Ao Ma

Zhiyuan Li, Heng Wang, Dongnan Liu et al.

Multimodal Large Language Models (MLLMs) have showcased exceptional Chain-of-Thought (CoT) reasoning ability in complex textual inference tasks including causal reasoning. However, will these causalities remain straightforward when crucial hints hide in visual details? If not, what factors might influence cross-modal generalization? Whether we can effectively enhance their capacity for robust causal inference across both text and vision? Motivated by these, we introduce MuCR - a novel Multimodal Causal Reasoning benchmark that leverages synthetic siamese images and text pairs to challenge MLLMs. Additionally, we develop tailored metrics from multiple perspectives, including image-level match, phrase-level understanding, and sentence-level explanation, to comprehensively assess MLLMs' comprehension abilities. Our experiments reveal that current MLLMs fall short in multimodal causal reasoning compared to their performance in purely textual settings. Additionally, we find that identifying visual cues across images is key to effective cross-modal generalization. Finally, we propose a VcCoT strategy that better highlights visual cues, and our results confirm its efficacy in enhancing multimodal causal reasoning. The project is available at: https://github.com/Zhiyuan-Li-John/MuCR

Jiahao Fan, Yuxin Qin, Wei Feng et al.

Product posters blend striking visuals with informative text to highlight the product and capture customer attention. However, crafting appealing posters and manually optimizing them based on online performance is laborious and resource-consuming. To address this, we introduce AutoPP, an automated pipeline for product poster generation and optimization that eliminates the need for human intervention. Specifically, the generator, relying solely on basic product information, first uses a unified design module to integrate the three key elements of a poster (background, text, and layout) into a cohesive output. Then, an element rendering module encodes these elements into condition tokens, efficiently and controllably generating the product poster. Based on the generated poster, the optimizer enhances its Click-Through Rate (CTR) by leveraging online feedback. It systematically replaces elements to gather fine-grained CTR comparisons and utilizes Isolated Direct Preference Optimization (IDPO) to attribute CTR gains to isolated elements. Our work is supported by AutoPP1M, the largest dataset specifically designed for product poster generation and optimization, which contains one million high-quality posters and feedback collected from over one million users. Experiments demonstrate that AutoPP achieves state-of-the-art results in both offline and online settings. Our code and dataset are publicly available at: https://github.com/JD-GenX/AutoPP

Shuo Lu, Haohan Wang, Wei Feng et al.

Advertising image generation has increasingly focused on online metrics like Click-Through Rate (CTR), yet existing approaches adopt a ``one-size-fits-all" strategy that optimizes for overall CTR while neglecting preference diversity among user groups. This leads to suboptimal performance for specific groups, limiting targeted marketing effectiveness. To bridge this gap, we present \textit{One Size, Many Fits} (OSMF), a unified framework that aligns diverse group-wise click preferences in large-scale advertising image generation. OSMF begins with product-aware adaptive grouping, which dynamically organizes users based on their attributes and product characteristics, representing each group with rich collective preference features. Building on these groups, preference-conditioned image generation employs a Group-aware Multimodal Large Language Model (G-MLLM) to generate tailored images for each group. The G-MLLM is pre-trained to simultaneously comprehend group features and generate advertising images. Subsequently, we fine-tune the G-MLLM using our proposed Group-DPO for group-wise preference alignment, which effectively enhances each group's CTR on the generated images. To further advance this field, we introduce the Grouped Advertising Image Preference Dataset (GAIP), the first large-scale public dataset of group-wise image preferences, including around 600K groups built from 40M users. Extensive experiments demonstrate that our framework achieves the state-of-the-art performance in both offline and online settings. Our code and datasets will be released at https://github.com/JD-GenX/OSMF.

Yexing Xu, Wei Feng, Shen Zhang et al.

Generating realistic and user-preferred advertisements is a key challenge in e-commerce. Existing approaches utilize multiple independent models driven by click-through-rate (CTR) to controllably create attractive image or text advertisements. However, their pipelines lack cross-modal perception and rely on CTR that only reflects average preferences. Therefore, we explore jointly generating personalized image-text advertisements from historical click behaviors. We first design a Unified Advertisement Generative model (Uni-AdGen) that employs a single autoregressive framework to produce both advertising images and texts. By incorporating a foreground perception module and instruction tuning, Uni-AdGen enhances the realism of the generated content. To further personalize advertisements, we equip Uni-AdGen with a coarse-to-fine preference understanding module that effectively captures user interests from noisy multimodal historical behaviors to drive personalized generation. Additionally, we construct the first large-scale Personalized Advertising image-text dataset (PAd1M) and introduce a Product Background Similarity (PBS) metric to facilitate training and evaluation. Extensive experiments show that our method outperforms baselines in general and personalized advertisement generation. Our project is available at https://github.com/JD-GenX/Uni-AdGen.

Zeyinzi Jiang, Chaojie Mao, Ziyuan Huang et al.

Parameter-efficient tuning has become a trend in transferring large-scale foundation models to downstream applications. Existing methods typically embed some light-weight tuners into the backbone, where both the design and the learning of the tuners are highly dependent on the base model. This work offers a new tuning paradigm, dubbed Res-Tuning, which intentionally unbinds tuners from the backbone. With both theoretical and empirical evidence, we show that popular tuning approaches have their equivalent counterparts under our unbinding formulation, and hence can be integrated into our framework effortlessly. Thanks to the structural disentanglement, we manage to free the design of tuners from the network architecture, facilitating flexible combination of various tuning strategies. We further propose a memory-efficient variant of Res-Tuning, where the bypass i.e., formed by a sequence of tuners) is effectively detached from the main branch, such that the gradients are back-propagated only to the tuners but not to the backbone. Such a detachment also allows one-time backbone forward for multi-task inference. Extensive experiments on both discriminative and generative tasks demonstrate the superiority of our method over existing alternatives from the perspectives of efficacy and efficiency. Project page: $\href{https://res-tuning.github.io/}{\textit{https://res-tuning.github.io/}}$.

Qiugang Zhan, Anning Jiang, Ran Tao et al.

Pre-trained vision models have found widespread application across diverse domains. Prompt tuning-based methods have emerged as a parameter-efficient paradigm for adapting pre-trained vision models. While effective on standard benchmarks, the continuous and dense nature of learned prompts can lead to sensitivity against input noise, as the high-capacity prompts tend to overfit task-irrelevant details. To address this trade-off, we propose Spike-NVPT, a noise-robust visual prompt tuning method. Specifically, we design a Signal Filtering Layer based on spiking neurons, which uses the integrate-and-fire (IF) mechanism to accumulate task-relevant signals over time and filter transient noise fluctuations. A subsequent Spike Discretization Unit converts filtered signals into sparse binary prompts. This discretization acts as a strong regularizer, forcing the model to anchor decision boundaries on the most discriminative and robust features. Notably, the resulting binary prompts remain static during deployment, ensuring zero additional computational overhead during inference. Experimental results demonstrate that Spike-NVPT achieves superior robustness performance, with a maximum improvement of 11.2% over conventional methods, and retains competitive accuracy on clean datasets. To the best of our knowledge, this is the first attempt to leverage spiking neurons for fine-tuning traditional artificial neural network (ANN)-based visual models.

Jiasong Feng, Ao Ma, Jing Wang et al.

Synthesizing motion-rich and temporally consistent videos remains a challenge in artificial intelligence, especially when dealing with extended durations. Existing text-to-video (T2V) models commonly employ spatial cross-attention for text control, equivalently guiding different frame generations without frame-specific textual guidance. Thus, the model's capacity to comprehend the temporal logic conveyed in prompts and generate videos with coherent motion is restricted. To tackle this limitation, we introduce FancyVideo, an innovative video generator that improves the existing text-control mechanism with the well-designed Cross-frame Textual Guidance Module (CTGM). Specifically, CTGM incorporates the Temporal Information Injector (TII) and Temporal Affinity Refiner (TAR) at the beginning and end of cross-attention, respectively, to achieve frame-specific textual guidance. Firstly, TII injects frame-specific information from latent features into text conditions, thereby obtaining cross-frame textual conditions. Then, TAR refines the correlation matrix between cross-frame textual conditions and latent features along the time dimension. Extensive experiments comprising both quantitative and qualitative evaluations demonstrate the effectiveness of FancyVideo. Our approach achieves state-of-the-art T2V generation results on the EvalCrafter benchmark and facilitates the synthesis of dynamic and consistent videos. Note that the T2V process of FancyVideo essentially involves a text-to-image step followed by T+I2V. This means it also supports the generation of videos from user images, i.e., the image-to-video (I2V) task. A significant number of experiments have shown that its performance is also outstanding.

Run Ling, Ke Cao, Jian Lu et al.

Multi-subject video generation aims to synthesize videos from textual prompts and multiple reference images, ensuring that each subject preserves natural scale and visual fidelity. However, current methods face two challenges: scale inconsistency, where variations in subject size lead to unnatural generation, and permutation sensitivity, where the order of reference inputs causes subject distortion. In this paper, we propose MoFu, a unified framework that tackles both challenges. For scale inconsistency, we introduce Scale-Aware Modulation (SMO), an LLM-guided module that extracts implicit scale cues from the prompt and modulates features to ensure consistent subject sizes. To address permutation sensitivity, we present a simple yet effective Fourier Fusion strategy that processes the frequency information of reference features via the Fast Fourier Transform to produce a unified representation. Besides, we design a Scale-Permutation Stability Loss to jointly encourage scale-consistent and permutation-invariant generation. To further evaluate these challenges, we establish a dedicated benchmark with controlled variations in subject scale and reference permutation. Extensive experiments demonstrate that MoFu significantly outperforms existing methods in preserving natural scale, subject fidelity, and overall visual quality.

Jing Wang, Ao Ma, Jiasong Feng et al.

The global self-attention mechanism in diffusion transformers involves redundant computation due to the sparse and redundant nature of visual information, and the attention map of tokens within a spatial window shows significant similarity. To address this redundancy, we propose the Proxy-Tokenized Diffusion Transformer (PT-DiT), which employs sparse representative token attention (where the number of representative tokens is much smaller than the total number of tokens) to model global visual information efficiently. Specifically, within each transformer block, we compute an averaging token from each spatial-temporal window to serve as a proxy token for that region. The global semantics are captured through the self-attention of these proxy tokens and then injected into all latent tokens via cross-attention. Simultaneously, we introduce window and shift window attention to address the limitations in detail modeling caused by the sparse attention mechanism. Building on the well-designed PT-DiT, we further develop the Qihoo-T2X family, which includes a variety of models for T2I, T2V, and T2MV tasks. Experimental results show that PT-DiT achieves competitive performance while reducing the computational complexity in both image and video generation tasks (e.g., a 49% reduction compared to DiT and a 34% reduction compared to PixArt-$α$). The visual exhibition and source code of Qihoo-T2X is available at https://360cvgroup.github.io/Qihoo-T2X/.

Bo Cheng, Yuhang Ma, Liebucha Wu et al.

The task of layout-to-image generation involves synthesizing images based on the captions of objects and their spatial positions. Existing methods still struggle in complex layout generation, where common bad cases include object missing, inconsistent lighting, conflicting view angles, etc. To effectively address these issues, we propose a \textbf{Hi}erarchical \textbf{Co}ntrollable (HiCo) diffusion model for layout-to-image generation, featuring object seperable conditioning branch structure. Our key insight is to achieve spatial disentanglement through hierarchical modeling of layouts. We use a multi branch structure to represent hierarchy and aggregate them in fusion module. To evaluate the performance of multi-objective controllable layout generation in natural scenes, we introduce the HiCo-7K benchmark, derived from the GRIT-20M dataset and manually cleaned. https://github.com/360CVGroup/HiCo_T2I.

Yicheng Di, Wei Yuan, Tieke He et al.

Vision-language pretrained models offer strong transferable representations, yet adapting them in privacy-sensitive multi-party settings is challenging due to the high communication cost of federated optimization and the limited local data on clients. Federated prompt learning mitigates this issue by keeping the VLPM backbone frozen and collaboratively training lightweight prompt parameters. However, existing approaches typically enforce a unified prompt structure and length across clients, which is inadequate under practical client heterogeneity in both data distributions and system resources, and may further introduce conflicts between globally shared and locally optimal knowledge. To address these challenges, we propose \textbf{SDFed}, a heterogeneous federated prompt learning framework that bridges Local-Global Discrepancy via Subspace Refinement and Divergence Control. SDFed maintains a fixed-length global prompt for efficient aggregation while allowing each client to learn a variable-length local prompt to better match its data characteristics and capacity. To mitigate local-global conflicts and facilitate effective knowledge transfer, SDFed introduces a subspace refinement method for local prompts and an information retention and divergence control strategy that preserves key local information while maintaining appropriate separability between global and local representations. Extensive experiments on several datasets demonstrate that SDFed consistently improves performance and robustness in heterogeneous federated settings.

Runze He, Yiji Cheng, Tiankai Hang et al.

In-context image generation and editing (ICGE) enables users to specify visual concepts through interleaved image-text prompts, demanding precise understanding and faithful execution of user intent. Although recent unified multimodal models exhibit promising understanding capabilities, these strengths often fail to transfer effectively to image generation. We introduce Re-Align, a unified framework that bridges the gap between understanding and generation through structured reasoning-guided alignment. At its core lies the In-Context Chain-of-Thought (IC-CoT), a structured reasoning paradigm that decouples semantic guidance and reference association, providing clear textual target and mitigating confusion among reference images. Furthermore, Re-Align introduces an effective RL training scheme that leverages a surrogate reward to measure the alignment between structured reasoning text and the generated image, thereby improving the model's overall performance on ICGE tasks. Extensive experiments verify that Re-Align outperforms competitive methods of comparable model scale and resources on both in-context image generation and editing tasks.

Yun Wang, Longguang Wang, Chenghao Zhang et al.

Recently, learning-based stereo matching networks have advanced significantly. However, they often lack robustness and struggle to achieve impressive cross-domain performance due to domain shifts and imbalanced disparity distributions among diverse datasets. Leveraging Vision Foundation Models (VFMs) can intuitively enhance the model's robustness, but integrating such a model into stereo matching cost-effectively to fully realize their robustness remains a key challenge. To address this, we propose SMoEStereo, a novel framework that adapts VFMs for stereo matching through a tailored, scene-specific fusion of Low-Rank Adaptation (LoRA) and Mixture-of-Experts (MoE) modules. SMoEStereo introduces MoE-LoRA with adaptive ranks and MoE-Adapter with adaptive kernel sizes. The former dynamically selects optimal experts within MoE to adapt varying scenes across domains, while the latter injects inductive bias into frozen VFMs to improve geometric feature extraction. Importantly, to mitigate computational overhead, we further propose a lightweight decision network that selectively activates MoE modules based on input complexity, balancing efficiency with accuracy. Extensive experiments demonstrate that our method exhibits state-of-the-art cross-domain and joint generalization across multiple benchmarks without dataset-specific adaptation. The code is available at \textcolor{red}{https://github.com/cocowy1/SMoE-Stereo}.

Yuxin Qin, Ke Cao, Haowei Liu et al.

E-commerce product poster generation aims to automatically synthesize a single image that effectively conveys product information by presenting a subject, text, and a designed style. Recent diffusion models with fine-grained and efficient controllability have advanced product poster synthesis, yet they typically rely on multi-stage pipelines, and simultaneous control over subject, text, and style remains underexplored. Such naive multi-stage pipelines also show three issues: poor subject fidelity, inaccurate text, and inconsistent style. To address these issues, we propose InnoAds-Composer, a single-stage framework that enables efficient tri-conditional control tokens over subject, glyph, and style. To alleviate the quadratic overhead introduced by naive tri-conditional token concatenation, we perform importance analysis over layers and timesteps and route each condition only to the most responsive positions, thereby shortening the active token sequence. Besides, to improve the accuracy of Chinese text rendering, we design a Text Feature Enhancement Module (TFEM) that integrates features from both glyph images and glyph crops. To support training and evaluation, we also construct a high-quality e-commerce product poster dataset and benchmark, which is the first dataset that jointly contains subject, text, and style conditions. Extensive experiments demonstrate that InnoAds-Composer significantly outperforms existing product poster methods without obviously increasing inference latency.

Runze He, Bo Cheng, Yuhang Ma et al.

In this paper, we propose a unified layout planning and image generation model, PlanGen, which can pre-plan spatial layout conditions before generating images. Unlike previous diffusion-based models that treat layout planning and layout-to-image as two separate models, PlanGen jointly models the two tasks into one autoregressive transformer using only next-token prediction. PlanGen integrates layout conditions into the model as context without requiring specialized encoding of local captions and bounding box coordinates, which provides significant advantages over the previous embed-and-pool operations on layout conditions, particularly when dealing with complex layouts. Unified prompting allows PlanGen to perform multitasking training related to layout, including layout planning, layout-to-image generation, image layout understanding, etc. In addition, PlanGen can be seamlessly expanded to layout-guided image manipulation thanks to the well-designed modeling, with teacher-forcing content manipulation policy and negative layout guidance. Extensive experiments verify the effectiveness of our PlanGen in multiple layoutrelated tasks, showing its great potential. Code is available at: https://360cvgroup.github.io/PlanGen.

Shuo Lu, Jianjie Cheng, Yinuo Xu et al.

Multimodal large language models (MLLMs) have achieved strong performance on perception-oriented tasks, yet their ability to perform mathematical spatial reasoning, defined as the capacity to parse and manipulate two- and three-dimensional relations, remains unclear. Humans easily solve textbook-style spatial reasoning problems with over 95\% accuracy, but we find that most leading MLLMs fail to reach even 60\% on the same tasks. This striking gap highlights spatial reasoning as a fundamental weakness of current models. To investigate this gap, we present MathSpatial, a unified framework for evaluating and improving spatial reasoning in MLLMs. MathSpatial includes three complementary components: (i) MathSpatial-Bench, a benchmark of 2K problems across three categories and eleven subtypes, designed to isolate reasoning difficulty from perceptual noise; (ii) MathSpatial-Corpus, a training dataset of 8K additional problems with verified solutions; and (iii) MathSpatial-SRT, which models reasoning as structured traces composed of three atomic operations--Correlate, Constrain, and Infer. Experiments show that fine-tuning Qwen2.5-VL-7B on MathSpatial achieves competitive accuracy while reducing tokens by 25\%. MathSpatial provides the first large-scale resource that disentangles perception from reasoning, enabling precise measurement and comprehensive understanding of mathematical spatial reasoning in MLLMs.

Asbjørn Munk, Ao Ma, Mads Nielsen

The current state-of-the art techniques for image segmentation are often based on U-Net architectures, a U-shaped encoder-decoder networks with skip connections. Despite the powerful performance, the architecture often does not perform well when used on data which has different characteristics than the data it was trained on. Many techniques for improving performance in the presence of domain shift have been developed, however typically only have loose connections to the theory of domain adaption. In this work, we propose an unsupervised domain adaptation framework for U-Nets with theoretical guarantees based on the Margin Disparity Discrepancy [1] called the MDD-UNet. We evaluate the proposed technique on the task of hippocampus segmentation, and find that the MDD-UNet is able to learn features which are domain-invariant with no knowledge about the labels in the target domain. The MDD-UNet improves performance over the standard U-Net on 11 out of 12 combinations of datasets. This work serves as a proof of concept by demonstrating an improvement on the U-Net in it's standard form without modern enhancements, which opens up a new avenue of studying domain adaptation for models with very large hypothesis spaces from both methodological and practical perspectives. Code is available at https://github.com/asbjrnmunk/mdd-unet.

Shenghan Zhang, Run Ling, Ke Cao et al.

Federated learning (FL) has emerged as a promising paradigm for privacy-preserving multi-camera video understanding. However, applying FL to cross-view scenarios faces three major challenges: (i) heterogeneous viewpoints and backgrounds lead to highly non-IID client distributions and overfitting to view-specific patterns, (ii) local distribution biases cause misaligned representations that hinder consistent cross-view semantics, and (iii) large video architectures incur prohibitive communication overhead. To address these issues, we propose FedCVU, a federated framework with three components: VS-Norm, which preserves normalization parameters to handle view-specific statistics; CV-Align, a lightweight contrastive regularization module to improve cross-view representation alignment; and SLA, a selective layer aggregation strategy that reduces communication without sacrificing accuracy. Extensive experiments on action understanding and person re-identification tasks under a cross-view protocol demonstrate that FedCVU consistently boosts unseen-view accuracy while maintaining strong seen-view performance, outperforming state-of-the-art FL baselines and showing robustness to domain heterogeneity and communication constraints.

Jing Wang, Ao Ma, Ke Cao et al.

Recent rapid advancements in text-to-video (T2V) generation, such as SoRA and Kling, have shown great potential for building world simulators. However, current T2V models struggle to grasp abstract physical principles and generate videos that adhere to physical laws. This challenge arises primarily from a lack of clear guidance on physical information due to a significant gap between abstract physical principles and generation models. To this end, we introduce the World Simulator Assistant (WISA), an effective framework for decomposing and incorporating physical principles into T2V models. Specifically, WISA decomposes physical principles into textual physical descriptions, qualitative physical categories, and quantitative physical properties. To effectively embed these physical attributes into the generation process, WISA incorporates several key designs, including Mixture-of-Physical-Experts Attention (MoPA) and a Physical Classifier, enhancing the model's physics awareness. Furthermore, most existing datasets feature videos where physical phenomena are either weakly represented or entangled with multiple co-occurring processes, limiting their suitability as dedicated resources for learning explicit physical principles. We propose a novel video dataset, WISA-32K, collected based on qualitative physical categories. It consists of 32,000 videos, representing 17 physical laws across three domains of physics: dynamics, thermodynamics, and optics. Experimental results demonstrate that WISA can effectively enhance the compatibility of T2V models with real-world physical laws, achieving a considerable improvement on the VideoPhy benchmark. The visual exhibitions of WISA and WISA-32K are available in the https://360cvgroup.github.io/WISA/.

Yihua Shao, Haojin He, Sijie Li et al.

Video Anomaly Detection~(VAD) focuses on identifying anomalies within videos. Supervised methods require an amount of in-domain training data and often struggle to generalize to unseen anomalies. In contrast, training-free methods leverage the intrinsic world knowledge of large language models (LLMs) to detect anomalies but face challenges in localizing fine-grained visual transitions and diverse events. Therefore, we propose EventVAD, an event-aware video anomaly detection framework that combines tailored dynamic graph architectures and multimodal LLMs through temporal-event reasoning. Specifically, EventVAD first employs dynamic spatiotemporal graph modeling with time-decay constraints to capture event-aware video features. Then, it performs adaptive noise filtering and uses signal ratio thresholding to detect event boundaries via unsupervised statistical features. The statistical boundary detection module reduces the complexity of processing long videos for MLLMs and improves their temporal reasoning through event consistency. Finally, it utilizes a hierarchical prompting strategy to guide MLLMs in performing reasoning before determining final decisions. We conducted extensive experiments on the UCF-Crime and XD-Violence datasets. The results demonstrate that EventVAD with a 7B MLLM achieves state-of-the-art (SOTA) in training-free settings, outperforming strong baselines that use 7B or larger MLLMs.

Ke Cao, Jing Wang, Ao Ma et al.

The Diffusion Transformer plays a pivotal role in advancing text-to-image and text-to-video generation, owing primarily to its inherent scalability. However, existing controlled diffusion transformer methods incur significant parameter and computational overheads and suffer from inefficient resource allocation due to their failure to account for the varying relevance of control information across different transformer layers. To address this, we propose the Relevance-Guided Efficient Controllable Generation framework, RelaCtrl, enabling efficient and resource-optimized integration of control signals into the Diffusion Transformer. First, we evaluate the relevance of each layer in the Diffusion Transformer to the control information by assessing the "ControlNet Relevance Score"-i.e., the impact of skipping each control layer on both the quality of generation and the control effectiveness during inference. Based on the strength of the relevance, we then tailor the positioning, parameter scale, and modeling capacity of the control layers to reduce unnecessary parameters and redundant computations. Additionally, to further improve efficiency, we replace the self-attention and FFN in the commonly used copy block with the carefully designed Two-Dimensional Shuffle Mixer (TDSM), enabling efficient implementation of both the token mixer and channel mixer. Both qualitative and quantitative experimental results demonstrate that our approach achieves superior performance with only 15% of the parameters and computational complexity compared to PixArt-delta.

Zhanjie Zhang, Ao Ma, Ke Cao et al.

Ultra-high quality artistic style transfer refers to repainting an ultra-high quality content image using the style information learned from the style image. Existing artistic style transfer methods can be categorized into style reconstruction-based and content-style disentanglement-based style transfer approaches. Although these methods can generate some artistic stylized images, they still exhibit obvious artifacts and disharmonious patterns, which hinder their ability to produce ultra-high quality artistic stylized images. To address these issues, we propose a novel artistic image style transfer method, U-StyDiT, which is built on transformer-based diffusion (DiT) and learns content-style disentanglement, generating ultra-high quality artistic stylized images. Specifically, we first design a Multi-view Style Modulator (MSM) to learn style information from a style image from local and global perspectives, conditioning U-StyDiT to generate stylized images with the learned style information. Then, we introduce a StyDiT Block to learn content and style conditions simultaneously from a style image. Additionally, we propose an ultra-high quality artistic image dataset, Aes4M, comprising 10 categories, each containing 400,000 style images. This dataset effectively solves the problem that the existing style transfer methods cannot produce high-quality artistic stylized images due to the size of the dataset and the quality of the images in the dataset. Finally, the extensive qualitative and quantitative experiments validate that our U-StyDiT can create higher quality stylized images compared to state-of-the-art artistic style transfer methods. To our knowledge, our proposed method is the first to address the generation of ultra-high quality stylized images using transformer-based diffusion.

Zhou Hong, Rongsheng Hu, Yicheng Di et al.

Image style transfer aims to integrate the visual patterns of a specific artistic style into a content image while preserving its content structure. Existing methods mainly rely on the generative adversarial network (GAN) or stable diffusion (SD). GAN-based approaches using CNNs or Transformers struggle to jointly capture local and global dependencies, leading to artifacts and disharmonious patterns. SD-based methods reduce such issues but often fail to preserve content structures and suffer from slow inference. To address these issues, we revisit GAN and propose a mamba-based generator, termed as StyMam, to produce high-quality stylized images without introducing artifacts and disharmonious patterns. Specifically, we introduce a mamba-based generator with a residual dual-path strip scanning mechanism and a channel-reweighted spatial attention module. The former efficiently captures local texture features, while the latter models global dependencies. Finally, extensive qualitative and quantitative experiments demonstrate that the proposed method outperforms state-of-the-art algorithms in both quality and speed.

Ao Ma, Jiasong Feng, Ke Cao et al.

Storytelling tasks involving generating consistent subjects have gained significant attention recently. However, existing methods, whether training-free or training-based, continue to face challenges in maintaining subject consistency due to the lack of fine-grained guidance and inter-frame interaction. Additionally, the scarcity of high-quality data in this field makes it difficult to precisely control storytelling tasks, including the subject's position, appearance, clothing, expression, and posture, thereby hindering further advancements. In this paper, we demonstrate that layout conditions, such as the subject's position and detailed attributes, effectively facilitate fine-grained interactions between frames. This not only strengthens the consistency of the generated frame sequence but also allows for precise control over the subject's position, appearance, and other key details. Building on this, we introduce an advanced storytelling task: Layout-Togglable Storytelling, which enables precise subject control by incorporating layout conditions. To address the lack of high-quality datasets with layout annotations for this task, we develop Lay2Story-1M, which contains over 1 million 720p and higher-resolution images, processed from approximately 11,300 hours of cartoon videos. Building on Lay2Story-1M, we create Lay2Story-Bench, a benchmark with 3,000 prompts designed to evaluate the performance of different methods on this task. Furthermore, we propose Lay2Story, a robust framework based on the Diffusion Transformers (DiTs) architecture for Layout-Togglable Storytelling tasks. Through both qualitative and quantitative experiments, we find that our method outperforms the previous state-of-the-art (SOTA) techniques, achieving the best results in terms of consistency, semantic correlation, and aesthetic quality.

Run Ling, Wenji Wang, Yuting Liu et al.

Personalized image generation is crucial for improving the user experience, as it renders reference images into preferred ones according to user visual preferences. Although effective, existing methods face two main issues. First, existing methods treat all items in the user historical sequence equally when extracting user preferences, overlooking the varying semantic similarities between historical items and the reference item. Disproportionately high weights for low-similarity items distort users' visual preferences for the reference item. Second, existing methods heavily rely on consistency between generated and reference images to optimize the generation, which leads to underfitting user preferences and hinders personalization. To address these issues, we propose Retrieval Augment Personalized Image GenerAtion guided by Recommendation (RAGAR). Our approach uses a retrieval mechanism to assign different weights to historical items according to their similarities to the reference item, thereby extracting more refined users' visual preferences for the reference item. Then we introduce a novel rank task based on the multi-modal ranking model to optimize the personalization of the generated images instead of forcing depend on consistency. Extensive experiments and human evaluations on three real-world datasets demonstrate that RAGAR achieves significant improvements in both personalization and semantic metrics compared to five baselines.

Yihua Shao, Xiaofeng Lin, Xinwei Long et al.

Enabling multi-task adaptation in pre-trained Low-Rank Adaptation (LoRA) models is crucial for enhancing their generalization capabilities. Most existing pre-trained LoRA fusion methods decompose weight matrices, sharing similar parameters while merging divergent ones. However, this paradigm inevitably induces inter-weight conflicts and leads to catastrophic domain forgetting. While incremental learning enables adaptation to multiple tasks, it struggles to achieve generalization in few-shot scenarios. Consequently, when the weight data follows a long-tailed distribution, it can lead to forgetting in the fused weights. To address this issue, we propose In-Context Meta LoRA Fusion (ICM-Fusion), a novel framework that synergizes meta-learning with in-context adaptation. The key innovation lies in our task vector arithmetic, which dynamically balances conflicting optimization directions across domains through learned manifold projections. ICM-Fusion obtains the optimal task vector orientation for the fused model in the latent space by adjusting the orientation of the task vectors. Subsequently, the fused LoRA is reconstructed by a self-designed Fusion VAE (F-VAE) to realize multi-task LoRA generation. We have conducted extensive experiments on visual and linguistic tasks, and the experimental results demonstrate that ICM-Fusion can be adapted to a wide range of architectural models and applied to various tasks. Compared to the current pre-trained LoRA fusion method, ICM-Fusion fused LoRA can significantly reduce the multi-tasking loss and can even achieve task enhancement in few-shot scenarios.

Jing Wang, Jiajun Liang, Jie Liu et al.

Recently, GRPO-based reinforcement learning has shown remarkable progress in optimizing flow-matching models, effectively improving their alignment with task-specific rewards. Within these frameworks, the policy update relies on importance-ratio clipping to constrain overconfident positive and negative gradients. However, in practice, we observe a systematic shift in the importance-ratio distribution-its mean falls below 1 and its variance differs substantially across timesteps. This left-shifted and inconsistent distribution prevents positive-advantage samples from entering the clipped region, causing the mechanism to fail in constraining overconfident positive updates. As a result, the policy model inevitably enters an implicit over-optimization stage-while the proxy reward continues to increase, essential metrics such as image quality and text-prompt alignment deteriorate sharply, ultimately making the learned policy impractical for real-world use. To address this issue, we introduce GRPO-Guard, a simple yet effective enhancement to existing GRPO frameworks. Our method incorporates ratio normalization, which restores a balanced and step-consistent importance ratio, ensuring that PPO clipping properly constrains harmful updates across denoising timesteps. In addition, a gradient reweighting strategy equalizes policy gradients over noise conditions, preventing excessive updates from particular timestep regions. Together, these designs act as a regulated clipping mechanism, stabilizing optimization and substantially mitigating implicit over-optimization without relying on heavy KL regularization. Extensive experiments on multiple diffusion backbones (e.g., SD3.5M, Flux.1-dev) and diverse proxy tasks demonstrate that GRPO-Guard significantly reduces over-optimization while maintaining or even improving generation quality.