Yulin Wang

Zanlin Ni, Yulin Wang, Jiangwei Yu et al. · tsinghua

Recent years have witnessed a remarkable success of large deep learning models. However, training these models is challenging due to high computational costs, painfully slow convergence, and overfitting issues. In this paper, we present Deep Incubation, a novel approach that enables the efficient and effective training of large models by dividing them into smaller sub-modules that can be trained separately and assembled seamlessly. A key challenge for implementing this idea is to ensure the compatibility of the independently trained sub-modules. To address this issue, we first introduce a global, shared meta model, which is leveraged to implicitly link all the modules together, and can be designed as an extremely small network with negligible computational overhead. Then we propose a module incubation algorithm, which trains each sub-module to replace the corresponding component of the meta model and accomplish a given learning task. Despite the simplicity, our approach effectively encourages each sub-module to be aware of its role in the target large model, such that the finally-learned sub-modules can collaborate with each other smoothly after being assembled. Empirically, our method outperforms end-to-end (E2E) training in terms of both final accuracy and training efficiency. For example, on top of ViT-Huge, it improves the accuracy by 2.7% on ImageNet or achieves similar performance with 4x less training time. Notably, the gains are significant for downstream tasks as well (e.g., object detection and image segmentation on COCO and ADE20K). Code is available at https://github.com/LeapLabTHU/Deep-Incubation.

Yifan Pu, Yiru Wang, Zhuofan Xia et al. · tsinghua

Rotated object detection aims to identify and locate objects in images with arbitrary orientation. In this scenario, the oriented directions of objects vary considerably across different images, while multiple orientations of objects exist within an image. This intrinsic characteristic makes it challenging for standard backbone networks to extract high-quality features of these arbitrarily orientated objects. In this paper, we present Adaptive Rotated Convolution (ARC) module to handle the aforementioned challenges. In our ARC module, the convolution kernels rotate adaptively to extract object features with varying orientations in different images, and an efficient conditional computation mechanism is introduced to accommodate the large orientation variations of objects within an image. The two designs work seamlessly in rotated object detection problem. Moreover, ARC can conveniently serve as a plug-and-play module in various vision backbones to boost their representation ability to detect oriented objects accurately. Experiments on commonly used benchmarks (DOTA and HRSC2016) demonstrate that equipped with our proposed ARC module in the backbone network, the performance of multiple popular oriented object detectors is significantly improved (\eg +3.03\% mAP on Rotated RetinaNet and +4.16\% on CFA). Combined with the highly competitive method Oriented R-CNN, the proposed approach achieves state-of-the-art performance on the DOTA dataset with 81.77\% mAP. Code is available at \url{https://github.com/LeapLabTHU/ARC}.

Yizeng Han, Dongchen Han, Zeyu Liu et al. · tsinghua

Early exiting has become a promising approach to improving the inference efficiency of deep networks. By structuring models with multiple classifiers (exits), predictions for ``easy'' samples can be generated at earlier exits, negating the need for executing deeper layers. Current multi-exit networks typically implement linear classifiers at intermediate layers, compelling low-level features to encapsulate high-level semantics. This sub-optimal design invariably undermines the performance of later exits. In this paper, we propose Dynamic Perceiver (Dyn-Perceiver) to decouple the feature extraction procedure and the early classification task with a novel dual-branch architecture. A feature branch serves to extract image features, while a classification branch processes a latent code assigned for classification tasks. Bi-directional cross-attention layers are established to progressively fuse the information of both branches. Early exits are placed exclusively within the classification branch, thus eliminating the need for linear separability in low-level features. Dyn-Perceiver constitutes a versatile and adaptable framework that can be built upon various architectures. Experiments on image classification, action recognition, and object detection demonstrate that our method significantly improves the inference efficiency of different backbones, outperforming numerous competitive approaches across a broad range of computational budgets. Evaluation on both CPU and GPU platforms substantiate the superior practical efficiency of Dyn-Perceiver. Code is available at https://www.github.com/LeapLabTHU/Dynamic_Perceiver.

Yulin Wang, Yang Yue, Rui Lu et al. · tsinghua

The superior performance of modern deep networks usually comes with a costly training procedure. This paper presents a new curriculum learning approach for the efficient training of visual backbones (e.g., vision Transformers). Our work is inspired by the inherent learning dynamics of deep networks: we experimentally show that at an earlier training stage, the model mainly learns to recognize some 'easier-to-learn' discriminative patterns within each example, e.g., the lower-frequency components of images and the original information before data augmentation. Driven by this phenomenon, we propose a curriculum where the model always leverages all the training data at each epoch, while the curriculum starts with only exposing the 'easier-to-learn' patterns of each example, and introduces gradually more difficult patterns. To implement this idea, we 1) introduce a cropping operation in the Fourier spectrum of the inputs, which enables the model to learn from only the lower-frequency components efficiently, 2) demonstrate that exposing the features of original images amounts to adopting weaker data augmentation, and 3) integrate 1) and 2) and design a curriculum learning schedule with a greedy-search algorithm. The resulting approach, EfficientTrain, is simple, general, yet surprisingly effective. As an off-the-shelf method, it reduces the wall-time training cost of a wide variety of popular models (e.g., ResNet, ConvNeXt, DeiT, PVT, Swin, and CSWin) by >1.5x on ImageNet-1K/22K without sacrificing accuracy. It is also effective for self-supervised learning (e.g., MAE). Code is available at https://github.com/LeapLabTHU/EfficientTrain.

Honglin Xiong, Yuxian Tang, Feng Li et al.

Motion artifacts in magnetic resonance imaging (MRI) degrade diagnostic reliability. Existing deep learning methods are typically contrast-specific and fail to generalize across diverse modalities and artifact severities. We propose a unified framework combining parameter-informed contrast disentanglement with severity-aware adaptive correction. ScanCLIP, pretrained on over 30,000 MRI text-image pairs, derives contrast embeddings from acquisition parameters to disentangle contrast style from anatomical content, yielding contrast-free features. A Vision Transformer then estimates motion severity and routes features through a Mixture-of-Experts network, enabling targeted artifact correction. A dual-pathway decoder reconstructs both the clean image and residual artifact map, enforcing image-space consistency. On IXI and HCP benchmarks, our method improves PSNR by 0.75 dB and SSIM by up to 0.0279 over state-of-the-art approaches, with larger gains at higher artifact severities. It further demonstrates robust zero-shot generalization on real-world clinical data acquired with unseen scanning parameters, where existing methods either fail to remove artifacts or introduce additional distortions.

Zanlin Ni, Yulin Wang, Renping Zhou et al. · tsinghua

Recent studies have demonstrated the effectiveness of token-based methods for visual content generation. As a representative work, non-autoregressive Transformers (NATs) are able to synthesize images with decent quality in a small number of steps. However, NATs usually necessitate configuring a complicated generation policy comprising multiple manually-designed scheduling rules. These heuristic-driven rules are prone to sub-optimality and come with the requirements of expert knowledge and labor-intensive efforts. Moreover, their one-size-fits-all nature cannot flexibly adapt to the diverse characteristics of each individual sample. To address these issues, we propose AdaNAT, a learnable approach that automatically configures a suitable policy tailored for every sample to be generated. In specific, we formulate the determination of generation policies as a Markov decision process. Under this framework, a lightweight policy network for generation can be learned via reinforcement learning. Importantly, we demonstrate that simple reward designs such as FID or pre-trained reward models, may not reliably guarantee the desired quality or diversity of generated samples. Therefore, we propose an adversarial reward design to guide the training of policy networks effectively. Comprehensive experiments on four benchmark datasets, i.e., ImageNet-256 & 512, MS-COCO, and CC3M, validate the effectiveness of AdaNAT. Code and pre-trained models will be released at https://github.com/LeapLabTHU/AdaNAT.

Yulin Wang, Yang Yue, Xinhong Xu et al. · gatech, tsinghua

Recent research has revealed that reducing the temporal and spatial redundancy are both effective approaches towards efficient video recognition, e.g., allocating the majority of computation to a task-relevant subset of frames or the most valuable image regions of each frame. However, in most existing works, either type of redundancy is typically modeled with another absent. This paper explores the unified formulation of spatial-temporal dynamic computation on top of the recently proposed AdaFocusV2 algorithm, contributing to an improved AdaFocusV3 framework. Our method reduces the computational cost by activating the expensive high-capacity network only on some small but informative 3D video cubes. These cubes are cropped from the space formed by frame height, width, and video duration, while their locations are adaptively determined with a light-weighted policy network on a per-sample basis. At test time, the number of the cubes corresponding to each video is dynamically configured, i.e., video cubes are processed sequentially until a sufficiently reliable prediction is produced. Notably, AdaFocusV3 can be effectively trained by approximating the non-differentiable cropping operation with the interpolation of deep features. Extensive empirical results on six benchmark datasets (i.e., ActivityNet, FCVID, Mini-Kinetics, Something-Something V1&V2 and Diving48) demonstrate that our model is considerably more efficient than competitive baselines.

Yulin Wang, Yizeng Han, Chaofei Wang et al. · tsinghua

Over the past decade, deep learning models have exhibited considerable advancements, reaching or even exceeding human-level performance in a range of visual perception tasks. This remarkable progress has sparked interest in applying deep networks to real-world applications, such as autonomous vehicles, mobile devices, robotics, and edge computing. However, the challenge remains that state-of-the-art models usually demand significant computational resources, leading to impractical power consumption, latency, or carbon emissions in real-world scenarios. This trade-off between effectiveness and efficiency has catalyzed the emergence of a new research focus: computationally efficient deep learning, which strives to achieve satisfactory performance while minimizing the computational cost during inference. This review offers an extensive analysis of this rapidly evolving field by examining four key areas: 1) the development of static or dynamic light-weighted backbone models for the efficient extraction of discriminative deep representations; 2) the specialized network architectures or algorithms tailored for specific computer vision tasks; 3) the techniques employed for compressing deep learning models; and 4) the strategies for deploying efficient deep networks on hardware platforms. Additionally, we provide a systematic discussion on the critical challenges faced in this domain, such as network architecture design, training schemes, practical efficiency, and more realistic model compression approaches, as well as potential future research directions.

Chaoqun Du, Yulin Wang, Jiayi Guo et al.

Test-Time Adaptation (TTA) aims to adapt pre-trained models to the target domain during testing. In reality, this adaptability can be influenced by multiple factors. Researchers have identified various challenging scenarios and developed diverse methods to address these challenges, such as dealing with continual domain shifts, mixed domains, and temporally correlated or imbalanced class distributions. Despite these efforts, a unified and comprehensive benchmark has yet to be established. To this end, we propose a Unified Test-Time Adaptation (UniTTA) benchmark, which is comprehensive and widely applicable. Each scenario within the benchmark is fully described by a Markov state transition matrix for sampling from the original dataset. The UniTTA benchmark considers both domain and class as two independent dimensions of data and addresses various combinations of imbalance/balance and i.i.d./non-i.i.d./continual conditions, covering a total of \( (2 \times 3)^2 = 36 \) scenarios. It establishes a comprehensive evaluation benchmark for realistic TTA and provides a guideline for practitioners to select the most suitable TTA method. Alongside this benchmark, we propose a versatile UniTTA framework, which includes a Balanced Domain Normalization (BDN) layer and a COrrelated Feature Adaptation (COFA) method--designed to mitigate distribution gaps in domain and class, respectively. Extensive experiments demonstrate that our UniTTA framework excels within the UniTTA benchmark and achieves state-of-the-art performance on average. Our code is available at \url{https://github.com/LeapLabTHU/UniTTA}.

Ziwei Zheng, Zechuan Zhang, Yulin Wang et al.

Generic event boundary detection (GEBD), inspired by human visual cognitive behaviors of consistently segmenting videos into meaningful temporal chunks, finds utility in various applications such as video editing and. In this paper, we demonstrate that SOTA GEBD models often prioritize final performance over model complexity, resulting in low inference speed and hindering efficient deployment in real-world scenarios. We contribute to addressing this challenge by experimentally reexamining the architecture of GEBD models and uncovering several surprising findings. Firstly, we reveal that a concise GEBD baseline model already achieves promising performance without any sophisticated design. Secondly, we find that the widely applied image-domain backbones in GEBD models can contain plenty of architecture redundancy, motivating us to gradually ``modernize'' each component to enhance efficiency. Thirdly, we show that the GEBD models using image-domain backbones conducting the spatiotemporal learning in a spatial-then-temporal greedy manner can suffer from a distraction issue, which might be the inefficient villain for GEBD. Using a video-domain backbone to jointly conduct spatiotemporal modeling is an effective solution for this issue. The outcome of our exploration is a family of GEBD models, named EfficientGEBD, significantly outperforms the previous SOTA methods by up to 1.7\% performance gain and 280\% speedup under the same backbone. Our research prompts the community to design modern GEBD methods with the consideration of model complexity, particularly in resource-aware applications. The code is available at \url{https://github.com/Ziwei-Zheng/EfficientGEBD}.

Wenxuan Ma, Jinming Zhang, Shuang Li et al. · tsinghua

Extensive studies on Unsupervised Domain Adaptation (UDA) have propelled the deployment of deep learning from limited experimental datasets into real-world unconstrained domains. Most UDA approaches align features within a common embedding space and apply a shared classifier for target prediction. However, since a perfectly aligned feature space may not exist when the domain discrepancy is large, these methods suffer from two limitations. First, the coercive domain alignment deteriorates target domain discriminability due to lacking target label supervision. Second, the source-supervised classifier is inevitably biased to source data, thus it may underperform in target domain. To alleviate these issues, we propose to simultaneously conduct feature alignment in two individual spaces focusing on different domains, and create for each space a domain-oriented classifier tailored specifically for that domain. Specifically, we design a Domain-Oriented Transformer (DOT) that has two individual classification tokens to learn different domain-oriented representations, and two classifiers to preserve domain-wise discriminability. Theoretical guaranteed contrastive-based alignment and the source-guided pseudo-label refinement strategy are utilized to explore both domain-invariant and specific information. Comprehensive experiments validate that our method achieves state-of-the-art on several benchmarks.

Jiaqi Wu, Zhen Wang, Enhao Huang et al.

Text-guided multispectral object detection uses text semantics to guide semantic-aware cross-modal interaction between RGB and IR for more robust perception. However, notable limitations remain: (1) existing methods often use text only as an auxiliary semantic enhancement signal, without exploiting its guiding role to bridge the inherent granularity asymmetry between RGB and IR; and (2) conventional data-driven attention-based fusion tends to emphasize stable consensus while overlooking potentially valuable cross-modal discrepancies. To address these issues, we propose a semantic bridge fusion framework with bi-support modeling for multispectral object detection. Specifically, text is used as a shared semantic bridge to align RGB and IR responses under a unified category condition, while the recalibrated thermal semantic prior is projected onto the RGB branch for semantic-level mapping fusion. We further formulate RGB-IR interaction evidence into the regular consensus support and the complementary discrepancy support that contains potentially discriminative cues, and introduce them into fusion via dynamic recalibration as a structured inductive bias. In addition, we design a bidirectional semantic alignment module for closed-loop vision-text guidance enhancement. Extensive experiments demonstrate the effectiveness of the proposed fusion framework and its superior detection performance on multispectral benchmarks. Code is available at https://github.com/zhenwang5372/Bridging-RGB-IR-Gap.

Dawei Li, Yaxuan Li, Dheeraj Mekala et al.

In-Context Learning (ICL) combined with pre-trained large language models has achieved promising results on various NLP tasks. However, ICL requires high-quality annotated demonstrations which might not be available in real-world scenarios. To overcome this limitation, we propose \textbf{D}ata \textbf{A}ugmentation for \textbf{I}n-Context \textbf{L}earning (\textbf{DAIL}). DAIL leverages the intuition that large language models are more familiar with the content generated by themselves. It first utilizes the language model to generate paraphrases of the test sample and employs majority voting to determine the final result based on individual predictions. Our extensive empirical evaluation shows that DAIL outperforms the standard ICL method and other ensemble-based methods in the low-resource scenario. Additionally, we explore the use of voting consistency as a confidence score of the model when the logits of predictions are inaccessible. We believe our work will stimulate further research on ICL in low-resource settings.

Haojun Jiang, Teng Wang, Zhenguo Sun et al.

Echocardiography is an essential medical technique for diagnosing cardiovascular diseases, but its high operational complexity has led to a shortage of trained professionals. To address this issue, we introduce a novel probe movement guidance algorithm that has the potential to be applied in guiding robotic systems or novices with probe pose adjustment for high-quality standard plane image acquisition.Cardiac ultrasound faces two major challenges: (1) the inherently complex structure of the heart, and (2) significant individual variations. Previous works have only learned the population-averaged structure of the heart rather than personalized cardiac structures, leading to a performance bottleneck. Clinically, we observe that sonographers dynamically adjust their interpretation of a patient's cardiac anatomy based on prior scanning sequences, consequently refining their scanning strategies. Inspired by this, we propose a novel sequence-aware self-supervised pre-training method. Specifically, our approach learns personalized three-dimensional cardiac structural features by predicting the masked-out image features and probe movement actions in a scanning sequence. We hypothesize that if the model can predict the missing content it has acquired a good understanding of personalized cardiac structure. Extensive experiments on a large-scale expert scanning dataset with 1.67 million samples demonstrate that our proposed sequence-aware paradigm can effectively reduce probe guidance errors compared to other advanced baseline methods.

Chaoqun Du, Yulin Wang, Shiji Song et al. · tsinghua

Long-tailed distributions frequently emerge in real-world data, where a large number of minority categories contain a limited number of samples. Such imbalance issue considerably impairs the performance of standard supervised learning algorithms, which are mainly designed for balanced training sets. Recent investigations have revealed that supervised contrastive learning exhibits promising potential in alleviating the data imbalance. However, the performance of supervised contrastive learning is plagued by an inherent challenge: it necessitates sufficiently large batches of training data to construct contrastive pairs that cover all categories, yet this requirement is difficult to meet in the context of class-imbalanced data. To overcome this obstacle, we propose a novel probabilistic contrastive (ProCo) learning algorithm that estimates the data distribution of the samples from each class in the feature space, and samples contrastive pairs accordingly. In fact, estimating the distributions of all classes using features in a small batch, particularly for imbalanced data, is not feasible. Our key idea is to introduce a reasonable and simple assumption that the normalized features in contrastive learning follow a mixture of von Mises-Fisher (vMF) distributions on unit space, which brings two-fold benefits. First, the distribution parameters can be estimated using only the first sample moment, which can be efficiently computed in an online manner across different batches. Second, based on the estimated distribution, the vMF distribution allows us to sample an infinite number of contrastive pairs and derive a closed form of the expected contrastive loss for efficient optimization. Our code is available at https://github.com/LeapLabTHU/ProCo.

Yang Yue, Yulin Wang, Bingyi Kang et al.

MLLMs have demonstrated remarkable comprehension and reasoning capabilities with complex language and visual data. These advances have spurred the vision of establishing a generalist robotic MLLM proficient in understanding complex human instructions and accomplishing various embodied tasks. However, developing MLLMs for real-world robots is challenging due to the typically limited computation and memory capacities available on robotic platforms. In contrast, the inference of MLLMs involves storing billions of parameters and performing tremendous computation, imposing significant hardware demands. In our paper, we propose a Dynamic Early-Exit Framework for Robotic Vision-Language-Action Model (DeeR-VLA, or simply DeeR) that automatically adjusts the size of the activated MLLM based on each situation at hand. The approach leverages a multi-exit architecture in MLLMs, which allows the model to terminate processing once a proper size of the model has been activated for a specific situation, thus avoiding further redundant computation. Additionally, we develop novel algorithms that establish early-termination criteria for DeeR, conditioned on predefined demands such as average computational cost (i.e., power consumption), as well as peak computational consumption (i.e., latency) and GPU memory usage. These enhancements ensure that DeeR operates efficiently under varying resource constraints while maintaining competitive performance. On the CALVIN robot manipulation benchmark, DeeR demonstrates significant reductions in computational costs of LLM by 5.2-6.5x and GPU memory of LLM by 2-6x without compromising performance. Code and checkpoints are available at https://github.com/yueyang130/DeeR-VLA.

Yulin Wang, Honglin Xiong, Kaicong Sun et al.

Multimodal brain magnetic resonance (MR) imaging is indispensable in neuroscience and neurology. However, due to the accessibility of MRI scanners and their lengthy acquisition time, multimodal MR images are not commonly available. Current MR image synthesis approaches are typically trained on independent datasets for specific tasks, leading to suboptimal performance when applied to novel datasets and tasks. Here, we present TUMSyn, a Text-guided Universal MR image Synthesis generalist model, which can flexibly generate brain MR images with demanded imaging metadata from routinely acquired scans guided by text prompts. To ensure TUMSyn's image synthesis precision, versatility, and generalizability, we first construct a brain MR database comprising 31,407 3D images with 7 MRI modalities from 13 centers. We then pre-train an MRI-specific text encoder using contrastive learning to effectively control MR image synthesis based on text prompts. Extensive experiments on diverse datasets and physician assessments indicate that TUMSyn can generate clinically meaningful MR images with specified imaging metadata in supervised and zero-shot scenarios. Therefore, TUMSyn can be utilized along with acquired MR scan(s) to facilitate large-scale MRI-based screening and diagnosis of brain diseases.

Fengxiang Wang, Hongzhen Wang, Mingshuo Chen et al.

The astonishing breakthrough of multimodal large language models (MLLMs) has necessitated new benchmarks to quantitatively assess their capabilities, reveal their limitations, and indicate future research directions. However, this is challenging in the context of remote sensing (RS), since the imagery features ultra-high resolution that incorporates extremely complex semantic relationships. Existing benchmarks usually adopt notably smaller image sizes than real-world RS scenarios, suffer from limited annotation quality, and consider insufficient dimensions of evaluation. To address these issues, we present XLRS-Bench: a comprehensive benchmark for evaluating the perception and reasoning capabilities of MLLMs in ultra-high-resolution RS scenarios. XLRS-Bench boasts the largest average image size (8500$\times$8500) observed thus far, with all evaluation samples meticulously annotated manually, assisted by a novel semi-automatic captioner on ultra-high-resolution RS images. On top of the XLRS-Bench, 16 sub-tasks are defined to evaluate MLLMs' 10 kinds of perceptual capabilities and 6 kinds of reasoning capabilities, with a primary emphasis on advanced cognitive processes that facilitate real-world decision-making and the capture of spatiotemporal changes. The results of both general and RS-focused MLLMs on XLRS-Bench indicate that further efforts are needed for real-world RS applications. We have open-sourced XLRS-Bench to support further research in developing more powerful MLLMs for remote sensing.

Zanlin Ni, Yulin Wang, Renping Zhou et al.

Recently, token-based generation have demonstrated their effectiveness in image synthesis. As a representative example, non-autoregressive Transformers (NATs) can generate decent-quality images in a few steps. NATs perform generation in a progressive manner, where the latent tokens of a resulting image are incrementally revealed. At each step, the unrevealed image regions are padded with mask tokens and inferred by NAT. In this paper, we delve into the mechanisms behind the effectiveness of NATs and uncover two important patterns that naturally emerge from NATs: Spatially (within a step), although mask and visible tokens are processed uniformly by NATs, the interactions between them are highly asymmetric. In specific, mask tokens mainly gather information for decoding, while visible tokens tend to primarily provide information, and their deep representations can be built only upon themselves. Temporally (across steps), the interactions between adjacent generation steps mostly concentrate on updating the representations of a few critical tokens, while the computation for the majority of tokens is generally repetitive. Driven by these findings, we propose EfficientNAT (ENAT), a NAT model that explicitly encourages these critical interactions inherent in NATs. At the spatial level, we disentangle the computations of visible and mask tokens by encoding visible tokens independently, while decoding mask tokens conditioned on the fully encoded visible tokens. At the temporal level, we prioritize the computation of the critical tokens at each step, while maximally reusing previously computed token representations to supplement necessary information. ENAT improves the performance of NATs notably with significantly reduced computational cost. Experiments on ImageNet-256, ImageNet-512 and MS-COCO validate the effectiveness of ENAT. Code is available at https://github.com/LeapLabTHU/ENAT.

Fengxiang Wang, Yulin Wang, Mingshuo Chen et al.

Recent advances in self-supervised learning for Vision Transformers (ViTs) have fueled breakthroughs in remote sensing (RS) foundation models. However, the quadratic complexity of self-attention poses a significant barrier to scalability, particularly for large models and high-resolution images. While the linear-complexity Mamba architecture offers a promising alternative, existing RS applications of Mamba remain limited to supervised tasks on small, domain-specific datasets. To address these challenges, we propose RoMA, a framework that enables scalable self-supervised pretraining of Mamba-based RS foundation models using large-scale, diverse, unlabeled data. RoMA enhances scalability for high-resolution images through a tailored auto-regressive learning strategy, incorporating two key innovations: 1) a rotation-aware pretraining mechanism combining adaptive cropping with angular embeddings to handle sparsely distributed objects with arbitrary orientations, and 2) multi-scale token prediction objectives that address the extreme variations in object scales inherent to RS imagery. Systematic empirical studies validate that Mamba adheres to RS data and parameter scaling laws, with performance scaling reliably as model and data size increase. Furthermore, experiments across scene classification, object detection, and semantic segmentation tasks demonstrate that RoMA-pretrained Mamba models consistently outperform ViT-based counterparts in both accuracy and computational efficiency. The source code and pretrained models will be released at https://github.com/MiliLab/RoMA.

Yang Yue, Yulin Wang, Haojun Jiang et al.

Echocardiography is crucial for cardiovascular disease detection but relies heavily on experienced sonographers. Echocardiography probe guidance systems, which provide real-time movement instructions for acquiring standard plane images, offer a promising solution for AI-assisted or fully autonomous scanning. However, developing effective machine learning models for this task remains challenging, as they must grasp heart anatomy and the intricate interplay between probe motion and visual signals. To address this, we present EchoWorld, a motion-aware world modeling framework for probe guidance that encodes anatomical knowledge and motion-induced visual dynamics, while effectively leveraging past visual-motion sequences to enhance guidance precision. EchoWorld employs a pre-training strategy inspired by world modeling principles, where the model predicts masked anatomical regions and simulates the visual outcomes of probe adjustments. Built upon this pre-trained model, we introduce a motion-aware attention mechanism in the fine-tuning stage that effectively integrates historical visual-motion data, enabling precise and adaptive probe guidance. Trained on more than one million ultrasound images from over 200 routine scans, EchoWorld effectively captures key echocardiographic knowledge, as validated by qualitative analysis. Moreover, our method significantly reduces guidance errors compared to existing visual backbones and guidance frameworks, excelling in both single-frame and sequential evaluation protocols. Code is available at https://github.com/LeapLabTHU/EchoWorld.

Feihong Yan, Qingyan Wei, Jiayi Tang et al.

Masked Autoregressive (MAR) models have emerged as a promising approach in image generation, expected to surpass traditional autoregressive models in computational efficiency by leveraging the capability of parallel decoding. However, their dependence on bidirectional self-attention inherently conflicts with conventional KV caching mechanisms, creating unexpected computational bottlenecks that undermine their expected efficiency. To address this problem, this paper studies the caching mechanism for MAR by leveraging two types of redundancy: Token Redundancy indicates that a large portion of tokens have very similar representations in the adjacent decoding steps, which allows us to first cache them in previous steps and then reuse them in the later steps. Condition Redundancy indicates that the difference between conditional and unconditional output in classifier-free guidance exhibits very similar values in adjacent steps. Based on these two redundancies, we propose LazyMAR, which introduces two caching mechanisms to handle them one by one. LazyMAR is training-free and plug-and-play for all MAR models. Experimental results demonstrate that our method achieves 2.83 times acceleration with almost no drop in generation quality. Our codes will be released in https://github.com/feihongyan1/LazyMAR.

Renping Zhou, Zanlin Ni, Tianyi Chen et al.

Recently, Masked Diffusion Models (MDMs) have shown promising potential across vision, language, and cross-modal generation. However, a notable discrepancy exists between their training and inference procedures. In particular, MDM inference is a multi-step, iterative process governed not only by the model itself but also by various schedules that dictate the token-decoding trajectory (e.g., how many tokens to decode at each step). In contrast, MDMs are typically trained using a simplified, single-step BERT-style objective that masks a subset of tokens and predicts all of them simultaneously. This step-level simplification fundamentally disconnects the training paradigm from the trajectory-level nature of inference, leaving the inference schedules never optimized during training. In this paper, we introduce Co-GRPO, which reformulates MDM generation as a unified Markov Decision Process (MDP) that jointly incorporates both the model and the inference schedule. By applying Group Relative Policy Optimization at the trajectory level, Co-GRPO cooperatively optimizes model parameters and schedule parameters under a shared reward, without requiring costly backpropagation through the multi-step generation process. This holistic optimization aligns training with inference more thoroughly and substantially improves generation quality. Empirical results across four benchmarks-ImageReward, HPS, GenEval, and DPG-Bench-demonstrate the effectiveness of our approach. For more details, please refer to our project page: https://co-grpo.github.io/ .

Yang Yue, Yulin Wang, Chenxin Tao et al.

Humans can develop internal world models that encode common sense knowledge, telling them how the world works and predicting the consequences of their actions. This concept has emerged as a promising direction for establishing general-purpose machine-learning models in recent preliminary works, e.g., for visual representation learning. In this paper, we present CheXWorld, the first effort towards a self-supervised world model for radiographic images. Specifically, our work develops a unified framework that simultaneously models three aspects of medical knowledge essential for qualified radiologists, including 1) local anatomical structures describing the fine-grained characteristics of local tissues (e.g., architectures, shapes, and textures); 2) global anatomical layouts describing the global organization of the human body (e.g., layouts of organs and skeletons); and 3) domain variations that encourage CheXWorld to model the transitions across different appearance domains of radiographs (e.g., varying clarity, contrast, and exposure caused by collecting radiographs from different hospitals, devices, or patients). Empirically, we design tailored qualitative and quantitative analyses, revealing that CheXWorld successfully captures these three dimensions of medical knowledge. Furthermore, transfer learning experiments across eight medical image classification and segmentation benchmarks showcase that CheXWorld significantly outperforms existing SSL methods and large-scale medical foundation models. Code & pre-trained models are available at https://github.com/LeapLabTHU/CheXWorld.

Jiakai Tang, Chuan Wang, Gaoming Yang et al.

Industrial recommender systems face two fundamental limitations under the log-driven paradigm: (1) knowledge poverty in ID-based item representations that causes brittle interest modeling under data sparsity, and (2) systemic blindness to beyond-log user interests that constrains model performance within platform boundaries. These limitations stem from an over-reliance on shallow interaction statistics and close-looped feedback while neglecting the rich world knowledge about product semantics and cross-domain behavioral patterns that Large Language Models have learned from vast corpora. To address these challenges, we introduce ReaSeq, a reasoning-enhanced framework that leverages world knowledge in Large Language Models to address both limitations through explicit and implicit reasoning. Specifically, ReaSeq employs explicit Chain-of-Thought reasoning via multi-agent collaboration to distill structured product knowledge into semantically enriched item representations, and latent reasoning via Diffusion Large Language Models to infer plausible beyond-log behaviors. Deployed on Taobao's ranking system serving hundreds of millions of users, ReaSeq achieves substantial gains: >6.0% in IPV and CTR, >2.9% in Orders, and >2.5% in GMV, validating the effectiveness of world-knowledge-enhanced reasoning over purely log-driven approaches.

Jiayi Guo, Chuanhao Yan, Xingqian Xu et al.

Ensuring precise multimodal alignment between diffusion-generated images and input prompts has been a long-standing challenge. Earlier works finetune diffusion weight using high-quality preference data, which tends to be limited and difficult to scale up. Recent editing-based methods further refine local regions of generated images but may compromise overall image quality. In this work, we propose Implicit Multimodal Guidance (IMG), a novel re-generation-based multimodal alignment framework that requires no extra data or editing operations. Specifically, given a generated image and its prompt, IMG a) utilizes a multimodal large language model (MLLM) to identify misalignments; b) introduces an Implicit Aligner that manipulates diffusion conditioning features to reduce misalignments and enable re-generation; and c) formulates the re-alignment goal into a trainable objective, namely Iteratively Updated Preference Objective. Extensive qualitative and quantitative evaluations on SDXL, SDXL-DPO, and FLUX show that IMG outperforms existing alignment methods. Furthermore, IMG acts as a flexible plug-and-play adapter, seamlessly enhancing prior finetuning-based alignment methods. Our code will be available at https://github.com/SHI-Labs/IMG-Multimodal-Diffusion-Alignment.

Yulin Wang, Yang Yue, Yang Yue et al. · tsinghua

Human vision is highly adaptive, efficiently sampling intricate environments by sequentially fixating on task-relevant regions. In contrast, prevailing machine vision models passively process entire scenes at once, resulting in excessive resource demands scaling with spatial-temporal input resolution and model size, yielding critical limitations impeding both future advancements and real-world application. Here we introduce AdaptiveNN, a general framework aiming to drive a paradigm shift from 'passive' to 'active, adaptive' vision models. AdaptiveNN formulates visual perception as a coarse-to-fine sequential decision-making process, progressively identifying and attending to regions pertinent to the task, incrementally combining information across fixations, and actively concluding observation when sufficient. We establish a theory integrating representation learning with self-rewarding reinforcement learning, enabling end-to-end training of the non-differentiable AdaptiveNN without additional supervision on fixation locations. We assess AdaptiveNN on 17 benchmarks spanning 9 tasks, including large-scale visual recognition, fine-grained discrimination, visual search, processing images from real driving and medical scenarios, language-driven embodied AI, and side-by-side comparisons with humans. AdaptiveNN achieves up to 28x inference cost reduction without sacrificing accuracy, flexibly adapts to varying task demands and resource budgets without retraining, and provides enhanced interpretability via its fixation patterns, demonstrating a promising avenue toward efficient, flexible, and interpretable computer vision. Furthermore, AdaptiveNN exhibits closely human-like perceptual behaviors in many cases, revealing its potential as a valuable tool for investigating visual cognition. Code is available at https://github.com/LeapLabTHU/AdaptiveNN.

Yulin Wang, Mengting Hu, Hongli Li et al.

In pose estimation for seen objects, a prevalent pipeline involves using neural networks to predict dense 3D coordinates of the object surface on 2D images, which are then used to establish dense 2D-3D correspondences. However, current methods primarily focus on more efficient encoding techniques to improve the precision of predicted 3D coordinates on the object's front surface, overlooking the potential benefits of incorporating the back surface and interior of the object. To better utilize the full surface and interior of the object, this study predicts 3D coordinates of both the object's front and back surfaces and densely samples 3D coordinates between them. This process creates ultra-dense 2D-3D correspondences, effectively enhancing pose estimation accuracy based on the Perspective-n-Point (PnP) algorithm. Additionally, we propose Hierarchical Continuous Coordinate Encoding (HCCE) to provide a more accurate and efficient representation of front and back surface coordinates. Experimental results show that, compared to existing state-of-the-art (SOTA) methods on the BOP website, the proposed approach outperforms across seven classic BOP core datasets. Code is available at https://github.com/WangYuLin-SEU/HCCEPose.

Zanlin Ni, Yulin Wang, Renping Zhou et al.

The field of image synthesis is currently flourishing due to the advancements in diffusion models. While diffusion models have been successful, their computational intensity has prompted the pursuit of more efficient alternatives. As a representative work, non-autoregressive Transformers (NATs) have been recognized for their rapid generation. However, a major drawback of these models is their inferior performance compared to diffusion models. In this paper, we aim to re-evaluate the full potential of NATs by revisiting the design of their training and inference strategies. Specifically, we identify the complexities in properly configuring these strategies and indicate the possible sub-optimality in existing heuristic-driven designs. Recognizing this, we propose to go beyond existing methods by directly solving the optimal strategies in an automatic framework. The resulting method, named AutoNAT, advances the performance boundaries of NATs notably, and is able to perform comparably with the latest diffusion models at a significantly reduced inference cost. The effectiveness of AutoNAT is validated on four benchmark datasets, i.e., ImageNet-256 & 512, MS-COCO, and CC3M. Our code is available at https://github.com/LeapLabTHU/ImprovedNAT.

Jiayi Guo, Junhao Zhao, Chaoqun Du et al.

Test-time adaptation (TTA) aims to improve the performance of source-domain pre-trained models on previously unseen, shifted target domains. Traditional TTA methods primarily adapt model weights based on target data streams, making model performance sensitive to the amount and order of target data. The recently proposed diffusion-driven TTA methods mitigate this by adapting model inputs instead of weights, where an unconditional diffusion model, trained on the source domain, transforms target-domain data into a synthetic domain that is expected to approximate the source domain. However, in this paper, we reveal that although the synthetic data in diffusion-driven TTA seems indistinguishable from the source data, it is unaligned with, or even markedly different from the latter for deep networks. To address this issue, we propose a \textbf{S}ynthetic-\textbf{D}omain \textbf{A}lignment (SDA) framework. Our key insight is to fine-tune the source model with synthetic data to ensure better alignment. Specifically, we first employ a conditional diffusion model to generate labeled samples, creating a synthetic dataset. Subsequently, we use the aforementioned unconditional diffusion model to add noise to and denoise each sample before fine-tuning. This Mix of Diffusion (MoD) process mitigates the potential domain misalignment between the conditional and unconditional models. Extensive experiments across classifiers, segmenters, and multimodal large language models (MLLMs, \eg, LLaVA) demonstrate that SDA achieves superior domain alignment and consistently outperforms existing diffusion-driven TTA methods. Our code is available at https://github.com/SHI-Labs/Diffusion-Driven-Test-Time-Adaptation-via-Synthetic-Domain-Alignment.

Gao Huang, Yulin Wang, Kangchen Lv et al.

Spatial redundancy widely exists in visual recognition tasks, i.e., discriminative features in an image or video frame usually correspond to only a subset of pixels, while the remaining regions are irrelevant to the task at hand. Therefore, static models which process all the pixels with an equal amount of computation result in considerable redundancy in terms of time and space consumption. In this paper, we formulate the image recognition problem as a sequential coarse-to-fine feature learning process, mimicking the human visual system. Specifically, the proposed Glance and Focus Network (GFNet) first extracts a quick global representation of the input image at a low resolution scale, and then strategically attends to a series of salient (small) regions to learn finer features. The sequential process naturally facilitates adaptive inference at test time, as it can be terminated once the model is sufficiently confident about its prediction, avoiding further redundant computation. It is worth noting that the problem of locating discriminant regions in our model is formulated as a reinforcement learning task, thus requiring no additional manual annotations other than classification labels. GFNet is general and flexible as it is compatible with any off-the-shelf backbone models (such as MobileNets, EfficientNets and TSM), which can be conveniently deployed as the feature extractor. Extensive experiments on a variety of image classification and video recognition tasks and with various backbone models demonstrate the remarkable efficiency of our method. For example, it reduces the average latency of the highly efficient MobileNet-V3 on an iPhone XS Max by 1.3x without sacrificing accuracy. Code and pre-trained models are available at https://github.com/blackfeather-wang/GFNet-Pytorch.

Yulin Wang, Yang Yue, Yuanze Lin et al.

Recent works have shown that the computational efficiency of video recognition can be significantly improved by reducing the spatial redundancy. As a representative work, the adaptive focus method (AdaFocus) has achieved a favorable trade-off between accuracy and inference speed by dynamically identifying and attending to the informative regions in each video frame. However, AdaFocus requires a complicated three-stage training pipeline (involving reinforcement learning), leading to slow convergence and is unfriendly to practitioners. This work reformulates the training of AdaFocus as a simple one-stage algorithm by introducing a differentiable interpolation-based patch selection operation, enabling efficient end-to-end optimization. We further present an improved training scheme to address the issues introduced by the one-stage formulation, including the lack of supervision, input diversity and training stability. Moreover, a conditional-exit technique is proposed to perform temporal adaptive computation on top of AdaFocus without additional training. Extensive experiments on six benchmark datasets (i.e., ActivityNet, FCVID, Mini-Kinetics, Something-Something V1&V2, and Jester) demonstrate that our model significantly outperforms the original AdaFocus and other competitive baselines, while being considerably more simple and efficient to train. Code is available at https://github.com/LeapLabTHU/AdaFocusV2.

Yulin Wang, Rui Huang, Shiji Song et al.

Vision Transformers (ViT) have achieved remarkable success in large-scale image recognition. They split every 2D image into a fixed number of patches, each of which is treated as a token. Generally, representing an image with more tokens would lead to higher prediction accuracy, while it also results in drastically increased computational cost. To achieve a decent trade-off between accuracy and speed, the number of tokens is empirically set to 16x16 or 14x14. In this paper, we argue that every image has its own characteristics, and ideally the token number should be conditioned on each individual input. In fact, we have observed that there exist a considerable number of "easy" images which can be accurately predicted with a mere number of 4x4 tokens, while only a small fraction of "hard" ones need a finer representation. Inspired by this phenomenon, we propose a Dynamic Transformer to automatically configure a proper number of tokens for each input image. This is achieved by cascading multiple Transformers with increasing numbers of tokens, which are sequentially activated in an adaptive fashion at test time, i.e., the inference is terminated once a sufficiently confident prediction is produced. We further design efficient feature reuse and relationship reuse mechanisms across different components of the Dynamic Transformer to reduce redundant computations. Extensive empirical results on ImageNet, CIFAR-10, and CIFAR-100 demonstrate that our method significantly outperforms the competitive baselines in terms of both theoretical computational efficiency and practical inference speed. Code and pre-trained models (based on PyTorch and MindSpore) are available at https://github.com/blackfeather-wang/Dynamic-Vision-Transformer and https://github.com/blackfeather-wang/Dynamic-Vision-Transformer-MindSpore.

Yulin Wang, Zhaoxi Chen, Haojun Jiang et al.

In this paper, we explore the spatial redundancy in video recognition with the aim to improve the computational efficiency. It is observed that the most informative region in each frame of a video is usually a small image patch, which shifts smoothly across frames. Therefore, we model the patch localization problem as a sequential decision task, and propose a reinforcement learning based approach for efficient spatially adaptive video recognition (AdaFocus). In specific, a light-weighted ConvNet is first adopted to quickly process the full video sequence, whose features are used by a recurrent policy network to localize the most task-relevant regions. Then the selected patches are inferred by a high-capacity network for the final prediction. During offline inference, once the informative patch sequence has been generated, the bulk of computation can be done in parallel, and is efficient on modern GPU devices. In addition, we demonstrate that the proposed method can be easily extended by further considering the temporal redundancy, e.g., dynamically skipping less valuable frames. Extensive experiments on five benchmark datasets, i.e., ActivityNet, FCVID, Mini-Kinetics, Something-Something V1&V2, demonstrate that our method is significantly more efficient than the competitive baselines. Code is available at https://github.com/blackfeather-wang/AdaFocus.

Yulin Wang, Zanlin Ni, Shiji Song et al.

Due to the need to store the intermediate activations for back-propagation, end-to-end (E2E) training of deep networks usually suffers from high GPUs memory footprint. This paper aims to address this problem by revisiting the locally supervised learning, where a network is split into gradient-isolated modules and trained with local supervision. We experimentally show that simply training local modules with E2E loss tends to collapse task-relevant information at early layers, and hence hurts the performance of the full model. To avoid this issue, we propose an information propagation (InfoPro) loss, which encourages local modules to preserve as much useful information as possible, while progressively discard task-irrelevant information. As InfoPro loss is difficult to compute in its original form, we derive a feasible upper bound as a surrogate optimization objective, yielding a simple but effective algorithm. In fact, we show that the proposed method boils down to minimizing the combination of a reconstruction loss and a normal cross-entropy/contrastive term. Extensive empirical results on five datasets (i.e., CIFAR, SVHN, STL-10, ImageNet and Cityscapes) validate that InfoPro is capable of achieving competitive performance with less than 40% memory footprint compared to E2E training, while allowing using training data with higher-resolution or larger batch sizes under the same GPU memory constraint. Our method also enables training local modules asynchronously for potential training acceleration. Code is available at: https://github.com/blackfeather-wang/InfoPro-Pytorch.

Yulin Wang, Kangchen Lv, Rui Huang et al.

The accuracy of deep convolutional neural networks (CNNs) generally improves when fueled with high resolution images. However, this often comes at a high computational cost and high memory footprint. Inspired by the fact that not all regions in an image are task-relevant, we propose a novel framework that performs efficient image classification by processing a sequence of relatively small inputs, which are strategically selected from the original image with reinforcement learning. Such a dynamic decision process naturally facilitates adaptive inference at test time, i.e., it can be terminated once the model is sufficiently confident about its prediction and thus avoids further redundant computation. Notably, our framework is general and flexible as it is compatible with most of the state-of-the-art light-weighted CNNs (such as MobileNets, EfficientNets and RegNets), which can be conveniently deployed as the backbone feature extractor. Experiments on ImageNet show that our method consistently improves the computational efficiency of a wide variety of deep models. For example, it further reduces the average latency of the highly efficient MobileNet-V3 on an iPhone XS Max by 20% without sacrificing accuracy. Code and pre-trained models are available at https://github.com/blackfeather-wang/GFNet-Pytorch.

Yulin Wang, Xuran Pan, Shiji Song et al.

In this paper, we propose a novel implicit semantic data augmentation (ISDA) approach to complement traditional augmentation techniques like flipping, translation or rotation. Our work is motivated by the intriguing property that deep networks are surprisingly good at linearizing features, such that certain directions in the deep feature space correspond to meaningful semantic transformations, e.g., adding sunglasses or changing backgrounds. As a consequence, translating training samples along many semantic directions in the feature space can effectively augment the dataset to improve generalization. To implement this idea effectively and efficiently, we first perform an online estimate of the covariance matrix of deep features for each class, which captures the intra-class semantic variations. Then random vectors are drawn from a zero-mean normal distribution with the estimated covariance to augment the training data in that class. Importantly, instead of augmenting the samples explicitly, we can directly minimize an upper bound of the expected cross-entropy (CE) loss on the augmented training set, leading to a highly efficient algorithm. In fact, we show that the proposed ISDA amounts to minimizing a novel robust CE loss, which adds negligible extra computational cost to a normal training procedure. Although being simple, ISDA consistently improves the generalization performance of popular deep models (ResNets and DenseNets) on a variety of datasets, e.g., CIFAR-10, CIFAR-100 and ImageNet. Code for reproducing our results is available at https://github.com/blackfeather-wang/ISDA-for-Deep-Networks.

Yining Li, Dongchen Han, Zeyu Liu et al.

While linear-complexity attention mechanisms offer a promising alternative to Softmax attention for overcoming the quadratic bottleneck, training such models from scratch remains prohibitively expensive. Inheriting weights from pretrained Transformers provides an appealing shortcut, yet the fundamental representational gap between Softmax and linear attention prevents effective weight transfer. In this work, we address this conversion challenge from two perspectives: architectural alignment and representational alignment. We identify Test-Time Training (TTT) as a linear-complexity architecture whose two-layer dynamic formulation is structurally aligned with Softmax attention, enabling direct inheritance of pretrained attention weights. To further align representational properties, including key shift-invariance and locality, we introduce key instance normalization and a lightweight locality enhancement module. We validate our approach by linearizing Stable Diffusion 3.5 and introduce SD3.5-T$^5$ (Transformer To Test Time Training). With only 1 hour of fine-tuning on 4$\times$H20 GPUs, SD3.5-T$^5$ achieves comparable text-to-image quality to the fine-tuned Softmax model, while accelerating inference by 1.32$\times$ and 1.47$\times$ at 1K and 2K resolutions.

Yulin Wang, Yang Yue, Rui Lu et al. · tsinghua

The superior performance of modern visual backbones usually comes with a costly training procedure. We contribute to this issue by generalizing the idea of curriculum learning beyond its original formulation, i.e., training models using easier-to-harder data. Specifically, we reformulate the training curriculum as a soft-selection function, which uncovers progressively more difficult patterns within each example during training, instead of performing easier-to-harder sample selection. Our work is inspired by an intriguing observation on the learning dynamics of visual backbones: during the earlier stages of training, the model predominantly learns to recognize some 'easier-to-learn' discriminative patterns in the data. These patterns, when observed through frequency and spatial domains, incorporate lower-frequency components, and the natural image contents without distortion or data augmentation. Motivated by these findings, we propose a curriculum where the model always leverages all the training data at every learning stage, yet the exposure to the 'easier-to-learn' patterns of each example is initiated first, with harder patterns gradually introduced as training progresses. To implement this idea in a computationally efficient way, we introduce a cropping operation in the Fourier spectrum of the inputs, enabling the model to learn from only the lower-frequency components. Then we show that exposing the contents of natural images can be readily achieved by modulating the intensity of data augmentation. Finally, we integrate these aspects and design curriculum schedules with tailored search algorithms. The resulting method, EfficientTrain++, is simple, general, yet surprisingly effective. It reduces the training time of a wide variety of popular models by 1.5-3.0x on ImageNet-1K/22K without sacrificing accuracy. It also demonstrates efficacy in self-supervised learning (e.g., MAE).

Yulin Wang, Haoji Zhang, Yang Yue et al.

This paper presents a comprehensive exploration of the phenomenon of data redundancy in video understanding, with the aim to improve computational efficiency. Our investigation commences with an examination of spatial redundancy, which refers to the observation that the most informative region in each video frame usually corresponds to a small image patch, whose shape, size and location shift smoothly across frames. Motivated by this phenomenon, we formulate the patch localization problem as a dynamic decision task, and introduce a spatially adaptive video recognition approach, termed AdaFocus. In specific, a lightweight encoder is first employed to quickly process the full video sequence, whose features are then utilized by a policy network to identify the most task-relevant regions. Subsequently, the selected patches are inferred by a high-capacity deep network for the final prediction. The full model can be trained in end-to-end conveniently. Furthermore, AdaFocus can be extended by further considering temporal and sample-wise redundancies, i.e., allocating the majority of computation to the most task-relevant frames, and minimizing the computation spent on relatively "easier" videos. Our resulting approach, Uni-AdaFocus, establishes a comprehensive framework that seamlessly integrates spatial, temporal, and sample-wise dynamic computation, while it preserves the merits of AdaFocus in terms of efficient end-to-end training and hardware friendliness. In addition, Uni-AdaFocus is general and flexible as it is compatible with off-the-shelf efficient backbones (e.g., TSM and X3D), which can be readily deployed as our feature extractor, yielding a significantly improved computational efficiency. Empirically, extensive experiments based on seven benchmark datasets and three application scenarios substantiate that Uni-AdaFocus is considerably more efficient than the competitive baselines.

Qiushi Li, Celi Lou, Yanfang Cheng et al.

The resolving ability of wide-field fluorescence microscopy is fundamentally limited by out-of-focus background owing to its low axial resolution, particularly for densely labeled biological samples. To address this, we developed ET2dNet, a deep learning-based EPI-TIRF cross-modality network that achieves TIRF-comparable background subtraction and axial super-resolution from a single wide-field image without requiring hardware modifications. The model employs a physics-informed hybrid architecture, synergizing supervised learning with registered EPI-TIRF image pairs and self-supervised physical modeling via convolution with the point spread function. This framework ensures exceptional generalization across microscope objectives, enabling few-shot adaptation to new imaging setups. Rigorous validation on cellular and tissue samples confirms ET2dNet's superiority in background suppression and axial resolution enhancement, while maintaining compatibility with deconvolution techniques for lateral resolution improvement. Furthermore, by extending this paradigm through knowledge distillation, we developed ET3dNet, a dedicated three-dimensional reconstruction network that produces artifact-reduced volumetric results. ET3dNet effectively removes out-of-focus background signals even when the input image stack lacks the source of background. This framework makes axial super-resolution imaging more accessible by providing an easy-to-deploy algorithm that avoids additional hardware costs and complexity, showing great potential for live cell studies and clinical histopathology.

Fengxiang Wang, Mingshuo Chen, Yueying Li et al.

Ultra-high-resolution (UHR) remote sensing (RS) imagery offers valuable data for Earth observation but pose challenges for existing multimodal foundation models due to two key bottlenecks: (1) limited availability of UHR training data, and (2) token explosion caused by the large image size. To address data scarcity, we introduce SuperRS-VQA (avg. 8,376$\times$8,376) and HighRS-VQA (avg. 2,000$\times$1,912), the highest-resolution vision-language datasets in RS to date, covering 22 real-world dialogue tasks. To mitigate token explosion, our pilot studies reveal significant redundancy in RS images: crucial information is concentrated in a small subset of object-centric tokens, while pruning background tokens (e.g., ocean or forest) can even improve performance. Motivated by these findings, we propose two strategies: Background Token Pruning and Anchored Token Selection, to reduce the memory footprint while preserving key semantics.Integrating these techniques, we introduce GeoLLaVA-8K, the first RS-focused multimodal large language model capable of handling inputs up to 8K$\times$8K resolution, built on the LLaVA framework. Trained on SuperRS-VQA and HighRS-VQA, GeoLLaVA-8K sets a new state-of-the-art on the XLRS-Bench.

Zanlin Ni, Yulin Wang, Yeguo Hua et al.

Recent advances in image synthesis have been propelled by powerful generative models, such as Masked Generative Transformers (MaskGIT), autoregressive models, diffusion models, and rectified flow models. A common principle behind their success is the decomposition of synthesis into multiple steps. However, this introduces a proliferation of step-specific parameters (e.g., noise level or temperature at each step). Existing approaches typically rely on manually-designed rules to manage this complexity, demanding expert knowledge and trial-and-error. Furthermore, these static schedules lack the flexibility to adapt to the unique characteristics of each sample, yielding sub-optimal performance. To address this issue, we present AdaGen, a general, learnable, and sample-adaptive framework for scheduling the iterative generation process. Specifically, we formulate the scheduling problem as a Markov Decision Process, where a lightweight policy network determines suitable parameters given the current generation state, and can be trained through reinforcement learning. Importantly, we demonstrate that simple reward designs, such as FID or pre-trained reward models, can be easily hacked and may not reliably guarantee the desired quality or diversity of generated samples. Therefore, we propose an adversarial reward design to guide the training of the policy networks. Finally, we introduce an inference-time refinement strategy and a controllable fidelity-diversity trade-off mechanism to further enhance the performance and flexibility of AdaGen. Comprehensive experiments on four generative paradigms validate the superiority of AdaGen. For example, AdaGen achieves better performance on DiT-XL with 3 times lower inference cost and improves the FID of VAR from 1.92 to 1.59 with negligible computational overhead.

Fengxiang Wang, Mingshuo Chen, Yueying Li et al.

Multimodal reasoning for ultra-high-resolution (UHR) remote sensing (RS) is usually bottlenecked by visual evidence acquisition: the model necessitates localizing tiny task-relevant regions in massive pixel spaces. While Agentic Reinforcement Learning with Verifiable Rewards (RLVR) using zoom-in tools offers a path forward, we find that standard reinforcement learning struggles to navigate these vast visual spaces without structured domain priors. In this paper, we investigate the interplay between post-training paradigms: comparing Cold-start Supervised Fine-Tuning (SFT), RLVR, and Agentic RLVR on the UHR RS benchmark.Our controlled studies yield a counter-intuitive finding: high-quality Earth-science text-only QA is a primary driver of UHR visual reasoning gains. Despite lacking images, domain-specific text injects the concepts, mechanistic explanations, and decision rules necessary to guide visual evidence retrieval.Based on this, we propose a staged knowledge injection recipe: (1) cold-starting with scalable, knowledge-graph-verified Earth-science text QA to instill reasoning structures;and (2) "pre-warming" on the same hard UHR image-text examples during SFT to stabilize and amplify subsequent tool-based RL. This approach achieves a 60.40% Pass@1 on XLRS-Bench, significantly outperforming larger general purpose models (e.g., GPT-5.2, Gemini 3.0 Pro, Intern-S1) and establishing a new state-of-the-art.

Fengxiang Wang, Mingshuo Chen, Yueying Li et al.

The "thinking-with-images" paradigm enables multimodal large language models (MLLMs) to actively explore visual scenes via zoom-in tools. This is essential for ultra-high-resolution (UHR) remote sensing VQA, where task-relevant cues are sparse and tiny. However, we observe a consistent failure mode in existing zoom-enabled MLLMs: Tool Usage Homogenization, where tool calls collapse into task-agnostic patterns, limiting effective evidence acquisition. To address this, we propose GeoEyes, a staged training framework consisting of (1) a cold-start SFT dataset, UHR Chain-of-Zoom (UHR-CoZ), which covers diverse zooming regimes, and (2) an agentic reinforcement learning method, AdaZoom-GRPO, that explicitly rewards evidence gain and answer improvement during zoom interactions. The resulting model learns on-demand zooming with proper stopping behavior and achieves substantial improvements on UHR remote sensing benchmarks, with 54.23% accuracy on XLRS-Bench.

Dongchen Han, Tianzhu Ye, Zhuofan Xia et al.

Scaling up network depth is a fundamental pursuit in neural architecture design, as theory suggests that deeper models offer exponentially greater capability. Benefiting from the residual connections, modern neural networks can scale up to more than one hundred layers and enjoy wide success. However, as networks continue to deepen, current architectures often struggle to realize their theoretical capacity improvements, calling for more advanced designs to further unleash the potential of deeper networks. In this paper, we identify two key barriers that obstruct residual models from scaling deeper: shortcut degradation and limited width. Shortcut degradation hinders deep-layer learning, while the inherent depth-width trade-off imposes limited width. To mitigate these issues, we propose a generalized residual architecture dubbed Step by Step Network (StepsNet) to bridge the gap between theoretical potential and practical performance of deep models. Specifically, we separate features along the channel dimension and let the model learn progressively via stacking blocks with increasing width. The resulting method mitigates the two identified problems and serves as a versatile macro design applicable to various models. Extensive experiments show that our method consistently outperforms residual models across diverse tasks, including image classification, object detection, semantic segmentation, and language modeling. These results position StepsNet as a superior generalization of the widely adopted residual architecture.

Siqing Yuan, Yulin Wang, Zirui Cao et al.

Cardiomyopathy, a principal contributor to heart failure and sudden cardiac mortality, demands precise early screening. Cardiac Magnetic Resonance (CMR), recognized as the diagnostic 'gold standard' through multiparametric protocols, holds the potential to serve as an accurate screening tool. However, its reliance on gadolinium contrast and labor-intensive interpretation hinders population-scale deployment. We propose CC-CMR, a Contrastive Learning and Cross-Modal alignment framework for gadolinium-free cardiomyopathy screening using cine CMR sequences. By aligning the latent spaces of cine CMR and Late Gadolinium Enhancement (LGE) sequences, our model encodes fibrosis-specific pathology into cine CMR embeddings. A Feature Interaction Module concurrently optimizes diagnostic precision and cross-modal feature congruence, augmented by an uncertainty-guided adaptive training mechanism that dynamically calibrates task-specific objectives to ensure model generalizability. Evaluated on multi-center data from 231 subjects, CC-CMR achieves accuracy of 0.943 (95% CI: 0.886-0.986), outperforming state-of-the-art cine-CMR-only models by 4.3% while eliminating gadolinium dependency, demonstrating its clinical viability for wide range of populations and healthcare environments.

Chaoqun Du, Yulin Wang, Shiji Song et al.

Bayesian decision theory advocates the Bayes classifier as the optimal approach for minimizing the risk in machine learning problems. Current deep learning algorithms usually solve for the optimal classifier by \emph{implicitly} estimating the posterior probabilities, \emph{e.g.}, by minimizing the Softmax cross-entropy loss. This simple methodology has been proven effective for meticulously balanced academic benchmark datasets. However, it is not applicable to the long-tailed data distributions in the real world, where it leads to the gradient imbalance issue and fails to ensure the Bayes optimal decision rule. To address these challenges, this paper presents a novel approach (BAPE) that provides a more precise theoretical estimation of the data distributions by \emph{explicitly} modeling the parameters of the posterior probabilities and solving them with point estimation. Consequently, our method directly learns the Bayes classifier without gradient descent based on Bayes' theorem, simultaneously alleviating the gradient imbalance and ensuring the Bayes optimal decision rule. Furthermore, we propose a straightforward yet effective \emph{distribution adjustment} technique. This method enables the Bayes classifier trained from the long-tailed training set to effectively adapt to the test data distribution with an arbitrary imbalance factor, thereby enhancing performance without incurring additional computational costs. In addition, we demonstrate the gains of our method are orthogonal to existing learning approaches for long-tailed scenarios, as they are mostly designed under the principle of \emph{implicitly} estimating the posterior probabilities. Extensive empirical evaluations on CIFAR-10-LT, CIFAR-100-LT, ImageNet-LT, and iNaturalist demonstrate that our method significantly improves the generalization performance of popular deep networks, despite its simplicity.

Runqi Meng, Sifan Song, Pengfei Jin et al.

Accurate tumor segmentation is crucial for cancer diagnosis and treatment. While foundation models have advanced general-purpose segmentation, existing methods still struggle with: (1) limited incorporation of medical priors, (2) imbalance between generic and tumor-specific features, and (3) high computational costs for clinical adaptation. To address these challenges, we propose MAST-Pro (Mixture-of-experts for Adaptive Segmentation of pan-Tumors with knowledge-driven Prompts), a novel framework that integrates dynamic Mixture-of-Experts (D-MoE) and knowledge-driven prompts for pan-tumor segmentation. Specifically, text and anatomical prompts provide domain-specific priors, guiding tumor representation learning, while D-MoE dynamically selects experts to balance generic and tumor-specific feature learning, improving segmentation accuracy across diverse tumor types. To enhance efficiency, we employ Parameter-Efficient Fine-Tuning (PEFT), optimizing MAST-Pro with significantly reduced computational overhead. Experiments on multi-anatomical tumor datasets demonstrate that MAST-Pro outperforms state-of-the-art approaches, achieving up to a 5.20% improvement in average DSC while reducing trainable parameters by 91.04%, without compromising accuracy.

Yifan Pu, Yizeng Han, Yulin Wang et al.

Fine-grained image recognition is a longstanding computer vision challenge that focuses on differentiating objects belonging to multiple subordinate categories within the same meta-category. Since images belonging to the same meta-category usually share similar visual appearances, mining discriminative visual cues is the key to distinguishing fine-grained categories. Although commonly used image-level data augmentation techniques have achieved great success in generic image classification problems, they are rarely applied in fine-grained scenarios, because their random editing-region behavior is prone to destroy the discriminative visual cues residing in the subtle regions. In this paper, we propose diversifying the training data at the feature-level to alleviate the discriminative region loss problem. Specifically, we produce diversified augmented samples by translating image features along semantically meaningful directions. The semantic directions are estimated with a covariance prediction network, which predicts a sample-wise covariance matrix to adapt to the large intra-class variation inherent in fine-grained images. Furthermore, the covariance prediction network is jointly optimized with the classification network in a meta-learning manner to alleviate the degenerate solution problem. Experiments on four competitive fine-grained recognition benchmarks (CUB-200-2011, Stanford Cars, FGVC Aircrafts, NABirds) demonstrate that our method significantly improves the generalization performance on several popular classification networks (e.g., ResNets, DenseNets, EfficientNets, RegNets and ViT). Combined with a recently proposed method, our semantic data augmentation approach achieves state-of-the-art performance on the CUB-200-2011 dataset. The source code will be released.