arXiv subjects

Zhizhong Huang

Publications and source records attributed to Zhizhong Huang.

At least 19 recordsLinked to original sources

arXiv2026

DiverseDiT++: Quantifying, Analyzing, and Promoting Representation Diversity in Diffusion Transformers

Recent advances in Diffusion Transformers (DiTs) have enabled remarkable progress in visual synthesis, benefiting from their superior scalability. To facilitate DiTs' capability of capturing meaningful internal representations, recent works such as REPA incorporate external pretrained encoders for representation alignment. However, the underlying mechanisms governing representation learning within DiTs remain poorly understood in the community. To this end, this paper first presents a systematic analysis of the representation dynamics of DiTs via quantifying the diversity of block-wise representations. Specifically, we introduce a novel metric, termed the Weighted Diversity Score (WDS), to measure the representational discrepancies across different blocks. Through extensive investigations on the evolution and influence of internal representations under various settings, we reveal that representation diversity across blocks is a critical factor for effective representation learning in DiTs. More importantly, WDS exhibits a strong correlation with synthesis quality across diverse settings, model scales, and training stages (Pearson's $r=-0.869$ with $\log(\text{FID})$), suggesting its potential as an indicator to reflect model performance and a principled guide for model optimization. Based on this key finding, we propose DiverseDiT++, a novel framework that explicitly promotes diverse representation learning. Concretely, our method incorporates long residual connections to diversify input representations across blocks and a representation diversity loss to encourage blocks to learn distinct features. Extensive experiments on ImageNet $256\times256$ and $512\times512$ demonstrate that our DiverseDiT++ yields consistent performance gains and convergence acceleration when applied to different backbones with various sizes,...

cs.CV↗

arXiv2026

Equidistribution of rational points and the geometric sieve for toric varieties

We prove the Manin--Peyre principle about the equidistribution of rational points on smooth proper split toric varieties. That is, rational points of bounded anticanonical height outside of the boundary divisors are equidistributed with respect to the Tamagawa measure in the adelic space. Based on a refinement of the universal torsor method due to Salberger, we provide asymptotic formulas with effective error terms for the number of rational points lying in an arbitrary adelic neighbourhood, and we also prove an effective version of the Ekedahl-type geometric sieve. Several applications to sieving rational points are derived.

math.NT↗

arXiv2026

Integral Diophantine approximation on varieties

We study the local behavior of integral points on log pairs near a fixed rational point in the boundary by means of an integral approximation constant. In light of Siegel's theorem about integral points on curves and McKinnon's conjecture on rational approximation constants, we conjecture that integral points that are close to the fixed point in archimedean topology should lie on certain rational curves with at most two points at infinity on weakly log Fano varieties. We verify this conjecture for a number of examples.

math.NT↗

arXiv2025

Quantitative strong approximation for ternary quadratic forms III

We prove asymptotic formulas for counting (primitive) integral points with local conditions on the (punctured) affine cone defined by a non-singular integral ternary quadratic form, and we relate our results to the Brauer--Manin obstruction. Our approach is based on the $δ$-variant of the Hardy--Littlewood circle method developed by Heath-Brown.

math.NT↗

arXiv2025

Local distribution of rational points in flag varieties

Given a flag variety $X$ defined over $\mathbb{Q}$ and a point $x$ in $X(\mathbb{R})$, we study approximations to $x$ by points $v$ in $X(\mathbb{Q})$, and show that, with an appropriate rescaling, those approximations equidistribute when $x$ is chosen randomly according to a Lebesgue measure on $X(\mathbb{R})$, or when $x$ is defined over $\mathbb{Q}$ and satisfies some non-degeneracy condition.

math.NT↗

arXiv2025

Effective equidistribution, arithmetic purity of strong approximation, and geometric sieve for affine quadrics

Let $k$ be a number field. Let $q(x_1,\cdots,x_n)$ be a non-degenerate integral quadratic form in $n\geq 3$ variables with coefficients in $k$ and $m\in k^\times$. Let $X$ be the affine quadric defined by $q=m$ in $\mathbb{A}^n_k$. Based on results on effective equidistribution of $S$-integral points in symmetric spaces, we establish the following: (i) The arithmetic purity of strong approximation off any single place of $k$ for $X$; (ii) The geometric sieve for $p_0$-integral points on $X$ when $k=\mathbb{Q}$ and $p_0$ is a prime number.

math.NT↗

arXiv2025

Generalizable Object Re-Identification via Visual In-Context Prompting

Current object re-identification (ReID) methods train domain-specific models (e.g., for persons or vehicles), which lack generalization and demand costly labeled data for new categories. While self-supervised learning reduces annotation needs by learning instance-wise invariance, it struggles to capture \textit{identity-sensitive} features critical for ReID. This paper proposes Visual In-Context Prompting~(VICP), a novel framework where models trained on seen categories can directly generalize to unseen novel categories using only \textit{in-context examples} as prompts, without requiring parameter adaptation. VICP synergizes LLMs and vision foundation models~(VFM): LLMs infer semantic identity rules from few-shot positive/negative pairs through task-specific prompting, which then guides a VFM (\eg, DINO) to extract ID-discriminative features via \textit{dynamic visual prompts}. By aligning LLM-derived semantic concepts with the VFM's pre-trained prior, VICP enables generalization to novel categories, eliminating the need for dataset-specific retraining. To support evaluation, we introduce ShopID10K, a dataset of 10K object instances from e-commerce platforms, featuring multi-view images and cross-domain testing. Experiments on ShopID10K and diverse ReID benchmarks demonstrate that VICP outperforms baselines by a clear margin on unseen categories. Code is available at https://github.com/Hzzone/VICP.

cs.CV↗

arXiv2025

Rational curves and the Hilbert Property on Jacobian Kummer varieties

A conjecture by Corvaja and Zannier predicts that smooth, projective, simply connected varieties over a number field with Zariski dense set of rational points have the Hilbert Property; this was proved by Demeio for Kummer surfaces which are associated to products of two elliptic curves. In this article, over a finitely generated field of characteristic zero, we establish the Hilbert Property for all Kummer surfaces associated to the Jacobian of a genus $2$ curve. In general we prove that all Jacobian Kummer varieties associated to a hyperelliptic curve of genus $\geq 2$ of odd degree also have the Hilbert Property.

math.NT↗

arXiv2025

Rational approximations on toric varieties

Using the universal torsor method due to Salberger, we study the approximation of a general fixed point by rational points on split toric varieties. We prove that under certain geometric hypothesis the best approximations (in the sense of McKinnon-Roth's work) can be achieved on rational curves passing through the fixed point of minimal degree, confirming a conjecture of McKinnon. These curves correspond, according to Batyrev's terminology, to the centred primitive collections of the fan.

math.NT↗

arXiv2025

Quantitative strong approximation for quaternary quadratic forms

The purpose of this article is twofold. On the one hand, we prove asymptotic formulas for the quantitative distribution of rational points on any smooth non-split projective quadratic surface. We obtain the optimal error term for the real place. On the other hand, we also study the growth of integral points on the three-dimensional punctured affine cone, as a quantitative version of strong approximation with Brauer--Manin obstruction for this quasi-affine variety.

math.NT↗

arXiv2024

Quantitative strong approximation for ternary quadratic forms II

Let $F$ be a non-degenerate integral ternary quadratic form and let $m_0\in\mathbb{Z}_{\neq 0}$. We study growth of rational points on the affine quadric $(F=m_0)$ and show that they are equidistributed in the adelic space off a finite place. This is closely related to Linnik's problem. Our approach is based on the $δ$-variant of the Hardy--Littlewood circle method developed by Heath-Brown.

math.NT↗

arXiv2024

Quantitative strong approximation for ternary quadratic forms I

We derive asymptotic formulas with a secondary term for the (smoothly weighted) count of number of integer solutions of height $\leqslant B$ with local conditions to the equation $F(x_1,x_2,x_3)=m$, where $F$ is a non-degenerate indefinite ternary integral quadratic form, and $m$ is a non-zero integer satisfying $-mΔ_F=\square$ which can grow like $O(B^{2-θ})$ for some fixed $θ>0$. Our approach is based on the $δ$-variant of the Hardy--Littlewood circle method developed by Heath-Brown.

math.NT↗

arXiv2024

DreamVideo-2: Zero-Shot Subject-Driven Video Customization with Precise Motion Control

Recent advances in customized video generation have enabled users to create videos tailored to both specific subjects and motion trajectories. However, existing methods often require complicated test-time fine-tuning and struggle with balancing subject learning and motion control, limiting their real-world applications. In this paper, we present DreamVideo-2, a zero-shot video customization framework capable of generating videos with a specific subject and motion trajectory, guided by a single image and a bounding box sequence, respectively, and without the need for test-time fine-tuning. Specifically, we introduce reference attention, which leverages the model's inherent capabilities for subject learning, and devise a mask-guided motion module to achieve precise motion control by fully utilizing the robust motion signal of box masks derived from bounding boxes. While these two components achieve their intended functions, we empirically observe that motion control tends to dominate over subject learning. To address this, we propose two key designs: 1) the masked reference attention, which integrates a blended latent mask modeling scheme into reference attention to enhance subject representations at the desired positions, and 2) a reweighted diffusion loss, which differentiates the contributions of regions inside and outside the bounding boxes to ensure a balance between subject and motion control. Extensive experimental results on a newly curated dataset demonstrate that DreamVideo-2 outperforms state-of-the-art methods in both subject customization and motion control. The dataset, code, and models will be made publicly available.

cs.CV↗

arXiv2024

FLDM-VTON: Faithful Latent Diffusion Model for Virtual Try-on

Despite their impressive generative performance, latent diffusion model-based virtual try-on (VTON) methods lack faithfulness to crucial details of the clothes, such as style, pattern, and text. To alleviate these issues caused by the diffusion stochastic nature and latent supervision, we propose a novel Faithful Latent Diffusion Model for VTON, termed FLDM-VTON. FLDM-VTON improves the conventional latent diffusion process in three major aspects. First, we propose incorporating warped clothes as both the starting point and local condition, supplying the model with faithful clothes priors. Second, we introduce a novel clothes flattening network to constrain generated try-on images, providing clothes-consistent faithful supervision. Third, we devise a clothes-posterior sampling for faithful inference, further enhancing the model performance over conventional clothes-agnostic Gaussian sampling. Extensive experimental results on the benchmark VITON-HD and Dress Code datasets demonstrate that our FLDM-VTON outperforms state-of-the-art baselines and is able to generate photo-realistic try-on images with faithful clothing details.

cs.CV↗

arXiv2024

Optimal quantitative weak approximation for projective quadrics

We derive asymptotic formulas for the number of rational points on a smooth projective quadratic hypersurface of dimension at least three inside of a shrinking adelic open neighbourhood. This is a quantitative version of weak approximation for quadrics and allows us to deduce the best growth rate of the size of such an adelic neighbourhood for which equidistribution is preserved.

math.NT↗

arXiv2024

Point, Segment and Count: A Generalized Framework for Object Counting

Class-agnostic object counting aims to count all objects in an image with respect to example boxes or class names, \emph{a.k.a} few-shot and zero-shot counting. In this paper, we propose a generalized framework for both few-shot and zero-shot object counting based on detection. Our framework combines the superior advantages of two foundation models without compromising their zero-shot capability: (\textbf{i}) SAM to segment all possible objects as mask proposals, and (\textbf{ii}) CLIP to classify proposals to obtain accurate object counts. However, this strategy meets the obstacles of efficiency overhead and the small crowded objects that cannot be localized and distinguished. To address these issues, our framework, termed PseCo, follows three steps: point, segment, and count. Specifically, we first propose a class-agnostic object localization to provide accurate but least point prompts for SAM, which consequently not only reduces computation costs but also avoids missing small objects. Furthermore, we propose a generalized object classification that leverages CLIP image/text embeddings as the classifier, following a hierarchical knowledge distillation to obtain discriminative classifications among hierarchical mask proposals. Extensive experimental results on FSC-147, COCO, and LVIS demonstrate that PseCo achieves state-of-the-art performance in both few-shot/zero-shot object counting/detection. Code: https://github.com/Hzzone/PseCo

cs.CV↗

arXiv2023

Deep Rank-Consistent Pyramid Model for Enhanced Crowd Counting

Most conventional crowd counting methods utilize a fully-supervised learning framework to establish a mapping between scene images and crowd density maps. They usually rely on a large quantity of costly and time-intensive pixel-level annotations for training supervision. One way to mitigate the intensive labeling effort and improve counting accuracy is to leverage large amounts of unlabeled images. This is attributed to the inherent self-structural information and rank consistency within a single image, offering additional qualitative relation supervision during training. Contrary to earlier methods that utilized the rank relations at the original image level, we explore such rank-consistency relation within the latent feature spaces. This approach enables the incorporation of numerous pyramid partial orders, strengthening the model representation capability. A notable advantage is that it can also increase the utilization ratio of unlabeled samples. Specifically, we propose a Deep Rank-consistEnt pyrAmid Model (DREAM), which makes full use of rank consistency across coarse-to-fine pyramid features in latent spaces for enhanced crowd counting with massive unlabeled images. In addition, we have collected a new unlabeled crowd counting dataset, FUDAN-UCC, comprising 4,000 images for training purposes. Extensive experiments on four benchmark datasets, namely UCF-QNRF, ShanghaiTech PartA and PartB, and UCF-CC-50, show the effectiveness of our method compared with previous semi-supervised methods. The codes are available at https://github.com/bridgeqiqi/DREAM.

cs.CV↗

arXiv2023

Semantic Latent Decomposition with Normalizing Flows for Face Editing

Navigating in the latent space of StyleGAN has shown effectiveness for face editing. However, the resulting methods usually encounter challenges in complicated navigation due to the entanglement among different attributes in the latent space. To address this issue, this paper proposes a novel framework, termed SDFlow, with a semantic decomposition in original latent space using continuous conditional normalizing flows. Specifically, SDFlow decomposes the original latent code into different irrelevant variables by jointly optimizing two components: (i) a semantic encoder to estimate semantic variables from input faces and (ii) a flow-based transformation module to map the latent code into a semantic-irrelevant variable in Gaussian distribution, conditioned on the learned semantic variables. To eliminate the entanglement between variables, we employ a disentangled learning strategy under a mutual information framework, thereby providing precise manipulation controls. Experimental results demonstrate that SDFlow outperforms existing state-of-the-art face editing methods both qualitatively and quantitatively. The source code is made available at https://github.com/phil329/SDFlow.

cs.CV↗