arXiv subjects

Yue Ma

Publications and source records attributed to Yue Ma.

At least 91 records · Page 5Linked to original sources

arXiv2025

EVCtrl: Efficient Control Adapter for Visual Generation

Visual generation includes both image and video generation, training probabilistic models to create coherent, diverse, and semantically faithful content from scratch. While early research focused on unconditional sampling, practitioners now demand controllable generation that allows precise specification of layout, pose, motion, or style. While ControlNet grants precise spatial-temporal control, its auxiliary branch markedly increases latency and introduces redundant computation in both uncontrolled regions and denoising steps, especially for video. To address this problem, we introduce EVCtrl, a lightweight, plug-and-play control adapter that slashes overhead without retraining the model. Specifically, we propose a spatio-temporal dual caching strategy for sparse control information. For spatial redundancy, we first profile how each layer of DiT-ControlNet responds to fine-grained control, then partition the network into global and local functional zones. A locality-aware cache focuses computation on the local zones that truly need the control signal, skipping the bulk of redundant computation in global regions. For temporal redundancy, we selectively omit unnecessary denoising steps to improve efficiency. Extensive experiments on CogVideo-Controlnet, Wan2.1-Controlnet, and Flux demonstrate that our method is effective in image and video control generation without the need for training. For example, it achieves 2.16 and 2.05 times speedups on CogVideo-Controlnet and Wan2.1-Controlnet, respectively, with almost no degradation in generation quality.Codes are available in the supplementary materials.

cs.CV↗

arXiv2025

ContextFlow: Training-Free Video Object Editing via Adaptive Context Enrichment

Training-free video object editing aims to achieve precise object-level manipulation, including object insertion, swapping, and deletion. However, it faces significant challenges in maintaining fidelity and temporal consistency. Existing methods, often designed for U-Net architectures, suffer from two primary limitations: inaccurate inversion due to first-order solvers, and contextual conflicts caused by crude "hard" feature replacement. These issues are more challenging in Diffusion Transformers (DiTs), where the unsuitability of prior layer-selection heuristics makes effective guidance challenging. To address these limitations, we introduce ContextFlow, a novel training-free framework for DiT-based video object editing. In detail, we first employ a high-order Rectified Flow solver to establish a robust editing foundation. The core of our framework is Adaptive Context Enrichment (for specifying what to edit), a mechanism that addresses contextual conflicts. Instead of replacing features, it enriches the self-attention context by concatenating Key-Value pairs from parallel reconstruction and editing paths, empowering the model to dynamically fuse information. Additionally, to determine where to apply this enrichment (for specifying where to edit), we propose a systematic, data-driven analysis to identify task-specific vital layers. Based on a novel Guidance Responsiveness Metric, our method pinpoints the most influential DiT blocks for different tasks (e.g., insertion, swapping), enabling targeted and highly effective guidance. Extensive experiments show that ContextFlow significantly outperforms existing training-free methods and even surpasses several state-of-the-art training-based approaches, delivering temporally coherent, high-fidelity results.

cs.CV↗

arXiv2025

MILD: Multi-Layer Diffusion Strategy for Complex and Precise Multi-IP Aware Human Erasing

Recent years have witnessed the success of diffusion models in image customization tasks. However, existing mask-guided human erasing methods still struggle in complex scenarios such as human-human occlusion, human-object entanglement, and human-background interference, mainly due to the lack of large-scale multi-instance datasets and effective spatial decoupling to separate foreground from background. To bridge these gaps, we curate the MILD dataset capturing diverse poses, occlusions, and complex multi-instance interactions. We then define the Cross-Domain Attention Gap (CAG), an attention-gap metric to quantify semantic leakage. On top of these, we propose Multi-Layer Diffusion (MILD), which decomposes the generation process into independent denoising pathways, enabling separate reconstruction of each foreground instance and the background. To enhance human-centric understanding, we introduce Human Morphology Guidance, a plug-and-play module that incorporates pose, parsing, and spatial relationships into the diffusion process to improve structural awareness and restoration quality. Additionally, we present Spatially-Modulated Attention, an adaptive mechanism that leverages spatial mask priors to modulate attention across semantic regions, further widening the CAG to effectively minimize boundary artifacts and mitigate semantic leakage. Experiments show that MILD significantly outperforms existing methods. Datasets and code are publicly available at: https://mild-multi-layer-diffusion.github.io/.

cs.CV↗

arXiv2025

A Survey on Cache Methods in Diffusion Models: Toward Efficient Multi-Modal Generation

Diffusion Models have become a cornerstone of modern generative AI for their exceptional generation quality and controllability. However, their inherent \textit{multi-step iterations} and \textit{complex backbone networks} lead to prohibitive computational overhead and generation latency, forming a major bottleneck for real-time applications. Although existing acceleration techniques have made progress, they still face challenges such as limited applicability, high training costs, or quality degradation. Against this backdrop, \textbf{Diffusion Caching} offers a promising training-free, architecture-agnostic, and efficient inference paradigm. Its core mechanism identifies and reuses intrinsic computational redundancies in the diffusion process. By enabling feature-level cross-step reuse and inter-layer scheduling, it reduces computation without modifying model parameters. This paper systematically reviews the theoretical foundations and evolution of Diffusion Caching and proposes a unified framework for its classification and analysis. Through comparative analysis of representative methods, we show that Diffusion Caching evolves from \textit{static reuse} to \textit{dynamic prediction}. This trend enhances caching flexibility across diverse tasks and enables integration with other acceleration techniques such as sampling optimization and model distillation, paving the way for a unified, efficient inference framework for future multimodal and interactive applications. We argue that this paradigm will become a key enabler of real-time and efficient generative AI, injecting new vitality into both theory and practice of \textit{Efficient Generative Intelligence}.

cs.LG↗

arXiv2025

EEdit: Rethinking the Spatial and Temporal Redundancy for Efficient Image Editing

Inversion-based image editing is rapidly gaining momentum while suffering from significant computation overhead, hindering its application in real-time interactive scenarios. In this paper, we rethink that the redundancy in inversion-based image editing exists in both the spatial and temporal dimensions, such as the unnecessary computation in unedited regions and the redundancy in the inversion progress. To tackle these challenges, we propose a practical framework, named EEdit, to achieve efficient image editing. Specifically, we introduce three techniques to solve them one by one. For spatial redundancy, spatial locality caching is introduced to compute the edited region and its neighboring regions while skipping the unedited regions, and token indexing preprocessing is designed to further accelerate the caching. For temporal redundancy, inversion step skipping is proposed to reuse the latent for efficient editing. Our experiments demonstrate an average of 2.46 $\times$ acceleration without performance drop in a wide range of editing tasks including prompt-guided image editing, dragging and image composition. Our codes are available at https://github.com/yuriYanZeXuan/EEdit

cs.CV↗

arXiv2025

RECALL: REpresentation-aligned Catastrophic-forgetting ALLeviation via Hierarchical Model Merging

We unveil that internal representations in large language models (LLMs) serve as reliable proxies of learned knowledge, and propose RECALL, a novel representation-aware model merging framework for continual learning without access to historical data. RECALL computes inter-model similarity from layer-wise hidden representations over clustered typical samples, and performs adaptive, hierarchical parameter fusion to align knowledge across models. This design enables the preservation of domain-general features in shallow layers while allowing task-specific adaptation in deeper layers. Unlike prior methods that require task labels or incur performance trade-offs, RECALL achieves seamless multi-domain integration and strong resistance to catastrophic forgetting. Extensive experiments across five NLP tasks and multiple continual learning scenarios show that RECALL outperforms baselines in both knowledge retention and generalization, providing a scalable and data-free solution for evolving LLMs.

cs.CL↗

arXiv2025

The global nonlinear stability of Minkowski spacetime with self-gravitating massive Dirac fields

We consider the Einstein-Dirac system for a massive field, which describes the evolution of self-gravitating massive spinor fields, and we investigate the global evolution problem, when the initial data set is sufficiently close to data describing a spacelike, asymptotically Euclidean slice of the Minkowski spacetime. We establish the gauge-invariant nonlinear stability of such fields, namely the existence of a globally hyperbolic development, which remains asymptotic to Minkowski spacetime in future timelike, null, and spacelike directions. Previous results on this problem have been limited to the Einstein-Dirac system in the massless case. Our analysis follows the asymptotically hyperboloidal-Euclidean framework introduced by LeFloch and Y. Ma for the massive Klein-Gordon-Einstein system. The structure specific to spinor fields and the Dirac equation necessitates significantly new elements in the proof. In contrast with prior approaches, our treatment of spinor fields and the Dirac equation is gauge-invariant, relying on the formalism of Lorentz Clifford algebras, principal fiber bundles, etc. Our analysis is carried out with the metric expressed in light-bending wave coordinates, as we call them. This leads us to the study of a global existence problem for a system of wave equations with constraints and a Klein-Gordon-type equations. We derive L2 estimates for the Dirac equation and its coupling with the Einstein equations, along with $pointwise estimates. New Sobolev inequalities are proven for spinor fields in a gauge-invariant manner in the hyperboloidal-Euclidean foliation. The nonlinear coupling between the massive Dirac equation and the Einstein equations is investigated, and we establish a hierarchy of estimates, which distinguish between translations, rotations, and boosts.

gr-qc↗

arXiv2025

DiTraj: training-free trajectory control for video diffusion transformer

Diffusion Transformers (DiT)-based video generation models with 3D full attention exhibit strong generative capabilities. Trajectory control represents a user-friendly task in the field of controllable video generation. However, existing methods either require substantial training resources or are specifically designed for U-Net, do not take advantage of the superior performance of DiT. To address these issues, we propose DiTraj, a simple but effective training-free framework for trajectory control in text-to-video generation, tailored for DiT. Specifically, first, to inject the object's trajectory, we propose foreground-background separation guidance: we use the Large Language Model (LLM) to convert user-provided prompts into foreground and background prompts, which respectively guide the generation of foreground and background regions in the video. Then, we analyze 3D full attention and explore the tight correlation between inter-token attention scores and position embedding. Based on this, we propose inter-frame Spatial-Temporal Decoupled 3D-RoPE (STD-RoPE). By modifying only foreground tokens' position embedding, STD-RoPE eliminates their cross-frame spatial discrepancies, strengthening cross-frame attention among them and thus enhancing trajectory control. Additionally, we achieve 3D-aware trajectory control by regulating the density of position embedding. Extensive experiments demonstrate that our method outperforms previous methods in both video quality and trajectory controllability.

cs.CV↗

arXiv2025

Follow-Your-Preference: Towards Preference-Aligned Image Inpainting

This paper investigates image inpainting with preference alignment. Instead of introducing a novel method, we go back to basics and revisit fundamental problems in achieving such alignment. We leverage the prominent direct preference optimization approach for alignment training and employ public reward models to construct preference training datasets. Experiments are conducted across nine reward models, two benchmarks, and two baseline models with varying structures and generative algorithms. Our key findings are as follows: (1) Most reward models deliver valid reward scores for constructing preference data, even if some of them are not reliable evaluators. (2) Preference data demonstrates robust trends in both candidate scaling and sample scaling across models and benchmarks. (3) Observable biases in reward models, particularly in brightness, composition, and color scheme, render them susceptible to cause reward hacking. (4) A simple ensemble of these models yields robust and generalizable results by mitigating such biases. Built upon these observations, our alignment models significantly outperform prior models across standard metrics, GPT-4 assessments, and human evaluations, without any changes to model structures or the use of new datasets. We hope our work can set a simple yet solid baseline, pushing this promising frontier. Our code is open-sourced at: https://github.com/shenytzzz/Follow-Your-Preference.

cs.CV↗

arXiv2025

Follow-Your-Emoji-Faster: Towards Efficient, Fine-Controllable, and Expressive Freestyle Portrait Animation

We present Follow-Your-Emoji-Faster, an efficient diffusion-based framework for freestyle portrait animation driven by facial landmarks. The main challenges in this task are preserving the identity of the reference portrait, accurately transferring target expressions, and maintaining long-term temporal consistency while ensuring generation efficiency. To address identity preservation and accurate expression retargeting, we enhance Stable Diffusion with two key components: a expression-aware landmarks as explicit motion signals, which improve motion alignment, support exaggerated expressions, and reduce identity leakage; and a fine-grained facial loss that leverages both expression and facial masks to better capture subtle expressions and faithfully preserve the reference appearance. With these components, our model supports controllable and expressive animation across diverse portrait types, including real faces, cartoons, sculptures, and animals. However, diffusion-based frameworks typically struggle to efficiently generate long-term stable animation results, which remains a core challenge in this task. To address this, we propose a progressive generation strategy for stable long-term animation, and introduce a Taylor-interpolated cache, achieving a 2.6X lossless acceleration. These two strategies ensure that our method produces high-quality results efficiently, making it user-friendly and accessible. Finally, we introduce EmojiBench++, a more comprehensive benchmark comprising diverse portraits, driving videos, and landmark sequences. Extensive evaluations on EmojiBench++ demonstrate that Follow-Your-Emoji-Faster achieves superior performance in both animation quality and controllability. The code, training dataset and benchmark will be found in https://follow-your-emoji.github.io/.

cs.CV↗

arXiv2025

The generalized Turán number for K_3 in graphs without suspensions of a path on five vertices

Given graphs $H$ and $F$, the generalized Turán number $\ex(n, H, F)$ is defined as the maximum number of copies of $H$ in an $n$-vertex graph that contains no copy of $F$. The suspension $\widehat{F}$ of a graph $F$ is obtained by adding a new vertex that is adjacent to every vertex of $F$. Mubayi and Mukherjee (2023, DM) conjectured that $\ex(n, K_3, \widehat{P_k})=\left\lfloor \frac{k-2}{2}\right\rfloor \cdot \frac{n^2}{8}+o(n^2)$, where $P_k$ is a path on $k\ge 4$ vertices. Using the triangle removal lemma, they verified this conjecture for $k=4,5,6$. Later, Mukherjee (2024, DM) established the exact value $\ex(n, K_3, \widehat{P_4})=\left\lfloor n^2/8\right\rfloor$. In this paper, using the stability method, we determine the exact value of $\ex(n, K_3, \widehat{P_5})$ by showing that for sufficiently large $n$, $\ex(n,K_3, \widehat{P_5})=\left\lfloor n^2/8\right\rfloor.$

math.CO↗

arXiv2025

Forecast then Calibrate: Feature Caching as ODE for Efficient Diffusion Transformers

Diffusion Transformers (DiTs) have demonstrated exceptional performance in high-fidelity image and video generation. To reduce their substantial computational costs, feature caching techniques have been proposed to accelerate inference by reusing hidden representations from previous timesteps. However, current methods often struggle to maintain generation quality at high acceleration ratios, where prediction errors increase sharply due to the inherent instability of long-step forecasting. In this work, we adopt an ordinary differential equation (ODE) perspective on the hidden-feature sequence, modeling layer representations along the trajectory as a feature-ODE. We attribute the degradation of existing caching strategies to their inability to robustly integrate historical features under large skipping intervals. To address this, we propose FoCa (Forecast-then-Calibrate), which treats feature caching as a feature-ODE solving problem. Extensive experiments on image synthesis, video generation, and super-resolution tasks demonstrate the effectiveness of FoCa, especially under aggressive acceleration. Without additional training, FoCa achieves near-lossless speedups of 5.50 times on FLUX, 6.45 times on HunyuanVideo, 3.17 times on Inf-DiT, and maintains high quality with a 4.53 times speedup on DiT.

cs.CV↗

arXiv2025

Follow-Your-Color: Multi-Instance Sketch Colorization

We present Follow-Your-Color, a diffusion-based framework for multi-instance sketch colorization. The production of multi-instance 2D line art colorization adheres to an industry-standard workflow, which consists of three crucial stages: the design of line art characters, the coloring of individual objects, and the refinement process. The artists are required to repeat the process of coloring each instance one by one, which is inaccurate and inefficient. Meanwhile, current generative methods fail to solve this task due to the challenge of multi-instance pair data collection. To tackle these challenges, we incorporate three technical designs to ensure precise character detail transcription and achieve multi-instance sketch colorization in a single forward pass. Specifically, we first propose the self-play training strategy to address the lack of training data. Then we introduce an instance guider to feed the color of the instance. To achieve accurate color matching, we present fine-grained color matching with edge loss to enhance visual quality. Equipped with the proposed modules, Follow-Your-Color enables automatically transforming sketches into vividly-colored images with accurate consistency and multi-instance control. Experiments on our collected datasets show that our model outperforms existing methods regarding chromatic precision. Specifically, our model critically automates the colorization process with zero manual adjustments, so novice users can produce stylistically consistent artwork by providing reference instances and the original line art. Our code and additional details are available at https://yinhan-zhang.github.io/color.

cs.CV↗

arXiv2025

Follow-Your-Instruction: A Comprehensive MLLM Agent for World Data Synthesis

With the growing demands of AI-generated content (AIGC), the need for high-quality, diverse, and scalable data has become increasingly crucial. However, collecting large-scale real-world data remains costly and time-consuming, hindering the development of downstream applications. While some works attempt to collect task-specific data via a rendering process, most approaches still rely on manual scene construction, limiting their scalability and accuracy. To address these challenges, we propose Follow-Your-Instruction, a Multimodal Large Language Model (MLLM)-driven framework for automatically synthesizing high-quality 2D, 3D, and 4D data. Our \textbf{Follow-Your-Instruction} first collects assets and their associated descriptions through multimodal inputs using the MLLM-Collector. Then it constructs 3D layouts, and leverages Vision-Language Models (VLMs) for semantic refinement through multi-view scenes with the MLLM-Generator and MLLM-Optimizer, respectively. Finally, it uses MLLM-Planner to generate temporally coherent future frames. We evaluate the quality of the generated data through comprehensive experiments on the 2D, 3D, and 4D generative tasks. The results show that our synthetic data significantly boosts the performance of existing baseline models, demonstrating Follow-Your-Instruction's potential as a scalable and effective data engine for generative intelligence.

cs.CV↗

arXiv2025

GR-Gaussian: Graph-Based Radiative Gaussian Splatting for Sparse-View CT Reconstruction

3D Gaussian Splatting (3DGS) has emerged as a promising approach for CT reconstruction. However, existing methods rely on the average gradient magnitude of points within the view, often leading to severe needle-like artifacts under sparse-view conditions. To address this challenge, we propose GR-Gaussian, a graph-based 3D Gaussian Splatting framework that suppresses needle-like artifacts and improves reconstruction accuracy under sparse-view conditions. Our framework introduces two key innovations: (1) a Denoised Point Cloud Initialization Strategy that reduces initialization errors and accelerates convergence; and (2) a Pixel-Graph-Aware Gradient Strategy that refines gradient computation using graph-based density differences, improving splitting accuracy and density representation. Experiments on X-3D and real-world datasets validate the effectiveness of GR-Gaussian, achieving PSNR improvements of 0.67 dB and 0.92 dB, and SSIM gains of 0.011 and 0.021. These results highlight the applicability of GR-Gaussian for accurate CT reconstruction under challenging sparse-view conditions.

eess.IV↗

arXiv2025

UniPaint: Unified Space-time Video Inpainting via Mixture-of-Experts

In this paper, we present UniPaint, a unified generative space-time video inpainting framework that enables spatial-temporal inpainting and interpolation. Different from existing methods that treat video inpainting and video interpolation as two distinct tasks, we leverage a unified inpainting framework to tackle them and observe that these two tasks can mutually enhance synthesis performance. Specifically, we first introduce a plug-and-play space-time video inpainting adapter, which can be employed in various personalized models. The key insight is to propose a Mixture of Experts (MoE) attention to cover various tasks. Then, we design a spatial-temporal masking strategy during the training stage to mutually enhance each other and improve performance. UniPaint produces high-quality and aesthetically pleasing results, achieving the best quantitative results across various tasks and scale setups. The code and checkpoints are available at $\href{https://github.com/mmmmm-w/UniPaint}{this \ repository}$.

cs.CV↗

arXiv2025

MagicAnime: A Hierarchically Annotated, Multimodal and Multitasking Dataset with Benchmarks for Cartoon Animation Generation

Generating high-quality cartoon animations multimodal control is challenging due to the complexity of non-human characters, stylistically diverse motions and fine-grained emotions. There is a huge domain gap between real-world videos and cartoon animation, as cartoon animation is usually abstract and has exaggerated motion. Meanwhile, public multimodal cartoon data are extremely scarce due to the difficulty of large-scale automatic annotation processes compared with real-life scenarios. To bridge this gap, We propose the MagicAnime dataset, a large-scale, hierarchically annotated, and multimodal dataset designed to support multiple video generation tasks, along with the benchmarks it includes. Containing 400k video clips for image-to-video generation, 50k pairs of video clips and keypoints for whole-body annotation, 12k pairs of video clips for video-to-video face animation, and 2.9k pairs of video and audio clips for audio-driven face animation. Meanwhile, we also build a set of multi-modal cartoon animation benchmarks, called MagicAnime-Bench, to support the comparisons of different methods in the tasks above. Comprehensive experiments on four tasks, including video-driven face animation, audio-driven face animation, image-to-video animation, and pose-driven character animation, validate its effectiveness in supporting high-fidelity, fine-grained, and controllable generation.

cs.CV↗

arXiv2025

Uncertainty-aware Probabilistic 3D Human Motion Forecasting via Invertible Networks

3D human motion forecasting aims to enable autonomous applications. Estimating uncertainty for each prediction (i.e., confidence based on probability density or quantile) is essential for safety-critical contexts like human-robot collaboration to minimize risks. However, existing diverse motion forecasting approaches struggle with uncertainty quantification due to implicit probabilistic representations hindering uncertainty modeling. We propose ProbHMI, which introduces invertible networks to parameterize poses in a disentangled latent space, enabling probabilistic dynamics modeling. A forecasting module then explicitly predicts future latent distributions, allowing effective uncertainty quantification. Evaluated on benchmarks, ProbHMI achieves strong performance for both deterministic and diverse prediction while validating uncertainty calibration, critical for risk-aware decision making.

cs.RO↗