Zhonglong Zheng

Zihao Xue, Yan Wang, Zhen Bi et al.

Diffusion Transformers have become a powerful backbone for text-to-image generation, but their layered and cross-modal generation process makes safety control fundamentally different from prompt-level filtering or output-level detection. Harmful semantics may be weakly expressed in text representations, progressively bound to visual latents, and finally entangled with rendering dynamics. As a result, safety steering at a fixed layer can be unstable, and a steering mechanism learned from known risks may not transfer reliably to a shifted target risk domain. We propose SafeDIG, a safety steering framework that formulates DiT safety adaptation as position-aware sparse feature transfer. SafeDIG first constructs Sparse Autoencoders over functionally distinct DiT intervention positions and uses robustness-aware pre-training routing to prioritize intervention sites that are expected to remain stable under source-target risk shift. It then separates transferable safety features from domain-specific activation geometry by freezing the SAE encoder as a reusable sparse safety dictionary and adapting only the decoder to the target-domain activation manifold. During inference, SafeDIG combines Blend and Repel operations to steer unsafe activations toward transferred safety manifolds or away from harmful sparse directions. Experiments on FLUX.1 Dev and Stable Diffusion 3.5 Large show that SafeDIG consistently reduces target-domain and overall unsafe generation rates while preserving source-domain safety and image quality.

Xiongbin Wu, Zhihao Luo, Shanzhe Lei et al.

Recently, vision-language model (VLM) agents have shown promising progress in open-world tasks, where successful task completion often requires multiple turns of visual perception and action execution. However, existing methods still rely primarily on Supervised Fine-Tuning (SFT) with expert demonstrations, while the advanced reinforcement learning (RL) algorithm, specifically Group Relative Policy Optimization (GRPO), has not been effectively employed for multi-turn RL in these tasks because standard GRPO requires full trajectories as training samples which leads to excessively long context and noise. To address this issue, we propose GROW, a RL framework for open-world VLM agents that decomposes collected trajectories into state-action samples, and computes advantages between these samples rather than treating a full trajectory as a single entity. We further provide a surrogate analysis indicating that, even though the grouped samples are conditioned on different local states rather than an identical prompt context, the objective can preserve the core relative policy optimization signal of GRPO under simplifying assumptions. Experiments on more than 800 Minecraft tasks show that our method achieves state-of-the-art (SOTA) performance, demonstrating the effectiveness of our proposed RL framework for open-world VLM agents.

Yanchao Wang, Dawei Zhang, Run Li et al.

Multi-object tracking (MOT) is a rising topic in video processing technologies and has important application value in consumer electronics. Currently, tracking-by-detection (TBD) is the dominant paradigm for MOT, which performs target detection and association frame by frame. However, the association performance of TBD methods degrades in complex scenes with heavy occlusions, which hinders the application of such methods in real-world scenarios.To this end, we incorporate pseudo-depth cues to enhance the association performance and propose Pseudo-Depth SORT (PD-SORT). First, we extend the Kalman filter state vector with pseudo-depth states. Second, we introduce a novel depth volume IoU (DVIoU) by combining the conventional 2D IoU with pseudo-depth. Furthermore, we develop a quantized pseudo-depth measurement (QPDM) strategy for more robust data association. Besides, we also integrate camera motion compensation (CMC) to handle dynamic camera situations. With the above designs, PD-SORT significantly alleviates the occlusion-induced ambiguous associations and achieves leading performances on DanceTrack, MOT17, and MOT20. Note that the improvement is especially obvious on DanceTrack, where objects show complex motions, similar appearances, and frequent occlusions. The code is available at https://github.com/Wangyc2000/PD_SORT.

Bingkun Nian, Fenghe Tang, Jianrui Ding et al.

Dynamic convolution demonstrates outstanding representation capabilities, which are crucial for natural image segmentation. However, it fails when applied to medical image segmentation (MIS) and infrared small target segmentation (IRSTS) due to limited data and limited fitting capacity. In this paper, we propose a new type of dynamic convolution called dynamic parameter convolution (DPConv) which shows superior fitting capacity, and it can efficiently leverage features from deep layers of encoder in reconstruction tasks to generate DPConv kernels that adapt to input variations.Moreover, we observe that DPConv, built upon deep features derived from reconstruction tasks, significantly enhances downstream segmentation performance. We refer to the segmentation network integrated with DPConv generated from reconstruction network as the siamese reconstruction-segmentation network (SRS). We conduct extensive experiments on seven datasets including five medical datasets and two infrared datasets, and the experimental results demonstrate that our method can show superior performance over several recently proposed methods. Furthermore, the zero-shot segmentation under unseen modality demonstrates the generalization of DPConv. The code is available at: https://github.com/fidshu/SRSNet.

Yanchao Wang, Dawei Zhang, Chengzhuan Yang et al.

Multi-object tracking (MOT) is a fundamental task in computer vision that requires continuously tracking multiple targets while maintaining consistent identities across frames. However, most existing approaches primarily rely on instance-level object features for trajectory association, which often leads to degraded performance under challenging conditions such as object deformation, nonlinear motion, and occlusion. In this work, we propose SAMOFT, a robust tracker that leverages pixel-level cues to improve robustness under complex motion scenarios. Specifically, we introduce a Pixel Motion Matching (PMM) module that integrates the Segment Anything Model (SAM) with dense optical flow to refine Kalman filter-based motion prediction using instantaneous foreground pixel motion. To further enhance robustness under unreliable detections, we design a Centroid Distance Matching (CDM) module that performs flexible mask-based centroid matching for low-confidence or partially occluded observations. Moreover, a Distribution-Based Correction (DBC) module models long-tailed motion patterns in a training-free manner using historical optical flow statistics and dynamically corrects trajectory states online. We also incorporate a Cluster-Aware ReID (CA-ReID) strategy to improve the stability and discriminative power of trajectory appearance features. Extensive experiments on the DanceTrack and MOTChallenge benchmarks demonstrate that SAMOFT consistently improves baseline trackers and achieves competitive performance compared with recent state-of-the-art methods, validating the effectiveness of leveraging pixel-level cues for robust multi-object tracking.

Xiaojian Wang, Chaoli Zhang, Zhonglong Zheng et al.

Time series forecasting has various applications, such as meteorological rainfall prediction, traffic flow analysis, financial forecasting, and operational load monitoring for various systems. Due to the sparsity of time series data, relying solely on time-domain or frequency-domain modeling limits the model's ability to fully leverage multi-domain information. Moreover, when applied to time series forecasting tasks, traditional attention mechanisms tend to over-focus on irrelevant historical information, which may introduce noise into the prediction process, leading to biased results. We proposed WDformer, a wavelet-based differential Transformer model. This study employs the wavelet transform to conduct a multi-resolution analysis of time series data. By leveraging the advantages of joint representation in the time-frequency domain, it accurately extracts the key information components that reflect the essential characteristics of the data. Furthermore, we apply attention mechanisms on inverted dimensions, allowing the attention mechanism to capture relationships between multiple variables. When performing attention calculations, we introduced the differential attention mechanism, which computes the attention score by taking the difference between two separate softmax attention matrices. This approach enables the model to focus more on important information and reduce noise. WDformer has achieved state-of-the-art (SOTA) results on multiple challenging real-world datasets, demonstrating its accuracy and effectiveness. Code is available at https://github.com/xiaowangbc/WDformer.

Kerui Chen, Jianrong Zhang, Ming Li et al.

Existing stylized motion generation models have shown their remarkable ability to understand specific style information from the style motion, and insert it into the content motion. However, capturing intra-style diversity, where a single style should correspond to diverse motion variations, remains a significant challenge. In this paper, we propose a clustering-based framework, ClusterStyle, to address this limitation. Instead of learning an unstructured embedding from each style motion, we leverage a set of prototypes to effectively model diverse style patterns across motions belonging to the same style category. We consider two types of style diversity: global-level diversity among style motions of the same category, and local-level diversity within the temporal dynamics of motion sequences. These components jointly shape two structured style embedding spaces, i.e., global and local, optimized via alignment with non-learnable prototype anchors. Furthermore, we augment the pretrained text-to-motion generation model with the Stylistic Modulation Adapter (SMA) to integrate the style features. Extensive experiments demonstrate that our approach outperforms existing state-of-the-art models in stylized motion generation and motion style transfer.

Chang Wan, Ming-Hsuan Yang, Minglu Li et al.

Recently, researchers have proposed many deep generative models, including generative adversarial networks(GANs) and denoising diffusion models. Although significant breakthroughs have been made and empirical success has been achieved with the GAN, its mathematical underpinnings remain relatively unknown. This paper focuses on a rigorous mathematical theoretical framework: the composite-functional-gradient GAN (CFG)[1]. Specifically, we reveal the theoretical connection between the CFG model and score-based models. We find that the training objective of the CFG discriminator is equivalent to finding an optimal D(x). The optimal gradient of D(x) differentiates the integral of the differences between the score functions of real and synthesized samples. Conversely, training the CFG generator involves finding an optimal G(x) that minimizes this difference. In this paper, we aim to derive an annealed weight preceding the weight of the CFG discriminator. This new explicit theoretical explanation model is called the annealed CFG method. To overcome the limitation of the annealed CFG method, as the method is not readily applicable to the SOTA GAN model, we propose a nested annealed training scheme (NATS). This scheme keeps the annealed weight from the CFG method and can be seamlessly adapted to various GAN models, no matter their structural, loss, or regularization differences. We conduct thorough experimental evaluations on various benchmark datasets for image generation. The results show that our annealed CFG and NATS methods significantly improve the quality and diversity of the synthesized samples. This improvement is clear when comparing the CFG method and the SOTA GAN models.

Chang Wan, Ke Fan, Xinwei Sun et al.

This paper introduces a promising alternative method for training Generative Adversarial Networks (GANs) on large-scale datasets with clear theoretical guarantees. GANs are typically learned through a minimax game between a generator and a discriminator, which is known to be empirically unstable. Previous learning paradigms have encountered mode collapse issues without a theoretical solution. To address these challenges, we propose a novel Lipschitz-constrained Functional Gradient GANs learning (Li-CFG) method to stabilize the training of GAN and provide a theoretical foundation for effectively increasing the diversity of synthetic samples by reducing the neighborhood size of the latent vector. Specifically, we demonstrate that the neighborhood size of the latent vector can be reduced by increasing the norm of the discriminator gradient, resulting in enhanced diversity of synthetic samples. To efficiently enlarge the norm of the discriminator gradient, we introduce a novel ε-centered gradient penalty that amplifies the norm of the discriminator gradient using the hyper-parameter ε. In comparison to other constraints, our method enlarging the discriminator norm, thus obtaining the smallest neighborhood size of the latent vector. Extensive experiments on benchmark datasets for image generation demonstrate the efficacy of the Li-CFG method and the ε-centered gradient penalty. The results showcase improved stability and increased diversity of synthetic samples.

Lu Wang, Jie Yang, Masoumeh Zareapoor et al.

Cross-modal hashing still has some challenges needed to address: (1) most existing CMH methods take graphs as input to model data distribution. These methods omit to consider the correlation of graph structure among multiple modalities; (2) most existing CMH methods ignores considering the fusion affinity among multi-modalities data; (3) most existing CMH methods relax the discrete constraints to solve the optimization objective, significantly degrading the retrieval performance. To solve the above limitations, we propose a novel Anchor Graph Structure Fusion Hashing (AGSFH). AGSFH constructs the anchor graph structure fusion matrix from different anchor graphs of multiple modalities with the Hadamard product, which can fully exploit the geometric property of underlying data structure. Based on the anchor graph structure fusion matrix, AGSFH attempts to directly learn an intrinsic anchor graph, where the structure of the intrinsic anchor graph is adaptively tuned so that the number of components of the intrinsic graph is exactly equal to the number of clusters. Besides, AGSFH preserves the anchor fusion affinity into the common binary Hamming space. Furthermore, a discrete optimization framework is designed to learn the unified binary codes. Extensive experimental results on three public social datasets demonstrate the superiority of AGSFH.