Yunhong Wang

Zhihong Fu, Zehua Fu, Qingjie Liu et al.

Transformers have been successfully applied to the visual tracking task and significantly promote tracking performance. The self-attention mechanism designed to model long-range dependencies is the key to the success of Transformers. However, self-attention lacks focusing on the most relevant information in the search regions, making it easy to be distracted by background. In this paper, we relieve this issue with a sparse attention mechanism by focusing the most relevant information in the search regions, which enables a much accurate tracking. Furthermore, we introduce a double-head predictor to boost the accuracy of foreground-background classification and regression of target bounding boxes, which further improve the tracking performance. Extensive experiments show that, without bells and whistles, our method significantly outperforms the state-of-the-art approaches on LaSOT, GOT-10k, TrackingNet, and UAV123, while running at 40 FPS. Notably, the training time of our method is reduced by 75% compared to that of TransT. The source code and models are available at https://github.com/fzh0917/SparseTT.

Huanyu Zhou, Qingjie Liu, Yunhong Wang

Human action recognition aims at classifying the category of human action from a segment of a video. Recently, people have dived into designing GCN-based models to extract features from skeletons for performing this task, because skeleton representations are much more efficient and robust than other modalities such as RGB frames. However, when employing the skeleton data, some important clues like related items are also discarded. It results in some ambiguous actions that are hard to be distinguished and tend to be misclassified. To alleviate this problem, we propose an auxiliary feature refinement head (FR Head), which consists of spatial-temporal decoupling and contrastive feature refinement, to obtain discriminative representations of skeletons. Ambiguous samples are dynamically discovered and calibrated in the feature space. Furthermore, FR Head could be imposed on different stages of GCNs to build a multi-level refinement for stronger supervision. Extensive experiments are conducted on NTU RGB+D, NTU RGB+D 120, and NW-UCLA datasets. Our proposed models obtain competitive results from state-of-the-art methods and can help to discriminate those ambiguous samples. Codes are available at https://github.com/zhysora/FR-Head.

Weilai Xiang, Hongyu Yang, Di Huang et al.

Inspired by recent advances in diffusion models, which are reminiscent of denoising autoencoders, we investigate whether they can acquire discriminative representations for classification via generative pre-training. This paper shows that the networks in diffusion models, namely denoising diffusion autoencoders (DDAE), are unified self-supervised learners: by pre-training on unconditional image generation, DDAE has already learned strongly linear-separable representations within its intermediate layers without auxiliary encoders, thus making diffusion pre-training emerge as a general approach for generative-and-discriminative dual learning. To validate this, we conduct linear probe and fine-tuning evaluations. Our diffusion-based approach achieves 95.9% and 50.0% linear evaluation accuracies on CIFAR-10 and Tiny-ImageNet, respectively, and is comparable to contrastive learning and masked autoencoders for the first time. Transfer learning from ImageNet also confirms the suitability of DDAE for Vision Transformers, suggesting the potential to scale DDAEs as unified foundation models. Code is available at github.com/FutureXiang/ddae.

Huanyu Zhou, Qingjie Liu, Yunhong Wang

Pan-sharpening aims at producing a high-resolution (HR) multi-spectral (MS) image from a low-resolution (LR) multi-spectral (MS) image and its corresponding panchromatic (PAN) image acquired by a same satellite. Inspired by a new fashion in recent deep learning community, we propose a novel Transformer based model for pan-sharpening. We explore the potential of Transformer in image feature extraction and fusion. Following the successful development of vision transformers, we design a two-stream network with the self-attention to extract the modality-specific features from the PAN and MS modalities and apply a cross-attention module to merge the spectral and spatial features. The pan-sharpened image is produced from the enhanced fused features. Extensive experiments on GaoFen-2 and WorldView-3 images demonstrate that our Transformer based model achieves impressive results and outperforms many existing CNN based methods, which shows the great potential of introducing Transformer to the pan-sharpening task. Codes are available at https://github.com/zhysora/PanFormer.

Jinqing Zhang, Yanan Zhang, Qingjie Liu et al.

Recently, the pure camera-based Bird's-Eye-View (BEV) perception provides a feasible solution for economical autonomous driving. However, the existing BEV-based multi-view 3D detectors generally transform all image features into BEV features, without considering the problem that the large proportion of background information may submerge the object information. In this paper, we propose Semantic-Aware BEV Pooling (SA-BEVPool), which can filter out background information according to the semantic segmentation of image features and transform image features into semantic-aware BEV features. Accordingly, we propose BEV-Paste, an effective data augmentation strategy that closely matches with semantic-aware BEV feature. In addition, we design a Multi-Scale Cross-Task (MSCT) head, which combines task-specific and cross-task information to predict depth distribution and semantic segmentation more accurately, further improving the quality of semantic-aware BEV feature. Finally, we integrate the above modules into a novel multi-view 3D object detection framework, namely SA-BEV. Experiments on nuScenes show that SA-BEV achieves state-of-the-art performance. Code has been available at https://github.com/mengtan00/SA-BEV.git.

Wenrui Cai, Qingjie Liu, Yunhong Wang

Trackers that follow Siamese paradigm utilize similarity matching between template and search region features for tracking. Many methods have been explored to enhance tracking performance by incorporating tracking history to better handle scenarios involving target appearance variations such as deformation and occlusion. However, the utilization of historical information in existing methods is insufficient and incomprehensive, which typically requires repetitive training and introduces a large amount of computation. In this paper, we show that by providing a tracker that follows Siamese paradigm with precise and updated historical information, a significant performance improvement can be achieved with completely unchanged parameters. Based on this, we propose a historical prompt network that uses refined historical foreground masks and historical visual features of the target to provide comprehensive and precise prompts for the tracker. We build a novel tracker called HIPTrack based on the historical prompt network, which achieves considerable performance improvements without the need to retrain the entire model. We conduct experiments on seven datasets and experimental results demonstrate that our method surpasses the current state-of-the-art trackers on LaSOT, LaSOText, GOT-10k and NfS. Furthermore, the historical prompt network can seamlessly integrate as a plug-and-play module into existing trackers, providing performance enhancements. The source code is available at https://github.com/WenRuiCai/HIPTrack.

Zhiyuan Zhao, Qingjie Liu, Yunhong Wang

Recently, few-shot object detection~(FSOD) has received much attention from the community, and many methods are proposed to address this problem from a knowledge transfer perspective. Though promising results have been achieved, these methods fail to achieve shot-stable:~methods that excel in low-shot regimes are likely to struggle in high-shot regimes, and vice versa. We believe this is because the primary challenge of FSOD changes when the number of shots varies. In the low-shot regime, the primary challenge is the lack of inner-class variation. In the high-shot regime, as the variance approaches the real one, the main hindrance to the performance comes from misalignment between learned and true distributions. However, these two distinct issues remain unsolved in most existing FSOD methods. In this paper, we propose to overcome these challenges by exploiting rich knowledge the model has learned and effectively transferring them to the novel classes. For the low-shot regime, we propose a distribution calibration method to deal with the lack of inner-class variation problem. Meanwhile, a shift compensation method is proposed to compensate for possible distribution shift during fine-tuning. For the high-shot regime, we propose to use the knowledge learned from ImageNet as guidance for the feature learning in the fine-tuning stage, which will implicitly align the distributions of the novel classes. Although targeted toward different regimes, these two strategies can work together to further improve the FSOD performance. Experiments on both the VOC and COCO benchmarks show that our proposed method can significantly outperform the baseline method and produce competitive results in both low-shot settings (shot<5) and high-shot settings (shot>=5). Code is available at https://github.com/JulioZhao97/EffTrans_Fsdet.git.

Jinqing Zhang, Yanan Zhang, Yunlong Qi et al.

Bird's-Eye-View (BEV) representation has emerged as a mainstream paradigm for multi-view 3D object detection, demonstrating impressive perceptual capabilities. However, existing methods overlook the geometric quality of BEV representation, leaving it in a low-resolution state and failing to restore the authentic geometric information of the scene. In this paper, we identify the drawbacks of previous approaches that limit the geometric quality of BEV representation and propose Radial-Cartesian BEV Sampling (RC-Sampling), which outperforms other feature transformation methods in efficiently generating high-resolution dense BEV representation to restore fine-grained geometric information. Additionally, we design a novel In-Box Label to substitute the traditional depth label generated from the LiDAR points. This label reflects the actual geometric structure of objects rather than just their surfaces, injecting real-world geometric information into the BEV representation. In conjunction with the In-Box Label, Centroid-Aware Inner Loss (CAI Loss) is developed to capture the inner geometric structure of objects. Finally, we integrate the aforementioned modules into a novel multi-view 3D object detector, dubbed GeoBEV, which achieves a state-of-the-art result of 66.2\% NDS on the nuScenes test set. The code is available at https://github.com/mengtan00/GeoBEV.git.

Zheng Jiang, Jinqing Zhang, Yanan Zhang et al.

Although multi-view 3D object detection based on the Bird's-Eye-View (BEV) paradigm has garnered widespread attention as an economical and deployment-friendly perception solution for autonomous driving, there is still a performance gap compared to LiDAR-based methods. In recent years, several cross-modal distillation methods have been proposed to transfer beneficial information from teacher models to student models, with the aim of enhancing performance. However, these methods face challenges due to discrepancies in feature distribution originating from different data modalities and network structures, making knowledge transfer exceptionally challenging. In this paper, we propose a Foreground Self-Distillation (FSD) scheme that effectively avoids the issue of distribution discrepancies, maintaining remarkable distillation effects without the need for pre-trained teacher models or cumbersome distillation strategies. Additionally, we design two Point Cloud Intensification (PCI) strategies to compensate for the sparsity of point clouds by frame combination and pseudo point assignment. Finally, we develop a Multi-Scale Foreground Enhancement (MSFE) module to extract and fuse multi-scale foreground features by predicted elliptical Gaussian heatmap, further improving the model's performance. We integrate all the above innovations into a unified framework named FSD-BEV. Extensive experiments on the nuScenes dataset exhibit that FSD-BEV achieves state-of-the-art performance, highlighting its effectiveness. The code and models are available at: https://github.com/CocoBoom/fsd-bev.

Guodong Wang, Yunhong Wang, Jie Qin et al.

Video Anomaly Detection (VAD) is an important topic in computer vision. Motivated by the recent advances in self-supervised learning, this paper addresses VAD by solving an intuitive yet challenging pretext task, i.e., spatio-temporal jigsaw puzzles, which is cast as a multi-label fine-grained classification problem. Our method exhibits several advantages over existing works: 1) the spatio-temporal jigsaw puzzles are decoupled in terms of spatial and temporal dimensions, responsible for capturing highly discriminative appearance and motion features, respectively; 2) full permutations are used to provide abundant jigsaw puzzles covering various difficulty levels, allowing the network to distinguish subtle spatio-temporal differences between normal and abnormal events; and 3) the pretext task is tackled in an end-to-end manner without relying on any pre-trained models. Our method outperforms state-of-the-art counterparts on three public benchmarks. Especially on ShanghaiTech Campus, the result is superior to reconstruction and prediction-based methods by a large margin.

Rui He, Zehua Fu, Qingjie Liu et al.

Tracking multiple athletes in sports videos is a very challenging Multi-Object Tracking (MOT) task, since athletes often have the same appearance and are intimately covered with each other, making a common occlusion problem becomes an abhorrent duplicate detection. In this paper, the duplicate detection is newly and precisely defined as occlusion misreporting on the same athlete by multiple detection boxes in one frame. To address this problem, we meticulously design a novel transformer-based Duplicate Detection Decontaminator (D$^3$) for training, and a specific algorithm Rally-Hungarian (RH) for matching. Once duplicate detection occurs, D$^3$ immediately modifies the procedure by generating enhanced boxes losses. RH, triggered by the team sports substitution rules, is exceedingly suitable for sports videos. Moreover, to complement the tracking dataset that without shot changes, we release a new dataset based on sports video named RallyTrack. Extensive experiments on RallyTrack show that combining D$^3$ and RH can dramatically improve the tracking performance with 9.2 in MOTA and 4.5 in HOTA. Meanwhile, experiments on MOT-series and DanceTrack discover that D$^3$ can accelerate convergence during training, especially save up to 80 percent of the original training time on MOT17. Finally, our model, which is trained only with volleyball videos, can be applied directly to basketball and soccer videos for MAT, which shows priority of our method. Our dataset is available at https://github.com/heruihr/rallytrack.

Guozhen Peng, Yunhong Wang, Yuwei Zhao et al.

Gait recognition has attracted increasing attention from academia and industry as a human recognition technology from a distance in non-intrusive ways without requiring cooperation. Although advanced methods have achieved impressive success in lab scenarios, most of them perform poorly in the wild. Recently, some Convolution Neural Networks (ConvNets) based methods have been proposed to address the issue of gait recognition in the wild. However, the temporal receptive field obtained by convolution operations is limited for long gait sequences. If directly replacing convolution blocks with visual transformer blocks, the model may not enhance a local temporal receptive field, which is important for covering a complete gait cycle. To address this issue, we design a Global-Local Temporal Receptive Field Network (GLGait). GLGait employs a Global-Local Temporal Module (GLTM) to establish a global-local temporal receptive field, which mainly consists of a Pseudo Global Temporal Self-Attention (PGTA) and a temporal convolution operation. Specifically, PGTA is used to obtain a pseudo global temporal receptive field with less memory and computation complexity compared with a multi-head self-attention (MHSA). The temporal convolution operation is used to enhance the local temporal receptive field. Besides, it can also aggregate pseudo global temporal receptive field to a true holistic temporal receptive field. Furthermore, we also propose a Center-Augmented Triplet Loss (CTL) in GLGait to reduce the intra-class distance and expand the positive samples in the training stage. Extensive experiments show that our method obtains state-of-the-art results on in-the-wild datasets, $i.e.$, Gait3D and GREW. The code is available at https://github.com/bgdpgz/GLGait.

Zongye Zhang, Yuzhuo Cui, Qingjie Liu et al.

Generalizing motion representation across diverse characters remains challenging due to significant topological variations in skeletal structures across datasets and species, which hinder the development of scalable generative models. To bridge this gap, we propose a Semantic-Aware Topology-Agnostic framework that learns a unified latent manifold shared by disparate species. Unlike methods relying on fixed hierarchies or rigid padding strategies, our approach leverages a semantic modulation mechanism to align functional joint correspondences, thereby decoupling motion from topology. This design enables the construction of a continuous, generative-friendly motion space from large-scale, unaligned raw BVH data. Experiments on human and animal datasets demonstrate that our framework achieves high-fidelity reconstruction and supports downstream text-to-motion tasks. Notably, the model enables zero-shot cross-species retargeting without paired data. Code and demos are available at: https://github.com/zzysteve/SATA

Ziyue Huang, Yongchao Feng, Qingjie Liu et al.

Detection pre-training methods for the DETR series detector have been extensively studied in natural scenes, e.g., DETReg. However, the detection pre-training remains unexplored in remote sensing scenes. In existing pre-training methods, alignment between object embeddings extracted from a pre-trained backbone and detector features is significant. However, due to differences in feature extraction methods, a pronounced feature discrepancy still exists and hinders the pre-training performance. The remote sensing images with complex environments and more densely distributed objects exacerbate the discrepancy. In this work, we propose a novel Mutually optimizing pre-training framework for remote sensing object Detection, dubbed as MutDet. In MutDet, we propose a systemic solution against this challenge. Firstly, we propose a mutual enhancement module, which fuses the object embeddings and detector features bidirectionally in the last encoder layer, enhancing their information interaction.Secondly, contrastive alignment loss is employed to guide this alignment process softly and simultaneously enhances detector features' discriminativity. Finally, we design an auxiliary siamese head to mitigate the task gap arising from the introduction of enhancement module. Comprehensive experiments on various settings show new state-of-the-art transfer performance. The improvement is particularly pronounced when data quantity is limited. When using 10% of the DIOR-R data, MutDet improves DetReg by 6.1% in AP50. Codes and models are available at: https://github.com/floatingstarZ/MutDet.

Zihao Cheng, Zeming Liu, Yingyu Shan et al.

While large language model--powered agents can self-evolve by accumulating experience or by dynamically creating new assets (i.e., tools or expert agents), existing frameworks typically treat these two evolutionary processes in isolation. This separation overlooks their intrinsic interdependence: the former is inherently bounded by a manually predefined static toolset, while the latter generates new assets from scratch without experiential guidance, leading to limited capability growth and unstable evolution. To address this limitation, we introduce a novel paradigm of co-evolutionary Capability Expansion and Experience Distillation. Guided by this paradigm, we propose the \textbf{Mem$^{\textbf{2}}$Evolve}, which integrates two core components: \textbf{Experience Memory} and \textbf{Asset Memory}. Specifically, Mem$^{2}$Evolve leverages accumulated experience to guide the dynamic creation of assets, thereby expanding the agent's capability space while simultaneously acquiring new experience to achieve co-evolution. Extensive experiments across 6 task categories and 8 benchmarks demonstrate that Mem$^{2}$Evolve achieves improvement of 18.53\% over standard LLMs, 11.80\% over agents evolving solely through experience, and 6.46\% over those evolving solely through asset creation, establishing it as a substantially more effective and stable self-evolving agent framework. Code is available at: https://buaa-irip-llm.github.io/Mem2Evolve.

Jinqing Zhang, Zehua Fu, Zelin Xu et al.

The comprehensive understanding capabilities of world models for driving scenarios have significantly improved the planning accuracy of end-to-end autonomous driving frameworks. However, the redundant modeling of static regions and the lack of deep interaction with trajectories hinder world models from exerting their full effectiveness. In this paper, we propose Temporal Residual World Model (TR-World), which focuses on dynamic object modeling. By calculating the temporal residuals of scene representations, the information of dynamic objects can be extracted without relying on detection and tracking. TR-World takes only temporal residuals as input, thus predicting the future spatial distribution of dynamic objects more precisely. By combining the prediction with the static object information contained in the current BEV features, accurate future BEV features can be obtained. Furthermore, we propose Future-Guided Trajectory Refinement (FGTR) module, which conducts interaction between prior trajectories (predicted from the current scene representation) and the future BEV features. This module can not only utilize future road conditions to refine trajectories, but also provides sparse spatial-temporal supervision on future BEV features to prevent world model collapse. Comprehensive experiments conducted on the nuScenes and NAVSIM datasets demonstrate that our method, namely ResWorld, achieves state-of-the-art planning performance. The code is available at https://github.com/mengtan00/ResWorld.git.

Yibai Liu, Shihang Wang, Zeming Liu et al.

Despite large language models (LLMs) have achieved impressive achievements across numerous tasks, supervised fine-tuning (SFT) remains essential for adapting these models to specialized domains. However, SFT for domain specialization can be resource-intensive and sometimes leads to a deterioration in performance over general capabilities due to catastrophic forgetting (CF). To address these issues, we propose a self-adaptive gradient-aware data selection approach (GrADS) for supervised fine-tuning of LLMs, which identifies effective subsets of training data by analyzing gradients obtained from a preliminary training phase. Specifically, we design self-guided criteria that leverage the magnitude and statistical distribution of gradients to prioritize examples that contribute the most to the model's learning process. This approach enables the acquisition of representative samples that enhance LLMs understanding of domain-specific tasks. Through extensive experimentation with various LLMs across diverse domains such as medicine, law, and finance, GrADS has demonstrated significant efficiency and cost-effectiveness. Remarkably, utilizing merely 5% of the selected GrADS data, LLMs already surpass the performance of those fine-tuned on the entire dataset, and increasing to 50% of the data results in significant improvements! With catastrophic forgetting substantially mitigated simultaneously. We will release our code for GrADS later.

Jiankai Li, Yunhong Wang, Weixin Li

Scene Graph Generation (SGG) plays a pivotal role in downstream vision-language tasks. Existing SGG methods typically suffer from poor compositional generalizations on unseen triplets. They are generally trained on incompletely annotated scene graphs that contain dominant triplets and tend to bias toward these seen triplets during inference. To address this issue, we propose a Triplet Calibration and Reduction (T-CAR) framework in this paper. In our framework, a triplet calibration loss is first presented to regularize the representations of diverse triplets and to simultaneously excavate the unseen triplets in incompletely annotated training scene graphs. Moreover, the unseen space of scene graphs is usually several times larger than the seen space since it contains a huge number of unrealistic compositions. Thus, we propose an unseen space reduction loss to shift the attention of excavation to reasonable unseen compositions to facilitate the model training. Finally, we propose a contextual encoder to improve the compositional generalizations of unseen triplets by explicitly modeling the relative spatial relations between subjects and objects. Extensive experiments show that our approach achieves consistent improvements for zero-shot SGG over state-of-the-art methods. The code is available at https://github.com/jkli1998/T-CAR.

Mingming Zhang, Ye Du, Zhenghui Hu et al.

Extracting building footprints from remote sensing images has been attracting extensive attention recently. Dominant approaches address this challenging problem by generating vectorized building masks with cumbersome refinement stages, which limits the application of such methods. In this paper, we introduce a new refinement-free and end-to-end building footprint extraction method, which is conceptually intuitive, simple, and effective. Our method, termed as BiSVP, represents a building instance with ordered vertices and formulates the building footprint extraction as predicting the serialized vertices directly in a bidirectional fashion. Moreover, we propose a cross-scale feature fusion (CSFF) module to facilitate high resolution and rich semantic feature learning, which is essential for the dense building vertex prediction task. Without bells and whistles, our BiSVP outperforms state-of-the-art methods by considerable margins on three building instance segmentation benchmarks, clearly demonstrating its superiority. The code and datasets will be made public available.

Yajie Liu, Pu Ge, Qingjie Liu et al.

How to effectively leverage the plentiful existing datasets to train a robust and high-performance model is of great significance for many practical applications. However, a model trained on a naive merge of different datasets tends to obtain poor performance due to annotation conflicts and domain divergence.In this paper, we attempt to train a unified model that is expected to perform well across domains on several popularity segmentation datasets.We conduct a detailed analysis of the impact on model generalization from three aspects of data augmentation, training strategies, and model capacity.Based on the analysis, we propose a robust solution that is able to improve model generalization across domains.Our solution ranks 2nd on RVC 2022 semantic segmentation task, with a dataset only 1/3 size of the 1st model used.

Zhuguanyu Wu, Jiaxin Chen, Hanwen Zhong et al.

Vision Transformer (ViT) has become one of the most prevailing fundamental backbone networks in the computer vision community. Despite the high accuracy, deploying it in real applications raises critical challenges including the high computational cost and inference latency. Recently, the post-training quantization (PTQ) technique has emerged as a promising way to enhance ViT's efficiency. Nevertheless, existing PTQ approaches for ViT suffer from the inflexible quantization on the post-Softmax and post-GELU activations that obey the power-law-like distributions. To address these issues, we propose a novel non-uniform quantizer, dubbed the Adaptive Logarithm AdaLog (AdaLog) quantizer. It optimizes the logarithmic base to accommodate the power-law-like distribution of activations, while simultaneously allowing for hardware-friendly quantization and de-quantization. By employing the bias reparameterization, the AdaLog quantizer is applicable to both the post-Softmax and post-GELU activations. Moreover, we develop an efficient Fast Progressive Combining Search (FPCS) strategy to determine the optimal logarithm base for AdaLog, as well as the scaling factors and zero points for the uniform quantizers. Extensive experimental results on public benchmarks demonstrate the effectiveness of our approach for various ViT-based architectures and vision tasks including classification, object detection, and instance segmentation. Code is available at https://github.com/GoatWu/AdaLog.

Rui Cai, Jun Guo, Xinze He et al.

In this report, we introduce Xiaomi-Robotics-0, an advanced vision-language-action (VLA) model optimized for high performance and fast and smooth real-time execution. The key to our method lies in a carefully designed training recipe and deployment strategy. Xiaomi-Robotics-0 is first pre-trained on large-scale cross-embodiment robot trajectories and vision-language data, endowing it with broad and generalizable action-generation capabilities while avoiding catastrophic forgetting of the visual-semantic knowledge of the underlying pre-trained VLM. During post-training, we propose several techniques for training the VLA model for asynchronous execution to address the inference latency during real-robot rollouts. During deployment, we carefully align the timesteps of consecutive predicted action chunks to ensure continuous and seamless real-time rollouts. We evaluate Xiaomi-Robotics-0 extensively in simulation benchmarks and on two challenging real-robot tasks that require precise and dexterous bimanual manipulation. Results show that our method achieves state-of-the-art performance across all simulation benchmarks. Moreover, Xiaomi-Robotics-0 can roll out fast and smoothly on real robots using a consumer-grade GPU, achieving high success rates and throughput on both real-robot tasks. To facilitate future research, code and model checkpoints are open-sourced at https://xiaomi-robotics-0.github.io

Guodong Wang, Yunhong Wang, Jie Qin et al.

Anomaly detection (AD), aiming to find samples that deviate from the training distribution, is essential in safety-critical applications. Though recent self-supervised learning based attempts achieve promising results by creating virtual outliers, their training objectives are less faithful to AD which requires a concentrated inlier distribution as well as a dispersive outlier distribution. In this paper, we propose Unilaterally Aggregated Contrastive Learning with Hierarchical Augmentation (UniCon-HA), taking into account both the requirements above. Specifically, we explicitly encourage the concentration of inliers and the dispersion of virtual outliers via supervised and unsupervised contrastive losses, respectively. Considering that standard contrastive data augmentation for generating positive views may induce outliers, we additionally introduce a soft mechanism to re-weight each augmented inlier according to its deviation from the inlier distribution, to ensure a purified concentration. Moreover, to prompt a higher concentration, inspired by curriculum learning, we adopt an easy-to-hard hierarchical augmentation strategy and perform contrastive aggregation at different depths of the network based on the strengths of data augmentation. Our method is evaluated under three AD settings including unlabeled one-class, unlabeled multi-class, and labeled multi-class, demonstrating its consistent superiority over other competitors.

Yizhou Jin, Yuezhu Feng, Jinjin Zhang et al.

Multimodal large language models (MLLMs) have recently demonstrated remarkable reasoning and perceptual abilities for anomaly detection. However, most approaches remain confined to image-level anomaly detection and textual reasoning, while pixel-level localization still relies on external vision modules and dense annotations. In this work, we activate the intrinsic reasoning potential of MLLMs to perform anomaly detection, pixel-level localization, and interpretable reasoning solely from image-level supervision, without any auxiliary components or pixel-wise labels. Specifically, we propose Reasoning-Driven Anomaly Localization (ReAL), which extracts anomaly-related tokens from the autoregressive reasoning process and aggregates their attention responses to produce pixel-level anomaly maps. We further introduce a Consistency-Guided Reasoning Optimization (CGRO) module that leverages reinforcement learning to align reasoning tokens with visual attentions, resulting in more coherent reasoning and accurate anomaly localization. Extensive experiments on four public benchmarks demonstrate that our method significantly improves anomaly detection, localization, and interpretability. Remarkably, despite relying solely on image-level supervision, our approach achieves performance competitive with MLLM-based methods trained under dense pixel-level supervision. Code is available at https://github.com/YizhouJin313/ReADL.

Mingming Zhang, Qingjie Liu, Yunhong Wang

Learning representations through self-supervision on unlabeled data has proven highly effective for understanding diverse images. However, remote sensing images often have complex and densely populated scenes with multiple land objects and no clear foreground objects. This intrinsic property generates high object density, resulting in false positive pairs or missing contextual information in self-supervised learning. To address these problems, we propose a context-enhanced masked image modeling method (CtxMIM), a simple yet efficient MIM-based self-supervised learning for remote sensing image understanding. CtxMIM formulates original image patches as a reconstructive template and employs a Siamese framework to operate on two sets of image patches. A context-enhanced generative branch is introduced to provide contextual information through context consistency constraints in the reconstruction. With the simple and elegant design, CtxMIM encourages the pre-training model to learn object-level or pixel-level features on a large-scale dataset without specific temporal or geographical constraints. Finally, extensive experiments show that features learned by CtxMIM outperform fully supervised and state-of-the-art self-supervised learning methods on various downstream tasks, including land cover classification, semantic segmentation, object detection, and instance segmentation. These results demonstrate that CtxMIM learns impressive remote sensing representations with high generalization and transferability. Code and data will be made public available.

Mingming Zhang, Qingjie Liu, Yunhong Wang

Deep learning-based methods have been extensively explored for automatic building mapping from high-resolution remote sensing images over recent years. While most building mapping models produce vector polygons of buildings for geographic and mapping systems, dominant methods typically decompose polygonal building extraction in some sub-problems, including segmentation, polygonization, and regularization, leading to complex inference procedures, low accuracy, and poor generalization. In this paper, we propose a simple and novel building mapping method with Hierarchical Transformers, called HiT, improving polygonal building mapping quality from high-resolution remote sensing images. HiT builds on a two-stage detection architecture by adding a polygon head parallel to classification and bounding box regression heads. HiT simultaneously outputs building bounding boxes and vector polygons, which is fully end-to-end trainable. The polygon head formulates a building polygon as serialized vertices with the bidirectional characteristic, a simple and elegant polygon representation avoiding the start or end vertex hypothesis. Under this new perspective, the polygon head adopts a transformer encoder-decoder architecture to predict serialized vertices supervised by the designed bidirectional polygon loss. Furthermore, a hierarchical attention mechanism combined with convolution operation is introduced in the encoder of the polygon head, providing more geometric structures of building polygons at vertex and edge levels. Comprehensive experiments on two benchmarks (the CrowdAI and Inria datasets) demonstrate that our method achieves a new state-of-the-art in terms of instance segmentation and polygonal metrics compared with state-of-the-art methods. Moreover, qualitative results verify the superiority and effectiveness of our model under complex scenes.

Yongchao Feng, Shiwei Li, Yingjie Gao et al.

Though feature-alignment based Domain Adaptive Object Detection (DAOD) methods have achieved remarkable progress, they ignore the source bias issue, i.e., the detector tends to acquire more source-specific knowledge, impeding its generalization capabilities in the target domain. Furthermore, these methods face a more formidable challenge in achieving consistent classification and localization in the target domain compared to the source domain. To overcome these challenges, we propose a novel Distillation-based Source Debiasing (DSD) framework for DAOD, which can distill domain-agnostic knowledge from a pre-trained teacher model, improving the detector's performance on both domains. In addition, we design a Target-Relevant Object Localization Network (TROLN), which can mine target-related localization information from source and target-style mixed data. Accordingly, we present a Domain-aware Consistency Enhancing (DCE) strategy, in which these information are formulated into a new localization representation to further refine classification scores in the testing stage, achieving a harmonization between classification and localization. Extensive experiments have been conducted to manifest the effectiveness of this method, which consistently improves the strong baseline by large margins, outperforming existing alignment-based works.

Zehua Fu, Wenhang Zuo, Zhenghui Hu et al.

Human pose estimation and tracking are fundamental tasks for understanding human behaviors in videos. Existing top-down framework-based methods usually perform three-stage tasks: human detection, pose estimation and tracking. Although promising results have been achieved, these methods rely heavily on high-performance detectors and may fail to track persons who are occluded or miss-detected. To overcome these problems, in this paper, we develop a novel keypoint confidence network and a tracking pipeline to improve human detection and pose estimation in top-down approaches. Specifically, the keypoint confidence network is designed to determine whether each keypoint is occluded, and it is incorporated into the pose estimation module. In the tracking pipeline, we propose the Bbox-revision module to reduce missing detection and the ID-retrieve module to correct lost trajectories, improving the performance of the detection stage. Experimental results show that our approach is universal in human detection and pose estimation, achieving state-of-the-art performance on both PoseTrack 2017 and 2018 datasets.

Junfu Wang, Yuanfang Guo, Liang Yang et al.

Graph Neural Networks (GNNs) have shown remarkable success in graph representation learning. Unfortunately, current weight assignment schemes in standard GNNs, such as the calculation based on node degrees or pair-wise representations, can hardly be effective in processing the networks with heterophily, in which the connected nodes usually possess different labels or features. Existing heterophilic GNNs tend to ignore the modeling of heterophily of each edge, which is also a vital part in tackling the heterophily problem. In this paper, we firstly propose a heterophily-aware attention scheme and reveal the benefits of modeling the edge heterophily, i.e., if a GNN assigns different weights to edges according to different heterophilic types, it can learn effective local attention patterns, which enable nodes to acquire appropriate information from distinct neighbors. Then, we propose a novel Heterophily-Aware Graph Attention Network (HA-GAT) by fully exploring and utilizing the local distribution as the underlying heterophily, to handle the networks with different homophily ratios. To demonstrate the effectiveness of the proposed HA-GAT, we analyze the proposed heterophily-aware attention scheme and local distribution exploration, by seeking for an interpretation from their mechanism. Extensive results demonstrate that our HA-GAT achieves state-of-the-art performances on eight datasets with different homophily ratios in both the supervised and semi-supervised node classification tasks.

Junfu Wang, Yuanfang Guo, Liang Yang et al.

Graph Neural Networks (GNNs) have been generalized to process the heterogeneous graphs by various approaches. Unfortunately, these approaches usually model the heterogeneity via various complicated modules. This paper aims to propose a simple yet effective framework to assign adequate ability to the homogeneous GNNs to handle the heterogeneous graphs. Specifically, we propose Relation Embedding based Graph Neural Network (RE-GNN), which employs only one parameter per relation to embed the importance of distinct types of relations and node-type-specific self-loop connections. To optimize these relation embeddings and the model parameters simultaneously, a gradient scaling factor is proposed to constrain the embeddings to converge to suitable values. Besides, we interpret the proposed RE-GNN from two perspectives, and theoretically demonstrate that our RE-GCN possesses more expressive power than GTN (which is a typical heterogeneous GNN, and it can generate meta-paths adaptively). Extensive experiments demonstrate that our RE-GNN can effectively and efficiently handle the heterogeneous graphs and can be applied to various homogeneous GNNs.

Wenshuai Xu, Zhenghui Hu, Yu Lu et al.

The pretraining-finetuning paradigm has gained popularity in various computer vision tasks. In this paradigm, the emergence of active finetuning arises due to the abundance of large-scale data and costly annotation requirements. Active finetuning involves selecting a subset of data from an unlabeled pool for annotation, facilitating subsequent finetuning. However, the use of a limited number of training samples can lead to a biased distribution, potentially resulting in model overfitting. In this paper, we propose a new method called ActiveDC for the active finetuning tasks. Firstly, we select samples for annotation by optimizing the distribution similarity between the subset to be selected and the entire unlabeled pool in continuous space. Secondly, we calibrate the distribution of the selected samples by exploiting implicit category information in the unlabeled pool. The feature visualization provides an intuitive sense of the effectiveness of our approach to distribution calibration. We conducted extensive experiments on three image classification datasets with different sampling ratios. The results indicate that ActiveDC consistently outperforms the baseline performance in all image classification tasks. The improvement is particularly significant when the sampling ratio is low, with performance gains of up to 10%. Our code will be released.

Junfu Wang, Yuanfang Guo, Liang Yang et al.

The current success of Graph Neural Networks (GNNs) usually relies on loading the entire attributed graph for processing, which may not be satisfied with limited memory resources, especially when the attributed graph is large. This paper pioneers to propose a Binary Graph Convolutional Network (Bi-GCN), which binarizes both the network parameters and input node attributes and exploits binary operations instead of floating-point matrix multiplications for network compression and acceleration. Meanwhile, we also propose a new gradient approximation based back-propagation method to properly train our Bi-GCN. According to the theoretical analysis, our Bi-GCN can reduce the memory consumption by an average of ~31x for both the network parameters and input data, and accelerate the inference speed by an average of ~51x, on three citation networks, i.e., Cora, PubMed, and CiteSeer. Besides, we introduce a general approach to generalize our binarization method to other variants of GNNs, and achieve similar efficiencies. Although the proposed Bi-GCN and Bi-GNNs are simple yet efficient, these compressed networks may also possess a potential capacity problem, i.e., they may not have enough storage capacity to learn adequate representations for specific tasks. To tackle this capacity problem, an Entropy Cover Hypothesis is proposed to predict the lower bound of the width of Bi-GNN hidden layers. Extensive experiments have demonstrated that our Bi-GCN and Bi-GNNs can give comparable performances to the corresponding full-precision baselines on seven node classification datasets and verified the effectiveness of our Entropy Cover Hypothesis for solving the capacity problem.

Ziyue Huang, Yupeng He, Qingjie Liu et al.

In contrast to the incremental classification task, the incremental detection task is characterized by the presence of data ambiguity, as an image may have differently labeled bounding boxes across multiple continuous learning stages. This phenomenon often impairs the model's ability to effectively learn new classes. However, existing research has paid less attention to the forward compatibility of the model, which limits its suitability for incremental learning. To overcome this obstacle, we propose leveraging a visual-language model such as CLIP to generate text feature embeddings for different class sets, which enhances the feature space globally. We then employ super-classes to replace the unavailable novel classes in the early learning stage to simulate the incremental scenario. Finally, we utilize the CLIP image encoder to accurately identify potential objects. We incorporate the finely recognized detection boxes as pseudo-annotations into the training process, thereby further improving the detection performance. We evaluate our approach on various incremental learning settings using the PASCAL VOC 2007 dataset, and our approach outperforms state-of-the-art methods, particularly for recognizing the new classes.

Shigeki Matsutani, Yoshiyuki Shimosako, Yunhong Wang

We numerically investigate the electric potential distribution over a two-dimensional continuum percolation model between the electrodes. The model consists of overlapped conductive particles on the background with an infinitesimal conductivity. Using the finite difference method, we solve the generalized Laplace equation and show that in the potential distribution, there appear the {\it{quasi-equipotential clusters}} which approximately and locally have the same values like steps and stairs. Since the quasi-equipotential clusters has the fractal structure, we compute the fractal dimension of equipotential curves and its dependence on the volume fraction over $[0,1]$. The fractal dimension in [1.00, 1.257] has a peak at the percolation threshold $p_c$.

Ziyue Huang, Mingming Zhang, Qingjie Liu et al.

The field of building detection from remote sensing images has made significant progress, but faces challenges in achieving high-accuracy detection due to the diversity in building appearances and the complexity of vast scenes. To address these challenges, we propose a novel approach called Context-Enhanced Detector (CEDet). Our approach utilizes a three-stage cascade structure to enhance the extraction of contextual information and improve building detection accuracy. Specifically, we introduce two modules: the Semantic Guided Contextual Mining (SGCM) module, which aggregates multi-scale contexts and incorporates an attention mechanism to capture long-range interactions, and the Instance Context Mining Module (ICMM), which captures instance-level relationship context by constructing a spatial relationship graph and aggregating instance features. Additionally, we introduce a semantic segmentation loss based on pseudo-masks to guide contextual information extraction. Our method achieves state-of-the-art performance on three building detection benchmarks, including CNBuilding-9P, CNBuilding-23P, and SpaceNet.

Wenrui Cai, Zhenyi Lu, Yuzhe Li et al.

With the advent of Transformer-based one-stream trackers that possess strong capability in inter-frame relation modeling, recent research has increasingly focused on how to introduce spatio-temporal context. However, most existing methods rely on a limited number of historical frames, which not only leads to insufficient utilization of the context, but also inevitably increases the length of input and incurs prohibitive computational overhead. Methods that query an external memory bank, on the other hand, suffer from inadequate fusion between the retrieved spatio-temporal features and the backbone. Moreover, using discrete historical frames as context overlooks the rich dynamics of the target. To address the issues, we propose Uni-MDTrack, which consists of two core components: Memory-Aware Compression Prompt (MCP) module and Dynamic State Fusion (DSF) module. MCP effectively compresses rich memory features into memory-aware prompt tokens, which deeply interact with the input throughout the entire backbone, significantly enhancing the performance while maintaining a stable computational load. DSF complements the discrete memory by capturing the continuous dynamic, progressively introducing the updated dynamic state features from shallow to deep layers, while also preserving high efficiency. Uni-MDTrack also supports unified tracking across RGB, RGB-D/T/E, and RGB-Language modalities. Experiments show that in Uni-MDTrack, training only the MCP, DSF, and prediction head, keeping the proportion of trainable parameters around 30%, yields substantial performance gains, achieves state-of-the-art results on 10 datasets spanning five modalities. Furthermore, both MCP and DSF exhibit excellent generality, functioning as plug-and-play components that can boost the performance of various baseline trackers, while significantly outperforming existing parameter-efficient training approaches.

Hongxi Yan, Qingjie Liu, Yunhong Wang

Large Reasoning Models (LRMs) excel at complex reasoning tasks through extended chain-of-thought generation, but their reliance on lengthy intermediate steps incurs substantial computational cost. We find that the entropy of the model's output distribution in early reasoning steps reliably distinguishes correct from incorrect reasoning. Motivated by this observation, we propose EntroCut, a training-free method that dynamically truncates reasoning by identifying high-confidence states where reasoning can be safely terminated. To comprehensively evaluate the trade-off between efficiency and accuracy, we introduce the Efficiency-Performance Ratio (EPR), a unified metric that quantifies relative token savings per unit accuracy loss. Experiments on four benchmarks show that EntroCut reduces token usage by up to 40\% with minimal accuracy sacrifice, achieving superior efficiency-performance trade-offs compared with existing training-free methods. These results demonstrate that entropy-guided dynamic truncation provides a practical approach to mitigate the inefficiency of LRMs.

Shuai Yang, Ziyue Huang, Jiaxin Chen et al.

Open-vocabulary object detection in remote sensing commonly relies on text-only prompting to specify target categories, implicitly assuming that inference-time category queries can be reliably grounded through pretraining-induced text-visual alignment. In practice, this assumption often breaks down in remote sensing scenarios due to task- and application-specific category semantics, resulting in unstable category specification under open-vocabulary settings. To address this limitation, we propose RS-MPOD, a multimodal open-vocabulary detection framework that reformulates category specification beyond text-only prompting by incorporating instance-grounded visual prompts, textual prompts, and their multimodal integration. RS-MPOD introduces a visual prompt encoder to extract appearance-based category cues from exemplar instances, enabling text-free category specification, and a multimodal fusion module to integrate visual and textual information when both modalities are available. Extensive experiments on standard, cross-dataset, and fine-grained remote sensing benchmarks show that visual prompting yields more reliable category specification under semantic ambiguity and distribution shifts, while multimodal prompting provides a flexible alternative that remains competitive when textual semantics are well aligned.

Hui Miao, Yuanfang Guo, Yunhong Wang

With the rapid development of the image generation technologies, the malicious abuses of the GAN-generated fingerprint images poses a significant threat to the public safety in certain circumstances. Although the existing universal deep forgery detection approach can be applied to detect the fake fingerprint images, they are easily attacked and have poor robustness. Meanwhile, there is no specifically designed deep forgery detection method for fingerprint images. In this paper, we propose the first deep forgery detection approach for fingerprint images, which combines unique ridge features of fingerprint and generation artifacts of the GAN-generated images, to the best of our knowledge. Specifically, we firstly construct a ridge stream, which exploits the grayscale variations along the ridges to extract unique fingerprint-specific features. Then, we construct a generation artifact stream, in which the FFT-based spectrums of the input fingerprint images are exploited, to extract more robust generation artifact features. At last, the unique ridge features and generation artifact features are fused for binary classification (i.e., real or fake). Comprehensive experiments demonstrate that our proposed approach is effective and robust with low complexities.

Ruijie Yang, Yuanfang Guo, Junfu Wang et al.

This paper focuses on an important type of black-box attacks, i.e., transfer-based adversarial attacks, where the adversary generates adversarial examples by a substitute (source) model and utilize them to attack an unseen target model, without knowing its information. Existing methods tend to give unsatisfactory adversarial transferability when the source and target models are from different types of DNN architectures (e.g. ResNet-18 and Swin Transformer). In this paper, we observe that the above phenomenon is induced by the output inconsistency problem. To alleviate this problem while effectively utilizing the existing DNN models, we propose a common knowledge learning (CKL) framework to learn better network weights to generate adversarial examples with better transferability, under fixed network architectures. Specifically, to reduce the model-specific features and obtain better output distributions, we construct a multi-teacher framework, where the knowledge is distilled from different teacher architectures into one student network. By considering that the gradient of input is usually utilized to generated adversarial examples, we impose constraints on the gradients between the student and teacher models, to further alleviate the output inconsistency problem and enhance the adversarial transferability. Extensive experiments demonstrate that our proposed work can significantly improve the adversarial transferability.

Yutong Zhang, Jiaxin Chen, Honglin Chen et al.

Memory-efficient transfer learning (METL) approaches have recently achieved promising performance in adapting pre-trained models to downstream tasks. They avoid applying gradient backpropagation in large backbones, thus significantly reducing the number of trainable parameters and high memory consumption during fine-tuning. However, since they typically employ a lightweight and learnable side network, these methods inevitably introduce additional memory and time overhead during inference, which contradicts the ultimate goal of efficient transfer learning. To address the above issue, we propose a novel approach dubbed Masked Dual Path Distillation (MDPD) to accelerate inference while retaining parameter and memory efficiency in fine-tuning with fading side networks. Specifically, MDPD develops a framework that enhances the performance by mutually distilling the frozen backbones and learnable side networks in fine-tuning, and discard the side network during inference without sacrificing accuracy. Moreover, we design a novel feature-based knowledge distillation method for the encoder structure with multiple layers. Extensive experiments on distinct backbones across vision/language-only and vision-and-language tasks demonstrate that our method not only accelerates inference by at least 25.2\% while keeping parameter and memory consumption comparable, but also remarkably promotes the accuracy compared to SOTA approaches. The source code is available at https://github.com/Zhang-VKk/MDPD.

Chenguang Liu, Guangshuai Gao, Ziyue Huang et al.

Detecting objects from aerial images poses significant challenges due to the following factors: 1) Aerial images typically have very large sizes, generally with millions or even hundreds of millions of pixels, while computational resources are limited. 2) Small object size leads to insufficient information for effective detection. 3) Non-uniform object distribution leads to computational resource wastage. To address these issues, we propose YOLC (You Only Look Clusters), an efficient and effective framework that builds on an anchor-free object detector, CenterNet. To overcome the challenges posed by large-scale images and non-uniform object distribution, we introduce a Local Scale Module (LSM) that adaptively searches cluster regions for zooming in for accurate detection. Additionally, we modify the regression loss using Gaussian Wasserstein distance (GWD) to obtain high-quality bounding boxes. Deformable convolution and refinement methods are employed in the detection head to enhance the detection of small objects. We perform extensive experiments on two aerial image datasets, including Visdrone2019 and UAVDT, to demonstrate the effectiveness and superiority of our proposed approach. Code is available at https://github.com/dawn-ech/YOLC.

Zihao Cheng, Hongru Wang, Zeming Liu et al.

Terminal agents extend Large Language Models with the ability to execute tasks directly in command-line environments, but their progress is bottlenecked by the scarcity of high-quality training data. Existing approaches bootstrap from partial sources such as human-defined seeds or GitHub repositories to instantiate one component and then complete the rest, producing tasks confined to narrow seed distributions, environments misaligned with task semantics, and inefficient trajectories from unguided exploration. To address these limitations, we introduce Terminal-World, a fully automated pipeline that uses agent skills as the central synthesis primitive, which jointly encode what to accomplish, when to apply (preconditions and environment state), and how to execute, enabling task instructions, environments, and teacher trajectories to be co-derived. To further broaden the synthesis space, Terminal-World composes skills into skill teams and skill graphs for multi-role and cross-domain task synthesis. Using this pipeline, we construct 5,723 training environments and train Terminal-World-8B/14B/32B, evaluated across 6 benchmarks where the Terminal-World series consistently outperforms terminal-agent baselines. Notably, using the same teacher model and only 1.2% of the training data, Terminal-World-32B surpasses Nemotron-Terminal-32B on Terminal-Bench 2.0 by +4.5 Pass@1 (31.5) and achieves 43.8 Pass@3.

Jingjing Liu, Ziye Huang, Zihao Cheng et al.

While Graphical User Interface (GUI) agents have shown promising performance in automated device interaction, they primarily depend on static parametric knowledge from pre-training or instruction tuning. This reliance fundamentally limits their ability to handle long-tailed tasks that require explicit procedural knowledge absent from model parameters, often forcing agents to resort to inefficient and brittle trial-and-error exploration. To mitigate this limitation, we introduce \textbf{Proactive Document-Guided Action} for GUI agents in dynamic, open-web environments, a novel paradigm that mirrors human problem-solving by enabling agents to autonomously search for relevant documentation to resolve long-tailed tasks. To evaluate agents' capability in this paradigm, we propose \textbf{DocOS}, a benchmark designed to assess document-guided problem solving in fully interactive environments. DocOS requires agents to autonomously navigate a web browser, locate relevant online documentation, comprehend procedural instructions, and faithfully ground them into executable GUI actions. Extensive experiments reveal that progress is strictly constrained by dual bottlenecks: agents struggle to reliably locate relevant information during proactive search and frequently fail to faithfully ground retrieved instructions into precise actions, pointing toward document-guided interaction as a crucial pathway for enabling self-evolving GUI agents in dynamic environments.

Ziyue Huang, Hongxi Yan, Qiqi Zhan et al.

The rapid advancement of remote sensing foundation models, particularly vision and multimodal models, has significantly enhanced the capabilities of intelligent geospatial data interpretation. These models combine various data modalities, such as optical, radar, and LiDAR imagery, with textual and geographic information, enabling more comprehensive analysis and understanding of remote sensing data. The integration of multiple modalities allows for improved performance in tasks like object detection, land cover classification, and change detection, which are often challenged by the complex and heterogeneous nature of remote sensing data. However, despite these advancements, several challenges remain. The diversity in data types, the need for large-scale annotated datasets, and the complexity of multimodal fusion techniques pose significant obstacles to the effective deployment of these models. Moreover, the computational demands of training and fine-tuning multimodal models require significant resources, further complicating their practical application in remote sensing image interpretation tasks. This paper provides a comprehensive review of the state-of-the-art in vision and multimodal foundation models for remote sensing, focusing on their architecture, training methods, datasets and application scenarios. We discuss the key challenges these models face, such as data alignment, cross-modal transfer learning, and scalability, while also identifying emerging research directions aimed at overcoming these limitations. Our goal is to provide a clear understanding of the current landscape of remote sensing foundation models and inspire future research that can push the boundaries of what these models can achieve in real-world applications. The list of resources collected by the paper can be found in the https://github.com/IRIP-BUAA/A-Review-for-remote-sensing-vision-language-models.

Wenqiang Sun, Haiyu Zhang, Haoyuan Wang et al.

This paper presents WorldPlay, a streaming video diffusion model that enables real-time, interactive world modeling with long-term geometric consistency, resolving the trade-off between speed and memory that limits current methods. WorldPlay draws power from three key innovations. 1) We use a Dual Action Representation to enable robust action control in response to the user's keyboard and mouse inputs. 2) To enforce long-term consistency, our Reconstituted Context Memory dynamically rebuilds context from past frames and uses temporal reframing to keep geometrically important but long-past frames accessible, effectively alleviating memory attenuation. 3) We also propose Context Forcing, a novel distillation method designed for memory-aware model. Aligning memory context between the teacher and student preserves the student's capacity to use long-range information, enabling real-time speeds while preventing error drift. Taken together, WorldPlay generates long-horizon streaming 720p video at 24 FPS with superior consistency, comparing favorably with existing techniques and showing strong generalization across diverse scenes. Project page and online demo can be found: https://3d-models.hunyuan.tencent.com/world/ and https://3d.hunyuan.tencent.com/sceneTo3D.

Yongchao Feng, Yajie Liu, Shuai Yang et al.

Vision-Language Model (VLM) have gained widespread adoption in Open-Vocabulary (OV) object detection and segmentation tasks. Despite they have shown promise on OV-related tasks, their effectiveness in conventional vision tasks has thus far been unevaluated. In this work, we present the systematic review of VLM-based detection and segmentation, view VLM as the foundational model and conduct comprehensive evaluations across multiple downstream tasks for the first time: 1) The evaluation spans eight detection scenarios (closed-set detection, domain adaptation, crowded objects, etc.) and eight segmentation scenarios (few-shot, open-world, small object, etc.), revealing distinct performance advantages and limitations of various VLM architectures across tasks. 2) As for detection tasks, we evaluate VLMs under three finetuning granularities: \textit{zero prediction}, \textit{visual fine-tuning}, and \textit{text prompt}, and further analyze how different finetuning strategies impact performance under varied task. 3) Based on empirical findings, we provide in-depth analysis of the correlations between task characteristics, model architectures, and training methodologies, offering insights for future VLM design. 4) We believe that this work shall be valuable to the pattern recognition experts working in the fields of computer vision, multimodal learning, and vision foundation models by introducing them to the problem, and familiarizing them with the current status of the progress while providing promising directions for future research. A project associated with this review and evaluation has been created at https://github.com/better-chao/perceptual_abilities_evaluation.

Zhuguanyu Wu, Jiayi Zhang, Jiaxin Chen et al.

Vision Transformers (ViTs) have become one of the most commonly used backbones for vision tasks. Despite their remarkable performance, they often suffer significant accuracy drops when quantized for practical deployment, particularly by post-training quantization (PTQ) under ultra-low bits. Recently, reconstruction-based PTQ methods have shown promising performance in quantizing Convolutional Neural Networks (CNNs). However, they fail when applied to ViTs, primarily due to the inaccurate estimation of output importance and the substantial accuracy degradation in quantizing post-GELU activations. To address these issues, we propose \textbf{APHQ-ViT}, a novel PTQ approach based on importance estimation with Average Perturbation Hessian (APH). Specifically, we first thoroughly analyze the current approximation approaches with Hessian loss, and propose an improved average perturbation Hessian loss. To deal with the quantization of the post-GELU activations, we design an MLP Reconstruction (MR) method by replacing the GELU function in MLP with ReLU and reconstructing it by the APH loss on a small unlabeled calibration set. Extensive experiments demonstrate that APHQ-ViT using linear quantizers outperforms existing PTQ methods by substantial margins in 3-bit and 4-bit across different vision tasks. The source code is available at https://github.com/GoatWu/APHQ-ViT.

Zimeng Wu, Yunhong Wang, Donghao Wang et al.

Vision-Language Models (VLMs) have advanced rapidly within the unified Transformer architecture, yet their deployment on resource-constrained devices remains challenging due to high computational complexity. While pruning has emerged as an effective technique for compressing VLMs, existing approaches predominantly focus on a single mode by pruning either parameters or tokens, neglecting fully exploring the inherent redundancy in each mode, which leads to substantial performance degradation at high pruning ratios. To address the above limitations, we propose Collaborative Multi-Mode Pruning (CoMP), a novel framework tailored for VLMs by performing joint parameter and token pruning. Specifically, we first design a Collaborative Importance Metric (CIM) that investigates the mutual interference between the coupled parameters and tokens. It incorporates distinct significance of tokens into the computation of parameter importance scores, while simultaneously mitigating the affect of pruned parameters on token importance scores. Moreover, we develop a Multi-Mode Pruning Strategy (MPS) that decomposes the overall pruning process into a sequence of pruning stages, while in each stage we estimate the priory of different pruning modes based on their pruning cost and adaptively shift to the optimal one. Additionally, MPS integrates the historical cost and random exploration, in order to achieve a stable pruning process and avoid local optimum. Extensive experiments across various vision-language tasks and models demonstrate that our method effectively promotes the performance under high pruning ratios by comparing to the state-of-the-art approaches. The source code is available at https://github.com/Wuzimeng/CoMP.git.

Zihao Cheng, Hongru Wang, Zeming Liu et al.

While integrating external tools into large language models (LLMs) enhances their ability to access real-time information and domain-specific services, existing approaches focus narrowly on functional tool selection following user instructions, overlooking the context-aware personalization in tool selection. This oversight leads to suboptimal user satisfaction and inefficient tool utilization, particularly when overlapping toolsets require nuanced selection based on contextual factors. To bridge this gap, we introduce ToolSpectrum, a benchmark designed to evaluate LLMs' capabilities in personalized tool utilization. Specifically, we formalize two key dimensions of personalization, user profile and environmental factors, and analyze their individual and synergistic impacts on tool utilization. Through extensive experiments on ToolSpectrum, we demonstrate that personalized tool utilization significantly improves user experience across diverse scenarios. However, even state-of-the-art LLMs exhibit the limited ability to reason jointly about user profiles and environmental factors, often prioritizing one dimension at the expense of the other. Our findings underscore the necessity of context-aware personalization in tool-augmented LLMs and reveal critical limitations for current models. Our data and code are available at https://github.com/Chengziha0/ToolSpectrum.