f_{F, K_3}(n) = O\left(n^{\frac{1}{2-\alpha}}\left(\log n\right)^{\frac{3}{2- \alpha}}\right) \quad \textrm{and} \quad f_{F, K_4}(n) = O\left(n^{\frac{1}{3-2\alpha}}\left(\log n\right)^{\frac{6}{3-2\alpha}}\right). \] For the cases where $F$ is a complete multipartite graph, letting $s = \sum_{i=1}^r s_i$, we prove that \[ f_{K_{s_1,\ldots,s_r}, K_{r+2}}(n) = O \left( n^{\frac{2s -3}{4s -5}} (\log n)^{3} \right). \] We also make an observation which improves the bounds of $\mathrm{ex}(G(n,p),C_4)$ by a polylogarithmic factor.

math.CO↗

arXiv2024

SphereHead: Stable 3D Full-head Synthesis with Spherical Tri-plane Representation

While recent advances in 3D-aware Generative Adversarial Networks (GANs) have aided the development of near-frontal view human face synthesis, the challenge of comprehensively synthesizing a full 3D head viewable from all angles still persists. Although PanoHead proves the possibilities of using a large-scale dataset with images of both frontal and back views for full-head synthesis, it often causes artifacts for back views. Based on our in-depth analysis, we found the reasons are mainly twofold. First, from network architecture perspective, we found each plane in the utilized tri-plane/tri-grid representation space tends to confuse the features from both sides, causing "mirroring" artifacts (e.g., the glasses appear in the back). Second, from data supervision aspect, we found that existing discriminator training in 3D GANs mainly focuses on the quality of the rendered image itself, and does not care much about its plausibility with the perspective from which it was rendered. This makes it possible to generate "face" in non-frontal views, due to its easiness to fool the discriminator. In response, we propose SphereHead, a novel tri-plane representation in the spherical coordinate system that fits the human head's geometric characteristics and efficiently mitigates many of the generated artifacts. We further introduce a view-image consistency loss for the discriminator to emphasize the correspondence of the camera parameters and the images. The combination of these efforts results in visually superior outcomes with significantly fewer artifacts. Our code and dataset are publicly available at https://lhyfst.github.io/spherehead.

cs.CV↗

arXiv2023

On the Constructor-Blocker Game

In the Constructor-Blocker game, two players, Constructor and Blocker, alternatively claim unclaimed edges of the complete graph $K_n$. For given graphs $F$ and $H$, Constructor can only claim edges that leave her graph $F$-free, while Blocker has no restrictions. Constructor's goal is to build as many copies of $H$ as she can, while Blocker attempts to stop this. The game ends once there are no more edges that Constructor can claim. The score $g(n,H,F)$ of the game is the number of copies of $H$ in Constructor's graph at the end of the game, when both players play optimally and Constructor plays first. In this paper, we extend results of Patk\'os, Stojakovi\'c and Vizer on $g(n, H, F)$ to many pairs of $H$ and $F$: We determine $g(n, H, F)$ when $H=K_r$ and $\chi(F)>r$, also when both $H$ and $F$ are odd cycles, using Szemer\'edi's Regularity Lemma. We also obtain bounds of $g(n, H, F)$ when $H=K_3$ and $F=K_{2,2}$.

math.CO↗

arXiv2023

DenseMP: Unsupervised Dense Pre-training for Few-shot Medical Image Segmentation

Few-shot medical image semantic segmentation is of paramount importance in the domain of medical image analysis. However, existing methodologies grapple with the challenge of data scarcity during the training phase, leading to over-fitting. To mitigate this issue, we introduce a novel Unsupervised Dense Few-shot Medical Image Segmentation Model Training Pipeline (DenseMP) that capitalizes on unsupervised dense pre-training. DenseMP is composed of two distinct stages: (1) segmentation-aware dense contrastive pre-training, and (2) few-shot-aware superpixel guided dense pre-training. These stages collaboratively yield a pre-trained initial model specifically designed for few-shot medical image segmentation, which can subsequently be fine-tuned on the target dataset. Our proposed pipeline significantly enhances the performance of the widely recognized few-shot segmentation model, PA-Net, achieving state-of-the-art results on the Abd-CT and Abd-MRI datasets. Code will be released after acceptance.

cs.CV↗

arXiv2023

SCoDA: Domain Adaptive Shape Completion for Real Scans

3D shape completion from point clouds is a challenging task, especially from scans of real-world objects. Considering the paucity of 3D shape ground truths for real scans, existing works mainly focus on benchmarking this task on synthetic data, e.g. 3D computer-aided design models. However, the domain gap between synthetic and real data limits the generalizability of these methods. Thus, we propose a new task, SCoDA, for the domain adaptation of real scan shape completion from synthetic data. A new dataset, ScanSalon, is contributed with a bunch of elaborate 3D models created by skillful artists according to scans. To address this new task, we propose a novel cross-domain feature fusion method for knowledge transfer and a novel volume-consistent self-training framework for robust learning from real data. Extensive experiments prove our method is effective to bring an improvement of 6%~7% mIoU.

cs.CV↗

arXiv2022

Ramsey-Tur\'an Problems with small independence numbers

Given a graph $H$ and a function $f(n)$, the Ramsey-Tur\'an number $RT(n,H,f(n))$ is the maximum number of edges in an $n$-vertex $H$-free graph with independence number at most $f(n)$. For $H$ being a small clique, many results about $RT(n,H,f(n))$ are known and we focus our attention on $H=K_s$ for $s\leq 13$. By applying Szemer\'edi's Regularity Lemma, the dependent random choice method and some weighted Tur\'an-type results, we prove that these cliques have the so-called phase transitions when $f(n)$ is around the inverse function of the off-diagonal Ramsey number of $K_r$ versus a large clique $K_n$ for some $r\leq s$.

math.CO↗