论文检索

Guanghao Zhang, Tao Zhong, Yan Xia, Mushui Liu, Zhelun Yu, Haoyuan Li, Wanggui He, Dong She, Yi Wang, Hao Jiang

While previous multimodal slow-thinking methods have demonstrated remarkable success in single-image understanding scenarios, their effectiveness becomes fundamentally constrained when extended to more complex multi-image comprehension tasks. This limitation stems from their predominant reliance on text-based intermediate reasoning processes. While for human, when engaging in sophisticated multi-image analysis, they typically perform two complementary cognitive operations: (1) continuous cross-image visual comparison through region-of-interest matching, and (2) dynamic memorization of critical visual concepts throughout the reasoning chain. Motivated by these observations, we propose the Complex Multi-Modal Chain-of-Thought (CMMCoT) framework, a multi-step reasoning framework that mimics human-like "slow thinking" for multi-image understanding. Our approach incorporates two key innovations: (1) The construction of interleaved multimodal multi-step reasoning chains, which utilize critical visual region tokens, extracted from intermediate reasoning steps, as supervisory signals. This mechanism not only facilitates comprehensive cross-modal understanding but also enhances model interpretability. (2) The introduction of a test-time memory augmentation module that expands the model’s reasoning capacity during inference while preserving parameter efficiency. Furthermore, to facilitate research in this direction, we have curated a novel multi-image slow-thinking dataset. Extensive experiments demonstrate the effectiveness of our model.

Fan Zhang, Zhiwei Gu, Hua Wang

To address the limitations of Transformer decoders in capturing edge details, recognizing local textures and modeling spatial continuity, this paper proposes a novel decoder framework specifically designed for medical image segmentation, comprising three core modules. First, the Adaptive Cross-Fusion Attention (ACFA) module integrates channel feature enhancement with spatial attention mechanisms and introduces learnable guidance in three directions (planar, horizontal, and vertical) to enhance responsiveness to key regions and structural orientations. Second, the Triple Feature Fusion Attention (TFFA) module fuses features from Spatial, Fourier and Wavelet domains, achieving joint frequency-spatial representation that strengthens global dependency and structural modeling while preserving local information such as edges and textures, making it particularly effective in complex and blurred boundary scenarios. Finally, the Structural-aware Multi-scale Masking Module (SMMM) optimizes the skip connections between encoder and decoder by leveraging multi-scale context and structural saliency filtering, effectively reducing feature redundancy and improving semantic interaction quality. Working synergistically, these modules not only address the shortcomings of traditional decoders but also significantly enhance performance in high-precision tasks such as tumor segmentation and organ boundary extraction, improving both segmentation accuracy and model generalization. Experimental results demonstrate that this framework provides an efficient and practical solution for medical image segmentation.

Evelyn Zhang, Fufu Yu, Aoqi Wu, Zichen Wen, Ke Yan, Shouhong Ding, Biqing Qi, Linfeng Zhang

Processing long visual token sequences poses a significant computational burden on Multimodal Large Language Models (MLLMs). While token pruning offers a path to acceleration, we find that current methods, while adequate for general understanding, catastrophically fail on fine-grained localization tasks. We attribute this failure to the inherent flaws of the two prevailing strategies: importance-based methods suffer from a strong positional bias, an inherent model artifact that distracts from semantic content, while diversity-based methods exhibit structural blindness, disregarding the user's prompt and spatial redundancy. To address this, we introduce D²Pruner, a framework that rectifies these issues by uniquely combining debiased importance with a structural pruning mechanism. Our method first secures a core set of the most critical tokens as pivots based on a debiased attention score. It then performs a Maximal Independent Set (MIS) selection on the remaining tokens, which are modeled on a hybrid graph where edges signify spatial proximity and semantic similarity. This process iteratively preserves the most important and available token while removing its neighbors, ensuring that the supplementary tokens are chosen to maximize importance and diversity. Extensive experiments demonstrate that D²Pruner achieves exceptional efficiency and fidelity.

Dafeng Zhang, Yongqi Song, Shizhuo Liu

The method used to measure relationships between face embeddings plays a crucial role in determining the performance of face clustering. Existing methods employ the Jaccard similarity coefficient instead of the traditional cosine distance to enhance the measurement accuracy. However, these methods introduce an excessive number of irrelevant nodes, producing Jaccard coefficients with limited discriminative power and adversely affecting clustering performance. To address this issue, we propose a prediction-driven Top-K Jaccard similarity coefficient that enhances the purity of neighboring nodes, thereby improving the reliability of similarity measurements. Nevertheless, accurately predicting the optimal number of neighbors (Top-K) remains challenging, leading to suboptimal clustering results. To overcome this limitation, we develop a Transformer-based prediction model that examines the relationships between the central node and its neighboring nodes near the Top-K to further enhance the reliability of similarity estimation. However, vanilla Transformer, when applied to predict relationships between nodes, often introduces noise due to their overemphasis on irrelevant feature relationships. To address these challenges, we propose a Sparse Differential Transformer (SDT), instead of the vanilla Transformer, to eliminate noise and enhance the model's anti-noise capabilities. Extensive experiments on multiple datasets, such as MS-Celeb-1M, demonstrate that our approach achieves state-of-the-art (SOTA) performance, outperforming existing methods and providing a more robust solution for face clustering.

Chi Zhang, Xiang Zhang, Lei Yu, Gui-Song Xia, Yuming Fang, Wenhan Yang

Super-Resolution from a Blurry low-resolution image (SRB) constitutes a severely ill-posed inverse problem. Current learning-based SRB approaches primarily rely on synthetic, well-labeled paired datasets to regularize solution spaces, yet they exhibit limited generalizability in practical applications due to significant domain discrepancies between simulated degradations and real-world imaging conditions. To bridge this synthetic-to-real gap, we propose a novel Self-supervised Event-based SRB (SE-SRB) framework that leverages neuromorphic event streams as physical priors and adopts a lightweight neural architecture tailored for effective domain adaptation. Specifically, the proposed SE-SRB introduces a self-supervised learning paradigm based on asymmetric integral driven consistency, which enforces temporal coherence between predictions derived from RGB and asynchronous event streams at different time points. Extensive experiments validate that SE-SRB consistently outperforms state-of-the-art methods on both synthetic and real-world datasets. Built upon a lightweight parallel two-stream architecture, SE-SRB achieves high computational efficiency, featuring reduced parameter count, lower FLOPs, and real-time inference capability (40 FPS).

Cheng Zhang, Hanwen Liang, Donny Y. Chen, Qianyi Wu, Konstantinos N. Plataniotis, Camilo Cruz Gambardella, Jianfei Cai

Panoramic video generation has attracted growing attention due to its applications in virtual reality and immersive media. However, existing methods lack explicit motion control and struggle to generate scenes with large and complex motions. We propose PanFlow a novel approach that exploits the spherical nature of panoramas to decouple the highly dynamic camera rotation from the input optical flow condition, enabling more precise control over large and dynamic motions. We further introduce a spherical noise warping strategy to promote loop consistency in motion across panorama boundaries. To support effective training, we curate a large-scale, motion-rich panoramic video dataset with frame-level pose and flow annotations. We also showcase the effectiveness of our method in various applications, including motion transfer and video editing. Extensive experiments demonstrate that PanFlow significantly outperforms prior methods in motion fidelity, visual quality, and temporal coherence.

Bozhou Zhang, Jingyu Li, Nan Song, Li Zhang

End-to-end autonomous driving has achieved remarkable advancements in recent years. Existing methods primarily follow a perception–planning paradigm, where perception and planning are executed sequentially within a fully differentiable framework for planning-oriented optimization. We further advance this paradigm through a "perception-in-plan'' framework design, which integrates perception into the planning process. This design facilitates targeted perception guided by evolving planning objectives over time, ultimately enhancing planning performance. Building on this insight, we introduce VeteranAD, a coupled perception and planning framework for end-to-end autonomous driving. By incorporating multi-mode anchored trajectories as planning priors, the perception module is specifically designed to gather traffic elements along these trajectories, enabling comprehensive and targeted perception. Planning trajectories are then generated based on both the perception results and the planning priors. To make perception fully serve planning, we adopt an autoregressive strategy that progressively predicts future trajectories while focusing on relevant regions for targeted perception at each step. With this simple yet effective design, VeteranAD fully unleashes the potential of planning-oriented end-to-end methods, leading to more accurate and reliable driving behavior. Extensive experiments on the NAVSIM and Bench2Drive datasets demonstrate that our VeteranAD achieves state-of-the-art performance.

Bofei Zhang, Zirui Shang, Zhi Gao, Wang Zhang, Rui Xie, Xiaojian Ma, Tao Yuan, Xinxiao Wu, Song-Chun Zhu, Qing Li

Building Graphical User Interface (GUI) agents is a promising research direction, which simulates human interaction with computers or mobile phones to perform diverse GUI tasks. However, a major challenge in developing generalized GUI agents is the lack of sufficient trajectory data across various operating systems and applications, mainly due to the high cost of manual annotations. In this paper, we propose the TongUI framework that transforms millions of multimodal web tutorials into GUI trajectories for generalized GUI agents. Concretely, we crawl GUI videos and articles from the Internet and process them into GUI agent trajectory data. Based on this, we construct the GUI-Net-1M dataset, which contains 1 million trajectories across five operating systems and over 280 applications. To the best of our knowledge, this is the largest open-source GUI trajectory dataset. We develop the TongUI agent by fine-tuning Qwen2.5-VL-3B/7B/32B models on GUI-Net-1M, which shows consistent performance improvements on commonly used grounding and navigation benchmarks, outperforming baseline agents by 10\% on multiple benchmarks, showing the effectiveness of the GUI-Net-1M dataset and underscoring the significance of our TongUI framework.

Yu-Liang Zhan, Xinyu Tang, Han Wan, Jian Li, Jirong Wen, Hao Sun

Recently, Chain-of-Thought (CoT) reasoning has significantly enhanced the capabilities of large language models (LLMs), but Vision–Language Models (VLMs) still struggle with multi-step reasoning tasks due to limited multimodal reasoning data. To bridge this gap, researchers have explored methods to transfer CoT reasoning from LLMs to VLMs. However, existing approaches either need high training costs or require architectural alignment. In this paper, we use Linear Artificial Tomography (LAT) to empirically show that LLMs and VLMs share similar low-frequency latent representations of CoT reasoning despite architectural differences. Based on this insight, we propose L2V-CoT, a novel training-free latent intervention approach that transfers CoT reasoning from LLMs to VLMs. L2V-CoT extracts and resamples low-frequency CoT representations from LLMs in the frequency domain, enabling dimension matching and latent injection into VLMs during inference to enhance reasoning capabilities. Extensive experiments demonstrate that our approach consistently outperforms training-free baselines and even surpasses supervised methods.

Lixin Zhan, Jiang Jie, Tianjian Zhou, Yukun Du, Yan Zheng, Xuehu Duan

Current semantic segmentation approaches for point cloud scenes heavily rely on manual labeling, while research on unsupervised semantic segmentation methods specifically for raw point clouds is still in its early stages. Unsupervised point cloud learning poses significant challenges due to the absence of annotation information and the lack of pre-training. The development of effective strategies is crucial in this context. In this paper, we propose a novel prototype library-driven unsupervised point cloud semantic segmentation strategy that utilizes Structure Learning and Consistent Reasoning (P-SLCR). First, we propose a Consistent Structure Learning to establish structural feature learning between consistent points and the library of consistent prototypes by selecting high-quality features. Second, we propose a Semantic Relation Consistent Reasoning that constructs a prototype inter-relation matrix between consistent and ambiguous prototype libraries separately. This process ensures the preservation of semantic consistency by imposing constraints on consistent and ambiguous prototype libraries through the prototype inter-relation matrix. Finally, our method was extensively evaluated on the S3DIS, SemanticKITTI, and Scannet datasets, achieving the best performance compared to unsupervised methods. Specifically, the mIoU of 47.1% is achieved for Area-5 of the S3DIS dataset, surpassing the classical fully supervised method PointNet by 2.5%.

Yaohua Zha, Xue Yuerong, Chunlin Fan, Yuansong Wang, Tao Dai, Ke Chen, Shu-Tao Xia

Deep learning-based 3D anomaly detection methods have demonstrated significant potential in industrial manufacturing. However, many approaches are specifically designed for anomaly detection tasks, which limits their generalizability to other 3D tasks. In contrast, self-supervised point cloud models aim for general representation learning, yet our investigation reveals that these classical models are suboptimal at anomaly detection under the unified fine-tuning paradigm. This motivates us to develop a more generalizable 3D model that can effectively detect anomalies without relying on task-specific designs. Interestingly, we find that using only the curvature of each point as its anomaly score already outperforms several classical self-supervised and dedicated anomaly detection models, highlighting the critical role of curvature in 3D anomaly detection. In this paper, we propose a Curvature-Augmented Self-supervised Learning (CASL) framework based on a reconstruction paradigm. Built upon the classical U-Net architecture, our approach introduces multi-scale curvature prompts to guide the decoder in predicting the coordinates of each point. Without relying on any dedicated anomaly detection mechanisms, it achieves leading detection performance through straightforward anomaly classification fine-tuning. Moreover, the learned representations generalize well to standard 3D understanding tasks such as point cloud classification.

Mingfeng Zha, Guoqing Wang, Tianyu Li, Wei Dong, Peng Wang, Yang Yang

Reflective imaging enables the mirror imagings and physical entities to possess identical attributes, e.g., color and shape. Current mirror detection (MD) methods primarily rely on designing functional components to establish the correlation and disparities between the imagings and entities, thereby identifying the mirror regions. However, the exploration of extended scenes with dynamic content changes is rarely investigated. Therefore, we propose the MirrorSAM designed for MD based on the Segment Anything Model (SAM). Specifically, due to the varying reflections produced by mirrors in different positions and the complex visual space that interferes with localization, we design the hierarchical mixture of direction experts (HMDE) in the low-rank space to reduce biases towards entities in SAM and dynamically adjust experts based on the input scene. We observe differences in depth between mirrors and adjacent areas, and propose the depth token calibration (DTC), which introduces a learnable depth token to generate the depth map and serve as an error correction factor. We further formulate the selective pixel-prototype contrastive (SPPC) loss, selecting partially confusable samples to promote the decoupling of mirror and non-mirror representations. Extensive experiments conducted on four mirror benchmarks and two settings demonstrate that our approach surpasses state-of-the-art methods with few trainable parameters and FLOPs. We further extend to four transparent surface benchmarks to validate generalization.

Xinyi Zeng, Yuxiang Yang, Pinxian Zeng, Wenxia Yin, Bo Liu, Xi Wu, Yan Wang

Facial Expression Recognition (FER) is crucial to human-computer interaction. Existing cross-domain FER (CD-FER) methods mainly focus on single-source closed-set scenarios, transferring knowledge from a single source domain to a target domain with identical class sets. However, CD-FER faces two real-world challenges: 1) the need to leverage information from multiple sources, leading to multi-domain shift, and 2) the necessity to recognize unseen target classes, resulting in class shift. These issues give rise to a novel and challenging task, which we define as Multi-domain Open-set FER (MO-FER). In this paper, we propose PromptEmo, a novel CLIP-based framework that leverages bilateral textual prompts to address both shifts in the MO-FER task. Leveraging the generalizability of LLM, PromptEmo constructs trainable positive prompts with LLM-generated emotion descriptions for seen classes, as well as template-derived negative prompts to enhance the reasoning for unseen classes. Then, we introduce a modal-task optimization paradigm organized from two perspectives: textual semantics and visual domains, yielding Intra-modal Space-specific Optimization (ISO) and Cross-modal Emotion-aware Interaction (CEI) strategies. ISO refines the CLIP-based textual space to ensure semantic separation between bilateral prompts and improves the latent visual space by promoting inter-domain alignment. Founded on ISO, CEI facilitates effective vision-language interactions, resulting in four joint loss terms that improve emotion recognition by shaping a domain-invariant, discriminative feature space. PromptEmo surpasses the current SOTA method by 7.7% AUC on unseen classes across four FER datasets, serving as a strong baseline for the MO-FER task.

Shuang Zeng, Xinyuan Chang, Xinran Liu, Yujian Yuan, Shiyi Liang, Zheng Pan, Mu Xu, Xing Wei

High-Definition Maps (HD maps) are essential for the precise navigation and decision-making of autonomous vehicles, yet their creation and upkeep present significant cost and timeliness challenges. The online construction of HD maps using on-board sensors has emerged as a promising solution; however, these methods can be impeded by incomplete data due to occlusions and inclement weather, while their performance in distant regions remains unsatisfying. This paper proposes PriorDrive to address these limitations by directly harnessing the power of various vectorized prior maps, significantly enhancing the robustness and accuracy of online HD map construction. Our approach integrates a variety of prior maps uniformly, such as OpenStreetMap's Standard Definition Maps (SD maps), outdated HD maps from vendors, and locally constructed maps from historical vehicle data. To effectively integrate such prior information into online mapping models, we introduce a Hybrid Prior Representation (HPQuery) that standardizes the representation of diverse map elements. We further propose a Unified Vector Encoder (UVE), which employs fused prior embedding and a dual encoding mechanism to encode vector data. To improve the UVE's generalizability and performance, we propose a segment-level and point-level pre-training strategy that enables the UVE to learn the prior distribution of vector data. Through extensive testing on the nuScenes, Argoverse 2 and OpenLane-V2, we demonstrate that PriorDrive is highly compatible with various online mapping models and substantially improves map prediction capabilities. The integration of prior maps through PriorDrive offers a robust solution to the challenges of single-perception data, paving the way for more reliable autonomous driving.

Huimin Zeng, Yue Bai, Yun Fu

Existing 3D Gaussian Splatting (3DGS) super-resolution methods typically perform high-resolution (HR) rendering of fixed scale factors, making them impractical for resource-limited scenarios. Directly rendering arbitrary-scale HR views with vanilla 3DGS introduces aliasing artifacts due to the lack of scale-aware rendering ability, while adding a post-processing upsampler for 3DGS complicates the framework and reduces rendering efficiency. To tackle these issues, we build an integrated framework that incorporates scale-aware rendering, generative prior-guided optimization, and progressive super-resolving to enable 3D Gaussian super-resolution of arbitrary scale factors with a single 3D model. Notably, our approach supports both integer and non-integer scale rendering to provide more flexibility. Extensive experiments demonstrate the effectiveness of our model in producing high-quality arbitrary-scale HR views (6.59 dB PSNR gain over 3DGS) with a single model. It preserves structural consistency with LR views and across different scales, while maintaining real-time rendering speed (85 FPS at 1080p).

Zhaoquan Yuan, Zining Wang, Yuankang Pan, Ao Luo, Wei Li, Xiao Wu, Changsheng Xu

Compositional Zero-Shot Learning (CZSL) addresses the challenge of recognizing unseen attribute-object compositions in images, representing a fundamental challenge in artificial intelligence. Current approaches, which primarily focus on semantic alignment or distribution independence of primitives, have not achieved effective state-object decoupling and causal interventional invariance, limiting their performance on unseen compositions. To tackle this challenge, this study introduces I2CD (Invertible Causal framework via Disentangle-Compose-Disentangle), a novel framework that integrates invertible neural networks with causal intervention techniques to achieve state-object disentanglement. The framework employs a disentangle-compose-disentangle mechanism for counterfactual generation within the disentangled representation space, ensuring that modifications to one primitive (attribute or object) maintain independence from the other, thus enabling robust causal disentanglement. Representational consistency is maintained through semantic alignment between initial disentangled representations and their recomposed-then-disentangled counterparts with corresponding textual concepts. Comprehensive evaluations on three benchmark datasets—MIT-States, UT-Zappos, and C-GQA—demonstrate the framework's effectiveness in achieving both disentanglement and compositional generalization in CZSL tasks.

Yuxuan Yuan, Lichen Wei, Luyao Tang, Chaoqi Chen, Zheyuan Cai, Yue Huang, Xinghao Ding

Open-set object detection (OSOD) aims to recognize known object categories while localizing previously unseen instances. However, real-world scenarios often involve co-occurring domain shifts and novel object categories. Existing OSOD methods typically overlook domain shifts, relying on source-trained representations that entangle domain-specific style with semantic content, thereby hindering generalization to both unseen domains and novel categories. To address this challenge, we propose a unified framework, termed DecOmpose and ATtribute (DOAT), which disentangles domain-specific style from semantic structure, thereby facilitating generalizable object detection. DOAT employs wavelet-based feature decomposition to separate style information from high-frequency structural details, thus enabling an explicit separation of domain and category shifts. To account for domain shift, the low-frequency components are perturbed within a style subspace to simulate diverse domain appearances. For unknown object discovery, the high-frequency components are utilized to estimate objectness scores via an attribution mechanism that fuses wavelet energy with semantic distance to known-category prototypes. Extensive experiments on standard open-set benchmarks have demonstrated the superior generalization performance of DOAT.

Yujian Yuan, Changjie Wu, Xinyuan Chang, Sijin Wang, Hang Zhang, Shiyi Liang, Shuang Zeng, Mu Xu

Large-scale map construction is foundational for critical applications such as autonomous driving and navigation systems. Traditional large-scale map construction approaches mainly rely on costly and inefficient special data collection vehicles and labor-intensive annotation processes. While existing satellite-based methods have demonstrated promising potential in enhancing the efficiency and coverage of map construction, they exhibit two major limitations: (1) inherent drawbacks of satellite data (e.g., occlusions, outdatedness) and (2) inefficient vectorization from perception-based methods, resulting in discontinuous and rough roads that require extensive post-processing. This paper presents a novel generative framework, UniMapGen, for large-scale map construction, offering three key innovations: (1) representing lane lines as discrete sequence and establishing an iterative strategy to generate more complete and smooth map vectors than traditional perception-based methods. (2) proposing a flexible architecture that supports multi-modal inputs, enabling dynamic selection among BEV, PV, and text prompt, to overcome the drawbacks of satellite data. (3) developing a state update strategy for global continuity and consistency of the constructed large-scale map. UniMapGen achieves state-of-the-art performance on the OpenSatMap dataset. Furthermore, UniMapGen can infer occluded roads and predict roads missing from dataset annotations.

Xinpan Yuan, Shaomin Xie, Liujie Hua, Chengyuan Zhang, Guihu Zhao, Lin Yuanbo Wu

Text-Based Person Retrieval (TBPR) aims to accurately retrieve target individuals from large-scale image databases using only textual descriptions. Existing methods typically assume a ground-truth correspondence between text and images (i.e., strongly correlated). However, in real-world scenarios, this assumption may not be able to hold for the cross-modal matching due to weak or even corrupted correlations between textual descriptions and visual content, referred to as noisy correspondence (NC). Such NC largely disrupts the correspondence learning between visual and semantic modalities. Though prior works have improved single-modal robustness against noisy labels, systematic modeling of both cross-modal and intra-modal geometric structures in TBPR remains limited attention. In this paper, we propose Geometric Structure Consistency Alignment (GSCA) to TBPR, which leverages cross-modal cosine similarity and intra-modal nearest-neighbor affinity to learn visual-semantic consistency under noisy correspondence. To mitigate the structural corruption caused by noisy pairs, we introduce the Structure Refinement and Mining (SRAM) module. By partitioning training data into clean, ambiguous, and noisy subsets, SRAM enables the model to strategically refine the cross-modal correspondence by mining reliable pairs, thus enhancing the reliability of positive or negative samples discrimination and preserving structural consistency across modalities. Extensive experiments demonstrate that our method achieves state-of-the-art performance across three public datasets. On CUHK-PEDES, it boosts Rank-1 by 1.42% in noise-free conditions, sustaining a robust 74.25% Rank-1 under a 50% noise ratio.

Xinbin Yuan, Zhaohui Zheng, Yuxuan Li, Xialei Liu, Li Liu, Xiang Li, Qibin Hou, Ming-Ming Cheng

In this paper, we show that current approaches using large square kernels or transformer-based global modeling aggregate contextual information uniformly across spatial dimensions, leading to feature dilution and localization errors for elongated targets. To mitigate this issue, we propose Strip R-CNN, the first work to systematically explore large strip convolutions for remote sensing object detection. Our key insight is that strip convolutions enable directional feature aggregation along the dominant spatial dimension of slender objects, reducing background interference while preserving essential geometric information. We design two core components: (i) StripNet, a backbone network employing sequential orthogonal large strip convolutions to capture anisotropic spatial patterns, and (ii) Strip Head, which enhances localization precision by incorporating strip convolutions into the detection head. Unlike previous large-kernel approaches that suffer from computational redundancy and isotropic limitations, our method achieves superior performance with remarkable efficiency. Extensive experiments on multiple benchmarks (DOTA, FAIR1M, HRSC2016, and DIOR) demonstrate significant improvements, with our 30M parameter model achieving 82.75% mAP on DOTA-v1.0, establishing a new state-of-the-art record while providing new insights into anisotropic feature learning for remote sensing applications.