Bin Xia

Bin Xia, Yulun Zhang, Shiyin Wang et al. · eth-zurich

Diffusion model (DM) has achieved SOTA performance by modeling the image synthesis process into a sequential application of a denoising network. However, different from image synthesis, image restoration (IR) has a strong constraint to generate results in accordance with ground-truth. Thus, for IR, traditional DMs running massive iterations on a large model to estimate whole images or feature maps is inefficient. To address this issue, we propose an efficient DM for IR (DiffIR), which consists of a compact IR prior extraction network (CPEN), dynamic IR transformer (DIRformer), and denoising network. Specifically, DiffIR has two training stages: pretraining and training DM. In pretraining, we input ground-truth images into CPEN$_{S1}$ to capture a compact IR prior representation (IPR) to guide DIRformer. In the second stage, we train the DM to directly estimate the same IRP as pretrained CPEN$_{S1}$ only using LQ images. We observe that since the IPR is only a compact vector, DiffIR can use fewer iterations than traditional DM to obtain accurate estimations and generate more stable and realistic results. Since the iterations are few, our DiffIR can adopt a joint optimization of CPEN$_{S2}$, DIRformer, and denoising network, which can further reduce the estimation error influence. We conduct extensive experiments on several IR tasks and achieve SOTA performance while consuming less computational costs. Code is available at \url{https://github.com/Zj-BinXia/DiffIR}.

Bin Xia, Yulun Zhang, Yitong Wang et al. · eth-zurich

Blind image super-resolution (Blind-SR) aims to recover a high-resolution (HR) image from its corresponding low-resolution (LR) input image with unknown degradations. Most of the existing works design an explicit degradation estimator for each degradation to guide SR. However, it is infeasible to provide concrete labels of multiple degradation combinations (e.g., blur, noise, jpeg compression) to supervise the degradation estimator training. In addition, these special designs for certain degradation, such as blur, impedes the models from being generalized to handle different degradations. To this end, it is necessary to design an implicit degradation estimator that can extract discriminative degradation representation for all degradations without relying on the supervision of degradation ground-truth. In this paper, we propose a Knowledge Distillation based Blind-SR network (KDSR). It consists of a knowledge distillation based implicit degradation estimator network (KD-IDE) and an efficient SR network. To learn the KDSR model, we first train a teacher network: KD-IDE$_{T}$. It takes paired HR and LR patches as inputs and is optimized with the SR network jointly. Then, we further train a student network KD-IDE$_{S}$, which only takes LR images as input and learns to extract the same implicit degradation representation (IDR) as KD-IDE$_{T}$. In addition, to fully use extracted IDR, we design a simple, strong, and efficient IDR based dynamic convolution residual block (IDR-DCRB) to build an SR network. We conduct extensive experiments under classic and real-world degradation settings. The results show that KDSR achieves SOTA performance and can generalize to various degradation processes. The source codes and pre-trained models will be released.

Bin Xia, Yulun Zhang, Yitong Wang et al. · eth-zurich

Lighter and faster image restoration (IR) models are crucial for the deployment on resource-limited devices. Binary neural network (BNN), one of the most promising model compression methods, can dramatically reduce the computations and parameters of full-precision convolutional neural networks (CNN). However, there are different properties between BNN and full-precision CNN, and we can hardly use the experience of designing CNN to develop BNN. In this study, we reconsider components in binary convolution, such as residual connection, BatchNorm, activation function, and structure, for IR tasks. We conduct systematic analyses to explain each component's role in binary convolution and discuss the pitfalls. Specifically, we find that residual connection can reduce the information loss caused by binarization; BatchNorm can solve the value range gap between residual connection and binary convolution; The position of the activation function dramatically affects the performance of BNN. Based on our findings and analyses, we design a simple yet efficient basic binary convolution unit (BBCU). Furthermore, we divide IR networks into four parts and specially design variants of BBCU for each part to explore the benefit of binarizing these parts. We conduct experiments on different IR tasks, and our BBCU significantly outperforms other BNNs and lightweight models, which shows that BBCU can serve as a basic unit for binarized IR networks. All codes and models will be released.

Bin Xia, Jingwen He, Yulun Zhang et al. · eth-zurich

The high computational costs of video super-resolution (VSR) models hinder their deployment on resource-limited devices, (e.g., smartphones and drones). Existing VSR models contain considerable redundant filters, which drag down the inference efficiency. To prune these unimportant filters, we develop a structured pruning scheme called Structured Sparsity Learning (SSL) according to the properties of VSR. In SSL, we design pruning schemes for several key components in VSR models, including residual blocks, recurrent networks, and upsampling networks. Specifically, we develop a Residual Sparsity Connection (RSC) scheme for residual blocks of recurrent networks to liberate pruning restrictions and preserve the restoration information. For upsampling networks, we design a pixel-shuffle pruning scheme to guarantee the accuracy of feature channel-space conversion. In addition, we observe that pruning error would be amplified as the hidden states propagate along with recurrent networks. To alleviate the issue, we design Temporal Finetuning (TF). Extensive experiments show that SSL can significantly outperform recent methods quantitatively and qualitatively.

Bin Xia, Yulun Zhang, Shiyin Wang et al. · eth-zurich

The Diffusion Model (DM) has emerged as the SOTA approach for image synthesis. However, the existing DM cannot perform well on some image-to-image translation (I2I) tasks. Different from image synthesis, some I2I tasks, such as super-resolution, require generating results in accordance with GT images. Traditional DMs for image synthesis require extensive iterations and large denoising models to estimate entire images, which gives their strong generative ability but also leads to artifacts and inefficiency for I2I. To tackle this challenge, we propose a simple, efficient, and powerful DM framework for I2I, called DiffI2I. Specifically, DiffI2I comprises three key components: a compact I2I prior extraction network (CPEN), a dynamic I2I transformer (DI2Iformer), and a denoising network. We train DiffI2I in two stages: pretraining and DM training. For pretraining, GT and input images are fed into CPEN$_{S1}$ to capture a compact I2I prior representation (IPR) guiding DI2Iformer. In the second stage, the DM is trained to only use the input images to estimate the same IRP as CPEN$_{S1}$. Compared to traditional DMs, the compact IPR enables DiffI2I to obtain more accurate outcomes and employ a lighter denoising network and fewer iterations. Through extensive experiments on various I2I tasks, we demonstrate that DiffI2I achieves SOTA performance while significantly reducing computational burdens.

Bin Xia, Yapeng Tian, Yulun Zhang et al. · eth-zurich

The most of CNN based super-resolution (SR) methods assume that the degradation is known (\eg, bicubic). These methods will suffer a severe performance drop when the degradation is different from their assumption. Therefore, some approaches attempt to train SR networks with the complex combination of multiple degradations to cover the real degradation space. To adapt to multiple unknown degradations, introducing an explicit degradation estimator can actually facilitate SR performance. However, previous explicit degradation estimation methods usually predict Gaussian blur with the supervision of groundtruth blur kernels, and estimation errors may lead to SR failure. Thus, it is necessary to design a method that can extract implicit discriminative degradation representation. To this end, we propose a Meta-Learning based Region Degradation Aware SR Network (MRDA), including Meta-Learning Network (MLN), Degradation Extraction Network (DEN), and Region Degradation Aware SR Network (RDAN). To handle the lack of groundtruth degradation, we use the MLN to rapidly adapt to the specific complex degradation after several iterations and extract implicit degradation information. Subsequently, a teacher network MRDA$_{T}$ is designed to further utilize the degradation information extracted by MLN for SR. However, MLN requires iterating on paired low-resolution (LR) and corresponding high-resolution (HR) images, which is unavailable in the inference phase. Therefore, we adopt knowledge distillation (KD) to make the student network learn to directly extract the same implicit degradation representation (IDR) as the teacher from LR images.

Yuan Shi, Bin Xia, Rui Zhu et al.

Color-guided depth map super-resolution (CDSR) improve the spatial resolution of a low-quality depth map with the corresponding high-quality color map, benefiting various applications such as 3D reconstruction, virtual reality, and augmented reality. While conventional CDSR methods typically rely on convolutional neural networks or transformers, diffusion models (DMs) have demonstrated notable effectiveness in high-level vision tasks. In this work, we present a novel CDSR paradigm that utilizes a diffusion model within the latent space to generate guidance for depth map super-resolution. The proposed method comprises a guidance generation network (GGN), a depth map super-resolution network (DSRN), and a guidance recovery network (GRN). The GGN is specifically designed to generate the guidance while managing its compactness. Additionally, we integrate a simple but effective feature fusion module and a transformer-style feature extraction module into the DSRN, enabling it to leverage guided priors in the extraction, fusion, and reconstruction of multi-model images. Taking into account both accuracy and efficiency, our proposed method has shown superior performance in extensive experiments when compared to state-of-the-art methods. Our codes will be made available at https://github.com/shiyuan7/DSR-Diff.

Bin Xia, John H. Wise

We investigate conditional diffusion modeling for three-dimensional 21 cm lightcone emulation, focusing on cubes with a sky-plane size of $64\times64$ and a line-of-sight depth up to 1024 cells. Relative to earlier 2D studies, the 3D setting is substantially harder because memory limits enforce very small micro-batches while the underlying voxel distribution is highly skewed and long tailed. We perform controlled comparisons across preprocessing choices, dynamic-range compression settings, architecture depth, and training duration using $25{,}600$ training lightcones and validation ensembles at fixed parameter points. For validation, each reference parameter point contains 800 21cmFAST realizations with independent initial conditions, and we use 800 samples per model and per reference set for the reported ensemble comparisons. We evaluate generated lightcones with complementary diagnostics in both image and summary-statistic spaces: brightness-temperature slices, the global signal, the power spectrum, and reduced scattering coefficients. Across the tested configurations, preprocessing is the dominant factor governing stable training and the resulting physical fidelity. Among the configurations explored here, Yeo-Johnson preprocessing combined with moderate amplitude compression gives the most consistently favorable trade-off, with the strongest quantitative support coming from rankings based on the standard-deviation-normalized mean absolute error ($\mathrm{MAE}_{\rm std}$) of the global signal and qualitatively compatible behavior in the complementary diagnostics. At the same time, visually plausible 3D samples still retain measurable biases in two-point and higher-order statistics. We therefore view the present work as a simulation-level baseline for three-dimensional 21 cm emulation and for future studies that incorporate more realistic observational effects.

Juan Wang, Bin Xia

Weakly supervised image segmentation approaches in the literature usually achieve high segmentation performance using tight bounding box supervision and decrease the performance greatly when supervised by loose bounding boxes. However, compared with loose bounding box, it is much more difficult to acquire tight bounding box due to its strict requirements on the precise locations of the four sides of the box. To resolve this issue, this study investigates whether it is possible to maintain good segmentation performance when loose bounding boxes are used as supervision. For this purpose, this work extends our previous parallel transformation based multiple instance learning (MIL) for tight bounding box supervision by integrating an MIL strategy based on polar transformation to assist image segmentation. The proposed polar transformation based MIL formulation works for both tight and loose bounding boxes, in which a positive bag is defined as pixels in a polar line of a bounding box with one endpoint located inside the object enclosed by the box and the other endpoint located at one of the four sides of the box. Moreover, a weighted smooth maximum approximation is introduced to incorporate the observation that pixels closer to the origin of the polar transformation are more likely to belong to the object in the box. The proposed approach was evaluated on two public datasets using dice coefficient when bounding boxes at different precision levels were considered in the experiments. The results demonstrate that the proposed approach achieves state-of-the-art performance for bounding boxes at all precision levels and is robust to mild and moderate errors in the loose bounding box annotations. The codes are available at \url{https://github.com/wangjuan313/wsis-beyond-tightBB}.

Yucheng Hang, Bin Xia, Wenming Yang et al.

Image harmonization aims to achieve visual consistency in composite images by adapting a foreground to make it compatible with a background. However, existing methods always only use the real image as the positive sample to guide the training, and at most introduce the corresponding composite image as a single negative sample for an auxiliary constraint, which leads to limited distortion knowledge, and further causes a too large solution space, making the generated harmonized image distorted. Besides, none of them jointly constrain from the foreground self-style and foreground-background style consistency, which exacerbates this problem. Moreover, recent region-aware adaptive instance normalization achieves great success but only considers the global background feature distribution, making the aligned foreground feature distribution biased. To address these issues, we propose a self-consistent style contrastive learning scheme (SCS-Co). By dynamically generating multiple negative samples, our SCS-Co can learn more distortion knowledge and well regularize the generated harmonized image in the style representation space from two aspects of the foreground self-style and foreground-background style consistency, leading to a more photorealistic visual result. In addition, we propose a background-attentional adaptive instance normalization (BAIN) to achieve an attention-weighted background feature distribution according to the foreground-background feature similarity. Experiments demonstrate the superiority of our method over other state-of-the-art methods in both quantitative comparison and visual analysis.

Juan Wang, Bin Xia

This study investigates weakly supervised image segmentation using loose bounding box supervision. It presents a multiple instance learning strategy based on polar transformation to assist image segmentation when loose bounding boxes are employed as supervision. In this strategy, weighted smooth maximum approximation is introduced to incorporate the observation that pixels closer to the origin of the polar transformation are more likely to belong to the object in the bounding box. The proposed approach was evaluated on a public medical dataset using Dice coefficient. The results demonstrate its superior performance. The codes are available at \url{https://github.com/wangjuan313/wsis-polartransform}.

Jiehui Huang, Yuechen Zhang, Xu He et al.

Recent video generation models demonstrate impressive synthesis capabilities but remain limited by single-modality conditioning, constraining their holistic world understanding. This stems from insufficient cross-modal interaction and limited modal diversity for comprehensive world knowledge representation. To address these limitations, we introduce UnityVideo, a unified framework for world-aware video generation that jointly learns across multiple modalities (segmentation masks, human skeletons, DensePose, optical flow, and depth maps) and training paradigms. Our approach features two core components: (1) dynamic noising to unify heterogeneous training paradigms, and (2) a modality switcher with an in-context learner that enables unified processing via modular parameters and contextual learning. We contribute a large-scale unified dataset with 1.3M samples. Through joint optimization, UnityVideo accelerates convergence and significantly enhances zero-shot generalization to unseen data. We demonstrate that UnityVideo achieves superior video quality, consistency, and improved alignment with physical world constraints. Code and data can be found at: https://github.com/dvlab-research/UnityVideo

Zhengchao Huang, Bin Xia, Zicheng Lin et al.

The rapid advancement of deepfake technologies has sparked widespread public concern, particularly as face forgery poses a serious threat to public information security. However, the unknown and diverse forgery techniques, varied facial features and complex environmental factors pose significant challenges for face forgery analysis. Existing datasets lack descriptive annotations of these aspects, making it difficult for models to distinguish between real and forged faces using only visual information amid various confounding factors. In addition, existing methods fail to yield user-friendly and explainable results, hindering the understanding of the model's decision-making process. To address these challenges, we introduce a novel Open-World Face Forgery Analysis VQA (OW-FFA-VQA) task and its corresponding benchmark. To tackle this task, we first establish a dataset featuring a diverse collection of real and forged face images with essential descriptions and reliable forgery reasoning. Based on this dataset, we introduce FFAA: Face Forgery Analysis Assistant, consisting of a fine-tuned Multimodal Large Language Model (MLLM) and Multi-answer Intelligent Decision System (MIDS). By integrating hypothetical prompts with MIDS, the impact of fuzzy classification boundaries is effectively mitigated, enhancing model robustness. Extensive experiments demonstrate that our method not only provides user-friendly and explainable results but also significantly boosts accuracy and robustness compared to previous methods.

Jialiang Yang, Bin Xia, Ruihang Chu et al.

Rapid advances in audio-video (AV) generation have enabled high-fidelity synthesis with synchronized sound, particularly for human-related scenarios involving speech and interactions. Yet evaluation for AV generation remains at an early stage, with only a few coarse-grained benchmarks for human-related scenarios and relying on limited preset evaluations with generic multimodal LLMs, leading to inaccurate assessments of model capabilities. To address these issues, we introduce AVBench, a fully automated benchmark tailored for human-centric AV generation. AVBench is built on two key designs for comprehensive and accurate evaluation: (i) Human-centric and fine-grained metrics. AVBench integrates ten evaluation dimensions designed for human-centered real-world scenarios, covering visual quality, audio quality, and multi-level consistency across modalities. These practical metrics capture human-related details that existing benchmarks often overlook. (ii) Specialized evaluators via preference learning. To address the lack of specialized training data, we construct large-scale supervision by transforming real-world videos into diverse training pairs with controlled perturbations. After fine-tuning on this high-quality dataset, the evaluators learn to reliably detect subtle cross-modal inconsistencies. Crucially, instead of producing discrete textual judgment, AVBench derives continuous evaluation scores from the model's prediction confidence on binary decisions. This probabilistic scoring mechanism enables a more reliable assessment than traditional VQA-style evaluation and aligns closely with human judgment. Taken together, AVBench offers automated evaluation for AV generation, demonstrates strong potential for data filtering, and serves as a differentiable reward signal for Reinforcement Learning from Human Feedback (RLHF).

Bin Xia, Shiyin Wang, Yingfan Tao et al.

In this paper, we introduce a Multimodal Large Language Model-based Generation Assistant (LLMGA), leveraging the vast reservoir of knowledge and proficiency in reasoning, comprehension, and response inherent in Large Language Models (LLMs) to assist users in image generation and editing. Diverging from existing approaches where Multimodal Large Language Models (MLLMs) generate fixed-size embeddings to control Stable Diffusion (SD), our LLMGA provides a detailed language generation prompt for precise control over SD. This not only augments LLM context understanding but also reduces noise in generation prompts, yields images with more intricate and precise content, and elevates the interpretability of the network. To this end, we curate a comprehensive dataset comprising prompt refinement, similar image generation, inpainting \& outpainting, and instruction-based editing. Moreover, we propose a two-stage training scheme. In the first stage, we train the MLLM to grasp the properties of image generation and editing, enabling it to generate detailed prompts. In the second stage, we optimize SD to align with the MLLM's generation prompts. Additionally, we propose a reference-based restoration network to alleviate texture, brightness, and contrast disparities between generated and preserved regions during inpainting and outpainting. Extensive results show that LLMGA has promising generation and editing capabilities and can enable more flexible and expansive applications in an interactive manner.

Ruihang Chu, Yefei He, Zhekai Chen et al.

We present Wan-Move, a simple and scalable framework that brings motion control to video generative models. Existing motion-controllable methods typically suffer from coarse control granularity and limited scalability, leaving their outputs insufficient for practical use. We narrow this gap by achieving precise and high-quality motion control. Our core idea is to directly make the original condition features motion-aware for guiding video synthesis. To this end, we first represent object motions with dense point trajectories, allowing fine-grained control over the scene. We then project these trajectories into latent space and propagate the first frame's features along each trajectory, producing an aligned spatiotemporal feature map that tells how each scene element should move. This feature map serves as the updated latent condition, which is naturally integrated into the off-the-shelf image-to-video model, e.g., Wan-I2V-14B, as motion guidance without any architecture change. It removes the need for auxiliary motion encoders and makes fine-tuning base models easily scalable. Through scaled training, Wan-Move generates 5-second, 480p videos whose motion controllability rivals Kling 1.5 Pro's commercial Motion Brush, as indicated by user studies. To support comprehensive evaluation, we further design MoveBench, a rigorously curated benchmark featuring diverse content categories and hybrid-verified annotations. It is distinguished by larger data volume, longer video durations, and high-quality motion annotations. Extensive experiments on MoveBench and the public dataset consistently show Wan-Move's superior motion quality. Code, models, and benchmark data are made publicly available.

Jintong Hu, Bin Xia, Bin Chen et al.

Implicit neural representations (INRs) have significantly advanced the field of arbitrary-scale super-resolution (ASSR) of images. Most existing INR-based ASSR networks first extract features from the given low-resolution image using an encoder, and then render the super-resolved result via a multi-layer perceptron decoder. Although these approaches have shown promising results, their performance is constrained by the limited representation ability of discrete latent codes in the encoded features. In this paper, we propose a novel ASSR method named GaussianSR that overcomes this limitation through 2D Gaussian Splatting (2DGS). Unlike traditional methods that treat pixels as discrete points, GaussianSR represents each pixel as a continuous Gaussian field. The encoded features are simultaneously refined and upsampled by rendering the mutually stacked Gaussian fields. As a result, long-range dependencies are established to enhance representation ability. In addition, a classifier is developed to dynamically assign Gaussian kernels to all pixels to further improve flexibility. All components of GaussianSR (i.e., encoder, classifier, Gaussian kernels, and decoder) are jointly learned end-to-end. Experiments demonstrate that GaussianSR achieves superior ASSR performance with fewer parameters than existing methods while enjoying interpretable and content-aware feature aggregations.

Chenhan Cao, Xiaoyu Fang, Bingqing Luo et al.

The ticket automation provides crucial support for the normal operation of IT software systems. An essential task of ticket automation is to assign experts to solve upcoming tickets. However, facing thousands of tickets, inappropriate assignments will make tickets transfer frequently among experts, which causes time delays and wasted resources. Effectively and efficiently finding an appropriate expert in fewer steps is vital to ticket automation. In this paper, we proposed a sequence to sequence based translation model combined with a recurrent recommendation network to recommend appropriate experts for tickets. The sequence to sequence model transforms the ticket description into the corresponding resolution for capturing the potential and useful features of representing tickets. The recurrent recommendation network recommends the appropriate expert based on the assumption that the previous expert in the recommendation sequence cannot solve the expert. To evaluate the performance, we conducted experiments to compare several baselines with SSR-TA on two real-world datasets, and the experimental results show that our proposed model outperforms the baselines. The comparative experiment results also show that SSR-TA has a better performance of expert recommendations for user-generated tickets.

Yuechen Zhang, Yaoyang Liu, Bin Xia et al.

We present Magic Mirror, a framework for generating identity-preserved videos with cinematic-level quality and dynamic motion. While recent advances in video diffusion models have shown impressive capabilities in text-to-video generation, maintaining consistent identity while producing natural motion remains challenging. Previous methods either require person-specific fine-tuning or struggle to balance identity preservation with motion diversity. Built upon Video Diffusion Transformers, our method introduces three key components: (1) a dual-branch facial feature extractor that captures both identity and structural features, (2) a lightweight cross-modal adapter with Conditioned Adaptive Normalization for efficient identity integration, and (3) a two-stage training strategy combining synthetic identity pairs with video data. Extensive experiments demonstrate that Magic Mirror effectively balances identity consistency with natural motion, outperforming existing methods across multiple metrics while requiring minimal parameters added. The code and model will be made publicly available at: https://github.com/dvlab-research/MagicMirror/

Jingyao Li, Pengguang Chen, Bin Xia et al.

Large Language Models (LLMs) have showcased impressive capabilities in handling straightforward programming tasks. However, their performance tends to falter when confronted with more challenging programming problems. We observe that conventional models often generate solutions as monolithic code blocks, restricting their effectiveness in tackling intricate questions. To overcome this limitation, we present Module-of-Thought Coder (MoTCoder). We introduce a framework for MoT instruction tuning, designed to promote the decomposition of tasks into logical sub-tasks and sub-modules. Our investigations reveal that, through the cultivation and utilization of sub-modules, MoTCoder significantly improves both the modularity and correctness of the generated solutions, leading to substantial pass@1 improvements of 5.9% on APPS and 5.8% on CodeContests. MoTCoder also achieved significant improvements in self-correction capabilities, surpassing the current SOTA by 3.3%. Additionally, we provide an analysis of between problem complexity and optimal module decomposition and evaluate the maintainability index, confirming that the code generated by MoTCoder is easier to understand and modify, which can be beneficial for long-term code maintenance and evolution. Our codes are available at https://github.com/dvlab-research/MoTCoder.

Junjia Huang, Binbin Yang, Pengxiang Yan et al.

Storyboard synthesis plays a crucial role in visual storytelling, aiming to generate coherent shot sequences that visually narrate cinematic events with consistent characters, scenes, and transitions. However, existing approaches are mostly adapted from text-to-image diffusion models, which struggle to maintain long-range temporal coherence, consistent character identities, and narrative flow across multiple shots. In this paper, we introduce DreamShot, a video generative model based storyboard framework that fully exploits powerful video diffusion priors for controllable multi-shot synthesis. DreamShot supports both Text-to-Shot and Reference-to-Shot generation, as well as story continuation conditioned on previous frames, enabling flexible and context-aware storyboard generation. By leveraging the spatial-temporal consistency inherent in video generative models, DreamShot produces visually and semantically coherent sequences with improved narrative fidelity and character continuity. Furthermore, DreamShot incorporates a multi-reference role conditioning module that accepts multiple character reference images and enforces identity alignment via a Role-Attention Consistency Loss, explicitly constraining attention between reference and generated roles. Extensive experiments demonstrate that DreamShot achieves superior scene coherence, role consistency, and generation efficiency compared to state-of-the-art text-to-image storyboard models, establishing a new direction toward controllable video model-driven visual storytelling.

Bin Xia, Bohao Peng, Jiyang Liu et al.

Recently unified generation and editing models have achieved remarkable success with their impressive performance. These models rely mainly on text prompts for instruction-based editing and generation, but language often fails to capture users intended edit locations and fine-grained visual details. To this end, we propose two tasks: scribble-based editing and generation, that enables more flexible creation on graphical user interface (GUI) combining user textual, images, and freehand sketches. We introduce DreamOmni3, tackling two challenges: data creation and framework design. Our data synthesis pipeline includes two parts: scribble-based editing and generation. For scribble-based editing, we define four tasks: scribble and instruction-based editing, scribble and multimodal instruction-based editing, image fusion, and doodle editing. Based on DreamOmni2 dataset, we extract editable regions and overlay hand-drawn boxes, circles, doodles or cropped image to construct training data. For scribble-based generation, we define three tasks: scribble and instruction-based generation, scribble and multimodal instruction-based generation, and doodle generation, following similar data creation pipelines. For the framework, instead of using binary masks, which struggle with complex edits involving multiple scribbles, images, and instructions, we propose a joint input scheme that feeds both the original and scribbled source images into the model, using different colors to distinguish regions and simplify processing. By applying the same index and position encodings to both images, the model can precisely localize scribbled regions while maintaining accurate editing. Finally, we establish comprehensive benchmarks for these tasks to promote further research. Experimental results demonstrate that DreamOmni3 achieves outstanding performance, and models and code will be publicly released.

Yuechen Zhang, Jinbo Xing, Bin Xia et al.

Despite the remarkable generation quality of video Diffusion Transformer (DiT) models, their practical deployment is severely hindered by extensive computational requirements. This inefficiency stems from two key challenges: the quadratic complexity of self-attention with respect to token length and the multi-step nature of diffusion models. To address these limitations, we present Jenga, a novel inference pipeline that combines dynamic attention carving with progressive resolution generation. Our approach leverages two key insights: (1) early denoising steps do not require high-resolution latents, and (2) later steps do not require dense attention. Jenga introduces a block-wise attention mechanism that dynamically selects relevant token interactions using 3D space-filling curves, alongside a progressive resolution strategy that gradually increases latent resolution during generation. Experimental results demonstrate that Jenga achieves substantial speedups across multiple state-of-the-art video diffusion models while maintaining comparable generation quality (8.83$\times$ speedup with 0.01\% performance drop on VBench). As a plug-and-play solution, Jenga enables practical, high-quality video generation on modern hardware by reducing inference time from minutes to seconds -- without requiring model retraining. Code: https://github.com/dvlab-research/Jenga

Chengyao Wang, Zhisheng Zhong, Bohao Peng et al.

We present MGM-Omni, a unified Omni LLM for omni-modal understanding and expressive, long-horizon speech generation. Unlike cascaded pipelines that isolate speech synthesis, MGM-Omni adopts a "brain-mouth" design with a dual-track, token-based architecture that cleanly decouples multimodal reasoning from real-time speech generation. This design enables efficient cross-modal interaction and low-latency, streaming speech generation. For understanding, a unified training strategy coupled with a dual audio encoder design enables long-form audio perception across diverse acoustic conditions. For generation, a chunk-based parallel decoding scheme narrows the text speech token-rate gap, accelerating inference and supporting streaming zero-shot voice cloning with stable timbre over extended durations. Compared to concurrent work, MGM-Omni achieves these capabilities with markedly data-efficient training. Extensive experiments demonstrate that MGM-Omni outperforms existing open source models in preserving timbre identity across extended sequences, producing natural and context-aware speech, and achieving superior long-form audio and omnimodal understanding. MGM-Omni establishes an efficient, end-to-end paradigm for omnimodal understanding and controllable, personalised long-horizon speech generation.

Yuan Shi, Bin Xia, Xiaoyu Jin et al.

Image restoration is a critical task in low-level computer vision, aiming to restore high-quality images from degraded inputs. Various models, such as convolutional neural networks (CNNs), generative adversarial networks (GANs), transformers, and diffusion models (DMs), have been employed to address this problem with significant impact. However, CNNs have limitations in capturing long-range dependencies. DMs require large prior models and computationally intensive denoising steps. Transformers have powerful modeling capabilities but face challenges due to quadratic complexity with input image size. To address these challenges, we propose VmambaIR, which introduces State Space Models (SSMs) with linear complexity into comprehensive image restoration tasks. We utilize a Unet architecture to stack our proposed Omni Selective Scan (OSS) blocks, consisting of an OSS module and an Efficient Feed-Forward Network (EFFN). Our proposed omni selective scan mechanism overcomes the unidirectional modeling limitation of SSMs by efficiently modeling image information flows in all six directions. Furthermore, we conducted a comprehensive evaluation of our VmambaIR across multiple image restoration tasks, including image deraining, single image super-resolution, and real-world image super-resolution. Extensive experimental results demonstrate that our proposed VmambaIR achieves state-of-the-art (SOTA) performance with much fewer computational resources and parameters. Our research highlights the potential of state space models as promising alternatives to the transformer and CNN architectures in serving as foundational frameworks for next-generation low-level visual tasks.

Zheng Chen, Yulun Zhang, Ding Liu et al.

Diffusion models (DMs) have recently been introduced in image deblurring and exhibited promising performance, particularly in terms of details reconstruction. However, the diffusion model requires a large number of inference iterations to recover the clean image from pure Gaussian noise, which consumes massive computational resources. Moreover, the distribution synthesized by the diffusion model is often misaligned with the target results, leading to restrictions in distortion-based metrics. To address the above issues, we propose the Hierarchical Integration Diffusion Model (HI-Diff), for realistic image deblurring. Specifically, we perform the DM in a highly compacted latent space to generate the prior feature for the deblurring process. The deblurring process is implemented by a regression-based method to obtain better distortion accuracy. Meanwhile, the highly compact latent space ensures the efficiency of the DM. Furthermore, we design the hierarchical integration module to fuse the prior into the regression-based model from multiple scales, enabling better generalization in complex blurry scenarios. Comprehensive experiments on synthetic and real-world blur datasets demonstrate that our HI-Diff outperforms state-of-the-art methods. Code and trained models are available at https://github.com/zhengchen1999/HI-Diff.

Juan Wang, Bin Xia

The cup-to-disc ratio (CDR) is one of the most significant indicator for glaucoma diagnosis. Different from the use of costly fully supervised learning formulation with pixel-wise annotations in the literature, this study investigates the feasibility of accurate CDR measurement in fundus images using only tight bounding box supervision. For this purpose, we develop a two-task network named as CDRNet for accurate CDR measurement, one for weakly supervised image segmentation, and the other for bounding-box regression. The weakly supervised image segmentation task is implemented based on generalized multiple instance learning formulation and smooth maximum approximation, and the bounding-box regression task outputs class-specific bounding box prediction in a single scale at the original image resolution. To get accurate bounding box prediction, a class-specific bounding-box normalizer and an expected intersection-over-union are proposed. In the experiments, the proposed approach was evaluated by a testing set with 1200 images using CDR error and $F_1$ score for CDR measurement and dice coefficient for image segmentation. A grader study was conducted to compare the performance of the proposed approach with those of individual graders. The experimental results indicate that the proposed approach outperforms the state-of-the-art performance obtained from the fully supervised image segmentation (FSIS) approach using pixel-wise annotation for CDR measurement. Its performance is also better than those of individual graders. In addition, the proposed approach gets performance close to the state-of-the-art obtained from FSIS and the performance of individual graders for optic cup and disc segmentation. The codes are available at \url{https://github.com/wangjuan313/CDRNet}.

Juan Wang, Bin Xia

This paper presents a weakly supervised image segmentation method that adopts tight bounding box annotations. It proposes generalized multiple instance learning (MIL) and smooth maximum approximation to integrate the bounding box tightness prior into the deep neural network in an end-to-end manner. In generalized MIL, positive bags are defined by parallel crossing lines with a set of different angles, and negative bags are defined as individual pixels outside of any bounding boxes. Two variants of smooth maximum approximation, i.e., $α$-softmax function and $α$-quasimax function, are exploited to conquer the numeral instability introduced by maximum function of bag prediction. The proposed approach was evaluated on two pubic medical datasets using Dice coefficient. The results demonstrate that it outperforms the state-of-the-art methods. The codes are available at \url{https://github.com/wangjuan313/wsis-boundingbox}.

Junyuan Xiao, Dingkang Liang, Xin Zhou et al.

Emerging multi-modal world models attempt to jointly generate videos across diverse modalities (e.g., RGB, depth, and mask), yet they fail to fully exploit the rich priors of existing foundation models. We propose $M^2$-REPA, the first representation alignment method tailored for multi-modal video generation. Our key insight is that foundation models trained on different modality spaces naturally capture distinct domain-specific priors, acting as complementary "experts." Specifically, we first decouple modality-specific features from the diffusion model's intermediate representations, then align each with its corresponding expert foundation model. To this end, we design two synergistic objectives: a multi-modal representation alignment loss that enforces feature-to-expert matching, and a modality-specific decoupling regularization that encourages complementarity across different modalities. This design enables joint optimization, fully exploiting priors from multiple foundation models. Extensive experiments demonstrate that our method significantly outperforms baselines in visual quality and long-term consistency.

Meng Chu, Xuan Billy Zhang, Kevin Qinghong Lin et al.

As AI systems move from generating text to accomplishing goals through sustained interaction, the ability to model environment dynamics becomes a central bottleneck. Agents that manipulate objects, navigate software, coordinate with others, or design experiments require predictive environment models, yet the term world model carries different meanings across research communities. We introduce a "levels x laws" taxonomy organized along two axes. The first defines three capability levels: L1 Predictor, which learns one-step local transition operators; L2 Simulator, which composes them into multi-step, action-conditioned rollouts that respect domain laws; and L3 Evolver, which autonomously revises its own model when predictions fail against new evidence. The second identifies four governing-law regimes: physical, digital, social, and scientific. These regimes determine what constraints a world model must satisfy and where it is most likely to fail. Using this framework, we synthesize over 400 works and summarize more than 100 representative systems spanning model-based reinforcement learning, video generation, web and GUI agents, multi-agent social simulation, and AI-driven scientific discovery. We analyze methods, failure modes, and evaluation practices across level-regime pairs, propose decision-centric evaluation principles and a minimal reproducible evaluation package, and outline architectural guidance, open problems, and governance challenges. The resulting roadmap connects previously isolated communities and charts a path from passive next-step prediction toward world models that can simulate, and ultimately reshape, the environments in which agents operate.

Bin Xia, Yuechen Zhang, Jingyao Li et al.

Currently, the success of large language models (LLMs) illustrates that a unified multitasking approach can significantly enhance model usability, streamline deployment, and foster synergistic benefits across different tasks. However, in computer vision, while text-to-image (T2I) models have significantly improved generation quality through scaling up, their framework design did not initially consider how to unify with downstream tasks, such as various types of editing. To address this, we introduce DreamOmni, a unified model for image generation and editing. We begin by analyzing existing frameworks and the requirements of downstream tasks, proposing a unified framework that integrates both T2I models and various editing tasks. Furthermore, another key challenge is the efficient creation of high-quality editing data, particularly for instruction-based and drag-based editing. To this end, we develop a synthetic data pipeline using sticker-like elements to synthesize accurate, high-quality datasets efficiently, which enables editing data scaling up for unified model training. For training, DreamOmni jointly trains T2I generation and downstream tasks. T2I training enhances the model's understanding of specific concepts and improves generation quality, while editing training helps the model grasp the nuances of the editing task. This collaboration significantly boosts editing performance. Extensive experiments confirm the effectiveness of DreamOmni. The code and model will be released.

Zhun Mou, Bin Xia, Zhengchao Huang et al.

Recent great advances in video generation models have demonstrated their potential to produce high-quality videos, bringing challenges to effective evaluation. Unlike human evaluation, existing automated evaluation metrics lack high-level semantic understanding and reasoning capabilities for video, thus making them infeasible and unexplainable. To fill this gap, we curate GRADEO-Instruct, a multi-dimensional T2V evaluation instruction tuning dataset, including 3.3k videos from over 10 existing video generation models and multi-step reasoning assessments converted by 16k human annotations. We then introduce GRADEO, one of the first specifically designed video evaluation models, which grades AI-generated videos for explainable scores and assessments through multi-step reasoning. Experiments show that our method aligns better with human evaluations than existing methods. Furthermore, our benchmarking reveals that current video generation models struggle to produce content that aligns with human reasoning and complex real-world scenarios. The models, datasets, and codes will be released soon.

Dongxia Liu, Jie Ma, Xiaochen Yang et al.

The creation of cinematic-quality animal effects necessitates the precise modeling of muscle and fur dynamics, a process that remains both labor-intensive and computationally expensive within traditional production workflows. While generative diffusion models have shown promise in diverse artistic workflows, their capacity for high-fidelity animal simulation remains largely unexploited. We present MoZoo, a generative dynamics solver that bypasses conventional refinement to synthesize high-fidelity animal videos from coarse meshes under multimodal guidance. We propose Role-Aware RoPE (RAR-RoPE) which employs role-based index remapping to synchronize motion alignment while decoupling reference information via fixed temporal offsets. Complementing this, Asymmetric Decoupled Attention partitions the latent sequence to enforce a unidirectional information flow, effectively preventing feature interference and improving computational efficiency. To address the scarcity of high-quality training data, we introduce MoZoo-Data, a synthetic-to-real pipeline that leverages a rendering engine and an inverse mapping approach to construct a large-scale dataset of paired sequences. Furthermore, we establish MoZooBench, a comprehensive benchmark with 120 mesh-video pairs. Experimental results demonstrate that MoZoo achieves high-fidelity fur simulation across diverse animal skeletons and layouts, preserving superior temporal and structural consistency.

Bin Xia, Jiyang Liu, Yuechen Zhang et al.

Instruction-based editing holds vast potential due to its simple and efficient interactive editing format. However, instruction-based editing, particularly for video, has been constrained by limited training data, hindering its practical application. To this end, we introduce DreamVE, a unified model for instruction-based image and video editing. Specifically, We propose a two-stage training strategy: first image editing, then video editing. This offers two main benefits: (1) Image data scales more easily, and models are more efficient to train, providing useful priors for faster and better video editing training. (2) Unifying image and video generation is natural and aligns with current trends. Moreover, we present comprehensive training data synthesis pipelines, including collage-based and generative model-based data synthesis. The collage-based data synthesis combines foreground objects and backgrounds to generate diverse editing data, such as object manipulation, background changes, and text modifications. It can easily generate billions of accurate, consistent, realistic, and diverse editing pairs. We pretrain DreamVE on extensive collage-based data to achieve strong performance in key editing types and enhance generalization and transfer capabilities. However, collage-based data lacks some attribute editing cases, leading to a relative drop in performance. In contrast, the generative model-based pipeline, despite being hard to scale up, offers flexibility in handling attribute editing cases. Therefore, we use generative model-based data to further fine-tune DreamVE. Besides, we design an efficient and powerful editing framework for DreamVE. We build on the SOTA T2V model and use a token concatenation with early drop approach to inject source image guidance, ensuring strong consistency and editability. The codes and models will be released.

Huiyun Cao, Yuan Shi, Bin Xia et al.

Diffusion models (DMs) have achieved promising performance in image restoration but haven't been explored for stereo images. The application of DM in stereo image restoration is confronted with a series of challenges. The need to reconstruct two images exacerbates DM's computational cost. Additionally, existing latent DMs usually focus on semantic information and remove high-frequency details as redundancy during latent compression, which is precisely what matters for image restoration. To address the above problems, we propose a high-frequency aware diffusion model, DiffStereo for stereo image restoration as the first attempt at DM in this domain. Specifically, DiffStereo first learns latent high-frequency representations (LHFR) of HQ images. DM is then trained in the learned space to estimate LHFR for stereo images, which are fused into a transformer-based stereo image restoration network providing beneficial high-frequency information of corresponding HQ images. The resolution of LHFR is kept the same as input images, which preserves the inherent texture from distortion. And the compression in channels alleviates the computational burden of DM. Furthermore, we devise a position encoding scheme when integrating the LHFR into the restoration network, enabling distinctive guidance in different depths of the restoration network. Comprehensive experiments verify that by combining generative DM and transformer, DiffStereo achieves both higher reconstruction accuracy and better perceptual quality on stereo super-resolution, deblurring, and low-light enhancement compared with state-of-the-art methods.

Ruofan Zhang, Bin Xia, Zhen Cheng et al.

Adaptive reasoning is essential for aligning the computational effort of large language models (LLMs) with the intrinsic difficulty of problems. Current chain-of-thought methods boost reasoning ability but indiscriminately generate long explanations, leading to evident inefficiency. However, existing reinforcement learning approaches to adaptive thinking remain unstable and heavily reward-dependent. Here we propose \textbf{DART}, a supervised \textbf{D}ifficulty-\textbf{A}daptive \textbf{R}easoning \textbf{T}runcation framework that adjusts thinking length according to problem difficulty. By distilling concise reasoning patterns from stronger models, interpolating them into a continuum of reasoning styles, and curating optimal training data that balances correctness and compactness, DART learns when to ``stop thinking''. Across multiple mathematical benchmarks, experimental results demonstrate its remarkable efficiency while preserving or improving accuracy, achieving a significant 81.2\% reasoning truncation (DeepSeek-R1-Distill-Qwen-7B on GSM8K dataset) with 5.33$\times$ computational acceleration. DART provides a stable and general paradigm for efficient reasoning, advancing the development of adaptive intelligence in LLMs.

Bin Xia, Bohao Peng, Yuechen Zhang et al.

Recent advancements in instruction-based image editing and subject-driven generation have garnered significant attention, yet both tasks still face limitations in meeting practical user needs. Instruction-based editing relies solely on language instructions, which often fail to capture specific editing details, making reference images necessary. Meanwhile, subject-driven generation is limited to combining concrete objects or people, overlooking broader, abstract concepts. To address these challenges, we propose two novel tasks: multimodal instruction-based editing and generation. These tasks support both text and image instructions and extend the scope to include both concrete and abstract concepts, greatly enhancing their practical applications. We introduce DreamOmni2, tackling two primary challenges: data creation and model framework design. Our data synthesis pipeline consists of three steps: (1) using a feature mixing method to create extraction data for both abstract and concrete concepts, (2) generating multimodal instruction-based editing training data using the editing and extraction models, and (3) further applying the extraction model to create training data for multimodal instruction-based editing. For the framework, to handle multi-image input, we propose an index encoding and position encoding shift scheme, which helps the model distinguish images and avoid pixel confusion. Additionally, we introduce joint training with the VLM and our generation/editing model to better process complex instructions. In addition, we have proposed comprehensive benchmarks for these two new tasks to drive their development. Experiments show that DreamOmni2 has achieved impressive results. Models and codes will be released.

Tong Wu, Zhiyong Chen, Meixia Tao et al.

Degraded broadcast channels (DBC) are a typical multiuser communication scenario, Semantic communications over DBC still lack in-depth research. In this paper, we design a semantic communications approach based on multi-user semantic fusion for wireless image transmission over DBC. In the proposed method, the transmitter extracts semantic features for two users separately. It then effectively fuses these semantic features for broadcasting by leveraging semantic similarity. Unlike traditional allocation of time, power, or bandwidth, the semantic fusion scheme can dynamically control the weight of the semantic features of the two users to balance the performance between the two users. Considering the different channel state information (CSI) of both users over DBC, a DBC-Aware method is developed that embeds the CSI of both users into the joint source-channel coding encoder and fusion module to adapt to the channel. Experimental results show that the proposed system outperforms the traditional broadcasting schemes.

Jintong Hu, Bin Xia, Bingchen Li et al.

Deep learning-based denoiser has been the focus of recent development on image denoising. In the past few years, there has been increasing interest in developing self-supervised denoising networks that only require noisy images, without the need for clean ground truth for training. However, a performance gap remains between current self-supervised methods and their supervised counterparts. Additionally, these methods commonly depend on assumptions about noise characteristics, thereby constraining their applicability in real-world scenarios. Inspired by the properties of the Frobenius norm expansion, we discover that incorporating a trace term reduces the optimization goal disparity between self-supervised and supervised methods, thereby enhancing the performance of self-supervised learning. To exploit this insight, we propose a trace-constraint loss function and design the low-trace adaptation Noise2Noise (LoTA-N2N) model that bridges the gap between self-supervised and supervised learning. Furthermore, we have discovered that several existing self-supervised denoising frameworks naturally fall within the proposed trace-constraint loss as subcases. Extensive experiments conducted on natural and confocal image datasets indicate that our method achieves state-of-the-art performance within the realm of zero-shot self-supervised image denoising approaches, without relying on any assumptions regarding the noise.

Xiaoyu Jin, Yuan Shi, Bin Xia et al.

Multi-modal Large Language Models (MLLMs) have a significant impact on various tasks, due to their extensive knowledge and powerful perception and generation capabilities. However, it still remains an open research problem on applying MLLMs to low-level vision tasks. In this paper, we present a simple MLLM-based Image Restoration framework to address this gap, namely Multi-modal Large Language Model based Restoration Assistant (LLMRA). We exploit the impressive capabilities of MLLMs to obtain the degradation information for universal image restoration. By employing a pretrained multi-modal large language model and a vision language model, we generate text descriptions and encode them as context embedding with degradation information for the degraded image. Through the proposed Context Enhance Module (CEM) and Degradation Context based Transformer Network (DC-former), we integrate these context embedding into the restoration network, contributing to more accurate and adjustable image restoration. Based on the dialogue with the users, our method leverages image degradation priors from MLLMs, providing low-level attributes descriptions of the input low-quality images and the restored high-quality images simultaneously. Extensive experiments demonstrate the superior performance of our LLMRA in universal image restoration tasks.

Bin Xia, Yapeng Tian, Yucheng Hang et al.

Reference-based super-resolution (RefSR) has made significant progress in producing realistic textures using an external reference (Ref) image. However, existing RefSR methods obtain high-quality correspondence matchings consuming quadratic computation resources with respect to the input size, limiting its application. Moreover, these approaches usually suffer from scale misalignments between the low-resolution (LR) image and Ref image. In this paper, we propose an Accelerated Multi-Scale Aggregation network (AMSA) for Reference-based Super-Resolution, including Coarse-to-Fine Embedded PatchMatch (CFE-PatchMatch) and Multi-Scale Dynamic Aggregation (MSDA) module. To improve matching efficiency, we design a novel Embedded PatchMacth scheme with random samples propagation, which involves end-to-end training with asymptotic linear computational cost to the input size. To further reduce computational cost and speed up convergence, we apply the coarse-to-fine strategy on Embedded PatchMacth constituting CFE-PatchMatch. To fully leverage reference information across multiple scales and enhance robustness to scale misalignment, we develop the MSDA module consisting of Dynamic Aggregation and Multi-Scale Aggregation. The Dynamic Aggregation corrects minor scale misalignment by dynamically aggregating features, and the Multi-Scale Aggregation brings robustness to large scale misalignment by fusing multi-scale information. Experimental results show that the proposed AMSA achieves superior performance over state-of-the-art approaches on both quantitative and qualitative evaluations.

Bin Xia, Yucheng Hang, Yapeng Tian et al.

Non-Local Attention (NLA) brings significant improvement for Single Image Super-Resolution (SISR) by leveraging intrinsic feature correlation in natural images. However, NLA gives noisy information large weights and consumes quadratic computation resources with respect to the input size, limiting its performance and application. In this paper, we propose a novel Efficient Non-Local Contrastive Attention (ENLCA) to perform long-range visual modeling and leverage more relevant non-local features. Specifically, ENLCA consists of two parts, Efficient Non-Local Attention (ENLA) and Sparse Aggregation. ENLA adopts the kernel method to approximate exponential function and obtains linear computation complexity. For Sparse Aggregation, we multiply inputs by an amplification factor to focus on informative features, yet the variance of approximation increases exponentially. Therefore, contrastive learning is applied to further separate relevant and irrelevant features. To demonstrate the effectiveness of ENLCA, we build an architecture called Efficient Non-Local Contrastive Network (ENLCN) by adding a few of our modules in a simple backbone. Extensive experimental results show that ENLCN reaches superior performance over state-of-the-art approaches on both quantitative and qualitative evaluations.