arXiv subjects

Jiahong Wu

Publications and source records attributed to Jiahong Wu.

At least 55 records · Page 3Linked to original sources

arXiv2026

Artifact-Aware Evaluation for High-Quality Video Generation

With the rapid advancement of video generation techniques, evaluating and auditing generated videos has become increasingly crucial. Existing approaches typically offer coarse video quality scores, lacking detailed localization and categorization of specific artifacts. In this work, we introduce a comprehensive evaluation protocol focusing on three key aspects affecting human perception: Appearance, Motion, and Camera. We define these axes through a taxonomy of 10 prevalent artifact categories reflecting common generative failures observed in video generation. To enable robust artifact detection and categorization, we introduce GenVID, a large-scale dataset of 80k videos generated by various state-of-the-art video generation models, each carefully annotated for the defined artifact categories. Leveraging GenVID, we develop DVAR, a Dense Video Artifact Recognition framework for fine-grained identification and classification of generative artifacts. Extensive experiments show that our approach significantly improves artifact detection accuracy and enables effective filtering of low-quality content.

cs.CV↗

arXiv2026

Latent Temporal Discrepancy as Motion Prior: A Loss-Weighting Strategy for Dynamic Fidelity in T2V

Video generation models have achieved notable progress in static scenarios, yet their performance in motion video generation remains limited, with quality degrading under drastic dynamic changes. This is due to noise disrupting temporal coherence and increasing the difficulty of learning dynamic regions. {Unfortunately, existing diffusion models rely on static loss for all scenarios, constraining their ability to capture complex dynamics.} To address this issue, we introduce Latent Temporal Discrepancy (LTD) as a motion prior to guide loss weighting. LTD measures frame-to-frame variation in the latent space, assigning larger penalties to regions with higher discrepancy while maintaining regular optimization for stable regions. This motion-aware strategy stabilizes training and enables the model to better reconstruct high-frequency dynamics. Extensive experiments on the general benchmark VBench and the motion-focused VMBench show consistent gains, with our method outperforming strong baselines by 3.31% on VBench and 3.58% on VMBench, achieving significant improvements in motion quality.

cs.CV↗

arXiv2026

Ranking-aware Reinforcement Learning for Ordinal Ranking

Ordinal regression and ranking are challenging due to inherent ordinal dependencies that conventional methods struggle to model. We propose Ranking-Aware Reinforcement Learning (RARL), a novel RL framework that explicitly learns these relationships. At its core, RARL features a unified objective that synergistically integrates regression and Learning-to-Rank (L2R), enabling mutual improvement between the two tasks. This is driven by a ranking-aware verifiable reward that jointly assesses regression precision and ranking accuracy, facilitating direct model updates via policy optimization. To further enhance training, we introduce Response Mutation Operations (RMO), which inject controlled noise to improve exploration and prevent stagnation at saddle points. The effectiveness of RARL is validated through extensive experiments on three distinct benchmarks.

cs.LG↗

arXiv2025

Taming Hallucinations: Boosting MLLMs' Video Understanding via Counterfactual Video Generation

Multimodal Large Language Models (MLLMs) have made remarkable progress in video understanding. However, they suffer from a critical vulnerability: an over-reliance on language priors, which can lead to visual ungrounded hallucinations, especially when processing counterfactual videos that defy common sense. This limitation, stemming from the intrinsic data imbalance between text and video, is challenging to address due to the substantial cost of collecting and annotating counterfactual data. To address this, we introduce DualityForge, a novel counterfactual data synthesis framework that employs controllable, diffusion-based video editing to transform real-world videos into counterfactual scenarios. By embedding structured contextual information into the video editing and QA generation processes, the framework automatically produces high-quality QA pairs together with original-edited video pairs for contrastive training. Based on this, we build DualityVidQA, a large-scale video dataset designed to reduce MLLM hallucinations. In addition, to fully exploit the contrastive nature of our paired data, we propose Duality-Normalized Advantage Training (DNA-Train), a two-stage SFT-RL training regime where the RL phase applies pair-wise $\ell_1$ advantage normalization, thereby enabling a more stable and efficient policy optimization. Experiments on DualityVidQA-Test demonstrate that our method substantially reduces model hallucinations on counterfactual videos, yielding a relative improvement of 24.0% over the Qwen2.5-VL-7B baseline. Moreover, our approach achieves significant gains across both hallucination and general-purpose benchmarks, indicating strong generalization capability. We will open-source our dataset and code.

cs.CV↗

arXiv2025

Omni-Effects: Unified and Spatially-Controllable Visual Effects Generation

Visual effects (VFX) are essential visual enhancements fundamental to modern cinematic production. Although video generation models offer cost-efficient solutions for VFX production, current methods are constrained by per-effect LoRA training, which limits generation to single effects. This fundamental limitation impedes applications that require spatially controllable composite effects, i.e., the concurrent generation of multiple effects at designated locations. However, integrating diverse effects into a unified framework faces major challenges: interference from effect variations and spatial uncontrollability during multi-VFX joint training. To tackle these challenges, we propose Omni-Effects, a first unified framework capable of generating prompt-guided effects and spatially controllable composite effects. The core of our framework comprises two key innovations: (1) LoRA-based Mixture of Experts (LoRA-MoE), which employs a group of expert LoRAs, integrating diverse effects within a unified model while effectively mitigating cross-task interference. (2) Spatial-Aware Prompt (SAP) incorporates spatial mask information into the text token, enabling precise spatial control. Furthermore, we introduce an Independent-Information Flow (IIF) module integrated within the SAP, isolating the control signals corresponding to individual effects to prevent any unwanted blending. To facilitate this research, we construct a comprehensive VFX dataset Omni-VFX via a novel data collection pipeline combining image editing and First-Last Frame-to-Video (FLF2V) synthesis, and introduce a dedicated VFX evaluation framework for validating model performance. Extensive experiments demonstrate that Omni-Effects achieves precise spatial control and diverse effect generation, enabling users to specify both the category and location of desired effects.

cs.CV↗

arXiv2025

Stability threshold of the 2D Boussinesq system near Couette flow in an infinite channel

In this paper, we study the stability threshold of the two-dimensional Boussinesq equations around the Couette flow in an infinite channel $\mathbb{R} \times [-1, 1]$ under no-slip boundary conditions. We prove that the Couette flow is asymptotically stable under initial perturbations satisfying $\| \mathbf{v}^{\mathrm{in}} -(y,0)\|_{H^2} \le \varepsilon_0 ν^{\frac12}$, and $\| ρ^{\mathrm{in}}-1 \|_{H^1} + \big\| |\partial_x|^{\frac13} ρ^{\mathrm{in}} \big\|_{H^1} \le \varepsilon_1 ν^{\frac56}$. Compared with the work of Masmoudi, Zhai, and Zhao [J. Funct. Anal., 284 (2023), 109736], where the asymptotic stability of the 2D Navier-Stokes-Boussinesq system around Couette flow in a finite channel $\mathbb{T} \times [-1, 1]$ was established, our result improves the stability threshold for the temperature from $ν^{\frac{11}{12}}$ to $ν^{\frac56}$.

math.AP↗

arXiv2025

SVBench: A Benchmark with Temporal Multi-Turn Dialogues for Streaming Video Understanding

Despite the significant advancements of Large Vision-Language Models (LVLMs) on established benchmarks, there remains a notable gap in suitable evaluation regarding their applicability in the emerging domain of long-context streaming video understanding. Current benchmarks for video understanding typically emphasize isolated single-instance text inputs and fail to evaluate the capacity to sustain temporal reasoning throughout the entire duration of video streams. To address these limitations, we introduce SVBench, a pioneering benchmark with temporal multi-turn question-answering chains specifically designed to thoroughly assess the capabilities of streaming video understanding of current LVLMs. We design a semi-automated annotation pipeline to obtain 49,979 Question-Answer (QA) pairs of 1,353 streaming videos, which includes generating QA chains that represent a series of consecutive multi-turn dialogues over video segments and constructing temporal linkages between successive QA chains. Our experimental results, obtained from 14 models in dialogue and streaming evaluations, reveal that while the closed-source GPT-4o outperforms others, most open-source LVLMs struggle with long-context streaming video understanding. We also construct a StreamingChat model, which significantly outperforms open-source LVLMs on our SVBench and achieves comparable performance on diverse vision-language benchmarks. We expect SVBench to advance the research of streaming video understanding by providing a comprehensive and in-depth analysis of current LVLMs. Our benchmark and model can be accessed at https://github.com/sotayang/SVBench.

cs.CV↗

arXiv2025

ImagerySearch: Adaptive Test-Time Search for Video Generation Beyond Semantic Dependency Constraints

Video generation models have achieved remarkable progress, particularly excelling in realistic scenarios; however, their performance degrades notably in imaginative scenarios. These prompts often involve rarely co-occurring concepts with long-distance semantic relationships, falling outside training distributions. Existing methods typically apply test-time scaling for improving video quality, but their fixed search spaces and static reward designs limit adaptability to imaginative scenarios. To fill this gap, we propose ImagerySearch, a prompt-guided adaptive test-time search strategy that dynamically adjusts both the inference search space and reward function according to semantic relationships in the prompt. This enables more coherent and visually plausible videos in challenging imaginative settings. To evaluate progress in this direction, we introduce LDT-Bench, the first dedicated benchmark for long-distance semantic prompts, consisting of 2,839 diverse concept pairs and an automated protocol for assessing creative generation capabilities. Extensive experiments show that ImagerySearch consistently outperforms strong video generation baselines and existing test-time scaling approaches on LDT-Bench, and achieves competitive improvements on VBench, demonstrating its effectiveness across diverse prompt types. We will release LDT-Bench and code to facilitate future research on imaginative video generation.

cs.CV↗

arXiv2025

Iterative projection method for unsteady Navier-Stokes equations with high Reynolds numbers

A new iterative projection method is proposed to solve the unsteady Navier-Stokes equations with high Reynolds numbers. The convectional projection method attempts to project the intermediate velocity to the divergence free space only once per time step. However, such a velocity is not genuinely divergence free in general practice, which can yield large errors when the Reynolds number is high. The new method has several important features: the BDF2 time discretization, the skew-symmetric convection in a semi-implicit form, two modulating parameters, and the iterative projections in each time step. A major difficulty in the proof of iteration convergence is the nonlinear convection. We solve this problem by first analyzing the non-convective scheme with a focus on the spectral properties of the iterative matrix, and then employing a delicate perturbation analysis for the convective scheme. The work achieves the weakly divergence free velocity (strongly divergence free for divergence free finite element spaces), and the rigorous stability and error analysis when the iterations converge. The three dimensional numerical tests confirm that this new method can effectively treat high Reynolds numbers with only a few iterations per time step, where the convectional projection method and the iterative projection method with the explicit convection would fail.

math.NA↗

arXiv2025

On the uniqueness of strong solution to the nonhomogeneous incompressible Navier-Stokes-Cahn-Hilliard system

This paper is mainly concerned with an initial-boundary value problem of the nonhomogeneous incompressible Navier-Stokes-Cahn-Hilliard system with the Landau potential in a two and three dimensions. The existence of strong solutions with bounded and strictly positive density for this system was constructed by Giorgini and Temam \cite{GT}. However, whether uniqueness holds has remained an open question. The present work solves this question and we prove the uniqueness of strong solution. Our method mainly relies on some extra time weighted estimates and the Lagrangian approach.

math.AP↗

arXiv2025

Improved stability threshold for 2D Navier-Stokes Couette flow in an infinite channel

We study the nonlinear stability of the two-dimensional Navier-Stokes equations around the Couette shear flow in the channel domain $\mathbb{R}\times[-1,1]$ subject to Navier slip boundary conditions. We establish a quantitative stability threshold for perturbations of the initial vorticity $ω_{in}$, showing that stability holds for perturbations of order $ν^{1/2}$ measured in an anisotropic Sobolev space. This sharpens the recent work of Arbon and Bedrossian [Comm. Math. Phys., 406 (2025), Paper No. 129] who proved stability under the threshold $ν^{1/2}(1+\ln(1/ν))^{-1/2}$. Our result removes the logarithmic loss and identifies the natural scaling $ν^{1/2}$ as the critical size of perturbations for nonlinear stability in this setting.

math.AP↗

arXiv2025

Global uniform regularity for the 3D incompressible MHD equations with slip boundary condition near a background magnetic field

This paper resolves the global regularity problem for the three-dimensional incompressible magnetohydrodynamics (MHD) equations in the upper half-space with slip boundary conditions, in the presence of a background magnetic field. Motivated by geophysical applications, we consider an anisotropic MHD system with weak dissipation in the $x_2$ and $x_3$ directions and small vertical magnetic diffusion. By exploiting the stabilizing effect induced by the background magnetic field and constructing a hierarchy of four energy functionals, we establish global-in-time uniform bounds that are independent of the viscosity in the $x_2$ and $x_3$ directions and the vertical resistivity. A key innovation in our analysis is the development of a two-tier energy method, which couples the boundedness of conormal derivatives with the decay of tangential derivatives. These global conormal regularity estimates, together with sharp decay rates, enable us to rigorously justify the vanishing dissipation limit and derive explicit long-time convergence rates to the MHD system with vanishing dissipation in the $x_2$ and $x_3$ directions and no vertical magnetic diffusion. In the absence of a magnetic field, the global-in-time vanishing viscosity limit for the 3D incompressible Navier-Stokes equations with anisotropic dissipation remains a challenging open problem. This work reveals the mechanism by which the magnetic field enhances dissipation and stabilizes the fluid dynamics in the vanishing viscosity limit.

math.AP↗

arXiv2025

Magnetic Stabilization of Compressible Flows: Global Existence in 3D Inviscid Non-Isentropic MHD Equations

Solutions to the compressible Euler equations in all dimensions have been shown to develop finite-time singularities from smooth initial data such as shocks and cusps. There is an extraordinary list of results on this subject. When the inviscid compressible flow is coupled with the magnetic field in the 3D inviscid non-isentropic compressible magnetohydrodynamic (MHD) equations in $\mathbb{T}^3$, this paper rules out finite-time blowup and establishes the global existence of smooth and stable solutions near a suitable background magnetic field. This result rigorously confirms the stabilizing phenomenon observed in physical experiments involving electrically conducting fluids.

math.AP↗

arXiv2025

Incompressible limit for the 3D compressible FENE dumbbell model

In this work, we study the global-in-time incompressible limit of the compressible FENE dumbbell model on the three-dimensional torus T^3, where the incompressible limit is driven by large volume viscosity. To establish this limit, we develop time-weighted a priori estimates that yield decay rates for strong solutions. A key challenge arises from the fact that increasing the volume viscosity suppresses the decay of high-frequency components, thereby weakening the dissipation of the density and complicating the derivation of uniform-in-time decay estimates. To overcome this difficulty, we introduce a novel momentum-based estimate and show that the incompressible component of the momentum decays faster in time than the velocity itself. Exploiting this enhanced decay, we successfully close the a priori estimates and establish a time-decreasing convergence rate toward the incompressible limit.

math.AP↗

arXiv2025

Global uniform regularity for the 3D incompressible MHD equations with slip boundary condition near an equilibrium

This paper solves the global conormal regularity problem for the three-dimensional incompressible MHD equations with slip boundary condition near a background magnetic field. Motivated by applications in geophysics, the MHD system considered here is anisotropic with small vertical dissipation and small horizontal magnetic diffusion. By exploiting the enhanced dissipation due to the background magnetic field and introducing three layers of energy functionals, we are able to establish global-in-time uniform bounds that are independent of vertical viscosity and horizontal resistivity. These global conormal regularity estimates allow us to pass to the limit and obtain the convergence to the MHD system with no vertical dissipation and horizontal magnetic diffusion. In the special case of the 3D incompressible Navier-Stokes, explicit long-time rates are also extracted in the zero vertical viscosity limit.

math.AP↗

arXiv2025

Global solutions to 3D compressible MHD equations with partial magnetic diffusion

The global existence of strong solutions to the compressible viscous magnetohydrodynamic (MHD) equations in $\mathbb{R}^3$ remains a significant open problem. When there is no magnetic diffusion, even small data global well-posedness is unknown. This study investigates the Cauchy problem in $\mathbb{R}^3$ for the compressible viscous MHD equations with horizontal magnetic diffusion. Using various anisotropic Sobolev inequalities and sharp estimates, we establish the existence of global solutions under small initial data within the Sobolev space framework.

math.AP↗

arXiv2025

VMBench: A Benchmark for Perception-Aligned Video Motion Generation

Video generation has advanced rapidly, improving evaluation methods, yet assessing video's motion remains a major challenge. Specifically, there are two key issues: 1) current motion metrics do not fully align with human perceptions; 2) the existing motion prompts are limited. Based on these findings, we introduce VMBench--a comprehensive Video Motion Benchmark that has perception-aligned motion metrics and features the most diverse types of motion. VMBench has several appealing properties: 1) Perception-Driven Motion Evaluation Metrics, we identify five dimensions based on human perception in motion video assessment and develop fine-grained evaluation metrics, providing deeper insights into models' strengths and weaknesses in motion quality. 2) Meta-Guided Motion Prompt Generation, a structured method that extracts meta-information, generates diverse motion prompts with LLMs, and refines them through human-AI validation, resulting in a multi-level prompt library covering six key dynamic scene dimensions. 3) Human-Aligned Validation Mechanism, we provide human preference annotations to validate our benchmarks, with our metrics achieving an average 35.3% improvement in Spearman's correlation over baseline methods. This is the first time that the quality of motion in videos has been evaluated from the perspective of human perception alignment. Additionally, we will soon release VMBench at https://github.com/GD-AIGC/VMBench, setting a new standard for evaluating and advancing motion generation models.

cs.CV↗

arXiv2024

EVLM: An Efficient Vision-Language Model for Visual Understanding

In the field of multi-modal language models, the majority of methods are built on an architecture similar to LLaVA. These models use a single-layer ViT feature as a visual prompt, directly feeding it into the language models alongside textual tokens. However, when dealing with long sequences of visual signals or inputs such as videos, the self-attention mechanism of language models can lead to significant computational overhead. Additionally, using single-layer ViT features makes it challenging for large language models to perceive visual signals fully. This paper proposes an efficient multi-modal language model to minimize computational costs while enabling the model to perceive visual signals as comprehensively as possible. Our method primarily includes: (1) employing cross-attention to image-text interaction similar to Flamingo. (2) utilize hierarchical ViT features. (3) introduce the Mixture of Experts (MoE) mechanism to enhance model effectiveness. Our model achieves competitive scores on public multi-modal benchmarks and performs well in tasks such as image captioning and video captioning.

cs.CV↗