Le Wang

Zixin Zhu, Yixuan Wei, Jianfeng Wang et al. · microsoft-research

The image captioning task is typically realized by an auto-regressive method that decodes the text tokens one by one. We present a diffusion-based captioning model, dubbed the name DDCap, to allow more decoding flexibility. Unlike image generation, where the output is continuous and redundant with a fixed length, texts in image captions are categorical and short with varied lengths. Therefore, naively applying the discrete diffusion model to text decoding does not work well, as shown in our experiments. To address the performance gap, we propose several key techniques including best-first inference, concentrated attention mask, text length prediction, and image-free training. On COCO without additional caption pre-training, it achieves a CIDEr score of 117.8, which is +5.0 higher than the auto-regressive baseline with the same architecture in the controlled setting. It also performs +26.8 higher CIDEr score than the auto-regressive baseline (230.3 v.s.203.5) on a caption infilling task. With 4M vision-language pre-training images and the base-sized model, we reach a CIDEr score of 125.1 on COCO, which is competitive to the best well-developed auto-regressive frameworks. The code is available at https://github.com/buxiangzhiren/DDCap.

Zixin Zhu, Xuelu Feng, Dongdong Chen et al.

StableDiffusion is a revolutionary text-to-image generator that is causing a stir in the world of image generation and editing. Unlike traditional methods that learn a diffusion model in pixel space, StableDiffusion learns a diffusion model in the latent space via a VQGAN, ensuring both efficiency and quality. It not only supports image generation tasks, but also enables image editing for real images, such as image inpainting and local editing. However, we have observed that the vanilla VQGAN used in StableDiffusion leads to significant information loss, causing distortion artifacts even in non-edited image regions. To this end, we propose a new asymmetric VQGAN with two simple designs. Firstly, in addition to the input from the encoder, the decoder contains a conditional branch that incorporates information from task-specific priors, such as the unmasked image region in inpainting. Secondly, the decoder is much heavier than the encoder, allowing for more detailed recovery while only slightly increasing the total inference cost. The training cost of our asymmetric VQGAN is cheap, and we only need to retrain a new asymmetric decoder while keeping the vanilla VQGAN encoder and StableDiffusion unchanged. Our asymmetric VQGAN can be widely used in StableDiffusion-based inpainting and local editing methods. Extensive experiments demonstrate that it can significantly improve the inpainting and editing performance, while maintaining the original text-to-image capability. The code is available at \url{https://github.com/buxiangzhiren/Asymmetric_VQGAN}.

Zheng Qin, Sanping Zhou, Le Wang et al.

The main challenge of Multi-Object Tracking~(MOT) lies in maintaining a continuous trajectory for each target. Existing methods often learn reliable motion patterns to match the same target between adjacent frames and discriminative appearance features to re-identify the lost targets after a long period. However, the reliability of motion prediction and the discriminability of appearances can be easily hurt by dense crowds and extreme occlusions in the tracking process. In this paper, we propose a simple yet effective multi-object tracker, i.e., MotionTrack, which learns robust short-term and long-term motions in a unified framework to associate trajectories from a short to long range. For dense crowds, we design a novel Interaction Module to learn interaction-aware motions from short-term trajectories, which can estimate the complex movement of each target. For extreme occlusions, we build a novel Refind Module to learn reliable long-term motions from the target's history trajectory, which can link the interrupted trajectory with its corresponding detection. Our Interaction Module and Refind Module are embedded in the well-known tracking-by-detection paradigm, which can work in tandem to maintain superior performance. Extensive experimental results on MOT17 and MOT20 datasets demonstrate the superiority of our approach in challenging scenarios, and it achieves state-of-the-art performances at various MOT metrics.

Yongzi Yu, Ao Li, Le Wang et al.

Large language model (LLM) agents have shown promise in automating complex data-analysis workflows, but their reliable deployment remains challenging in high-stakes industrial scenarios. Industrial anomaly detection (IAD) is essential for manufacturing quality, safety, and efficiency, yet existing LLM-based IAD agents mainly focus on execution while under-exploiting strategy formulation. Consequently, they struggle to handle heterogeneous modalities in a unified and cost-effective manner. Inspired by the DMAIC quality-management framework, we propose DMAIC-IAD (DMAIC-inspired Agentic Industrial Anomaly Detection), a "Plan First, Judge Later" multi-agent system that aligns LLM agents with structured industrial problem-solving. DMAIC-IAD distills heterogeneous references into standardized operating procedures (SOPs) before strategy generation, and introduces a pre-trained execution-free judge model to rank candidate strategies without costly runtime trials. Extensive experiments across four modalities show that DMAIC-IAD improves average detection performance over applicable agentic baselines by 37.76%.

Liushuai Shi, Le Wang, Chengjiang Long et al.

Understanding the multiple socially-acceptable future behaviors is an essential task for many vision applications. In this paper, we propose a tree-based method, termed as Social Interpretable Tree (SIT), to address this multi-modal prediction task, where a hand-crafted tree is built depending on the prior information of observed trajectory to model multiple future trajectories. Specifically, a path in the tree from the root to leaf represents an individual possible future trajectory. SIT employs a coarse-to-fine optimization strategy, in which the tree is first built by high-order velocity to balance the complexity and coverage of the tree and then optimized greedily to encourage multimodality. Finally, a teacher-forcing refining operation is used to predict the final fine trajectory. Compared with prior methods which leverage implicit latent variables to represent possible future trajectories, the path in the tree can explicitly explain the rough moving behaviors (e.g., go straight and then turn right), and thus provides better interpretability. Despite the hand-crafted tree, the experimental results on ETH-UCY and Stanford Drone datasets demonstrate that our method is capable of matching or exceeding the performance of state-of-the-art methods. Interestingly, the experiments show that the raw built tree without training outperforms many prior deep neural network based approaches. Meanwhile, our method presents sufficient flexibility in long-term prediction and different best-of-$K$ predictions.

Yabing Wang, Le Wang, Qiang Zhou et al.

Cross-lingual cross-modal retrieval (CCR) aims to retrieve visually relevant content based on non-English queries, without relying on human-labeled cross-modal data pairs during training. One popular approach involves utilizing machine translation (MT) to create pseudo-parallel data pairs, establishing correspondence between visual and non-English textual data. However, aligning their representations poses challenges due to the significant semantic gap between vision and text, as well as the lower quality of non-English representations caused by pre-trained encoders and data noise. To overcome these challenges, we propose LECCR, a novel solution that incorporates the multi-modal large language model (MLLM) to improve the alignment between visual and non-English representations. Specifically, we first employ MLLM to generate detailed visual content descriptions and aggregate them into multi-view semantic slots that encapsulate different semantics. Then, we take these semantic slots as internal features and leverage them to interact with the visual features. By doing so, we enhance the semantic information within the visual features, narrowing the semantic gap between modalities and generating local visual semantics for subsequent multi-level matching. Additionally, to further enhance the alignment between visual and non-English features, we introduce softened matching under English guidance. This approach provides more comprehensive and reliable inter-modal correspondences between visual and non-English features. Extensive experiments on four CCR benchmarks, \ie Multi30K, MSCOCO, VATEX, and MSR-VTT-CN, demonstrate the effectiveness of our proposed method. Code: \url{https://github.com/LiJiaBei-7/leccr}.

Jianxiang Xi, Cheng Wang, Hao Liu et al.

The guaranteed-performance consensualization for high-order linear and nonlinear multiagent systems with switching topologies is respectively realized in a completely distributed manner in the sense that consensus design criteria are independent of interaction topologies and switching motions. The current paper firstly proposes an adaptive consensus protocol with guaranteed-performance constraints and switching topologies, where interaction weights among neighboring agents are adaptively adjusted and state errors among all agents can be regulated. Then, a new translation-adaptive strategy is shown to realize completely distributed guaranteed-performance consensus control and an adaptive guaranteed-performance consensualization criterion is given on the basis of the Riccati inequality. Furthermore, an approach to regulate the consensus control gain and the guaranteed-performance cost is proposed in terms of linear matrix inequalities. Moreover, main conclusions for linear multiagent systems are extended to Lipschitz nonlinear cases. Finally, two numerical examples are provided to demonstrate theoretical results.

Kun Xia, Le Wang, Sanping Zhou et al.

The main challenge of Temporal Action Localization is to retrieve subtle human actions from various co-occurring ingredients, e.g., context and background, in an untrimmed video. While prior approaches have achieved substantial progress through devising advanced action detectors, they still suffer from these co-occurring ingredients which often dominate the actual action content in videos. In this paper, we explore two orthogonal but complementary aspects of a video snippet, i.e., the action features and the co-occurrence features. Especially, we develop a novel auxiliary task by decoupling these two types of features within a video snippet and recombining them to generate a new feature representation with more salient action information for accurate action localization. We term our method RefactorNet, which first explicitly factorizes the action content and regularizes its co-occurrence features, and then synthesizes a new action-dominated video representation. Extensive experimental results and ablation studies on THUMOS14 and ActivityNet v1.3 demonstrate that our new representation, combined with a simple action detector, can significantly improve the action localization performance.

Ning Gao, Sanping Zhou, Le Wang et al.

Semi-supervised learning has emerged as a widely adopted technique in the field of medical image segmentation. The existing works either focuses on the construction of consistency constraints or the generation of pseudo labels to provide high-quality supervisory signals, whose main challenge mainly comes from how to keep the continuous improvement of model capabilities. In this paper, we propose a simple yet effective semi-supervised learning framework, termed Progressive Mean Teachers (PMT), for medical image segmentation, whose goal is to generate high-fidelity pseudo labels by learning robust and diverse features in the training process. Specifically, our PMT employs a standard mean teacher to penalize the consistency of the current state and utilizes two sets of MT architectures for co-training. The two sets of MT architectures are individually updated for prolonged periods to maintain stable model diversity established through performance gaps generated by iteration differences. Additionally, a difference-driven alignment regularizer is employed to expedite the alignment of lagging models with the representation capabilities of leading models. Furthermore, a simple yet effective pseudo-label filtering algorithm is employed for facile evaluation of models and selection of high-fidelity pseudo-labels outputted when models are operating at high performance for co-training purposes. Experimental results on two datasets with different modalities, i.e., CT and MRI, demonstrate that our method outperforms the state-of-the-art medical image segmentation approaches across various dimensions. The code is available at https://github.com/Axi404/PMT.

Zheyuan Zhou, Le Wang, Naiyu Fang et al.

3D anomaly detection plays a crucial role in monitoring parts for localized inherent defects in precision manufacturing. Embedding-based and reconstruction-based approaches are among the most popular and successful methods. However, there are two major challenges to the practical application of the current approaches: 1) the embedded models suffer the prohibitive computational and storage due to the memory bank structure; 2) the reconstructive models based on the MAE mechanism fail to detect anomalies in the unmasked regions. In this paper, we propose R3D-AD, reconstructing anomalous point clouds by diffusion model for precise 3D anomaly detection. Our approach capitalizes on the data distribution conversion of the diffusion process to entirely obscure the input's anomalous geometry. It step-wisely learns a strict point-level displacement behavior, which methodically corrects the aberrant points. To increase the generalization of the model, we further present a novel 3D anomaly simulation strategy named Patch-Gen to generate realistic and diverse defect shapes, which narrows the domain gap between training and testing. Our R3D-AD ensures a uniform spatial transformation, which allows straightforwardly generating anomaly results by distance comparison. Extensive experiments show that our R3D-AD outperforms previous state-of-the-art methods, achieving 73.4% Image-level AUROC on the Real3D-AD dataset and 74.9% Image-level AUROC on the Anomaly-ShapeNet dataset with an exceptional efficiency.

Yizhe Li, Sanping Zhou, Zheng Qin et al.

Multi-Object Tracking (MOT) remains a vital component of intelligent video analysis, which aims to locate targets and maintain a consistent identity for each target throughout a video sequence. Existing works usually learn a discriminative feature representation, such as motion and appearance, to associate the detections across frames, which are easily affected by mutual occlusion and background clutter in practice. In this paper, we propose a simple yet effective two-stage feature learning paradigm to jointly learn single-shot and multi-shot features for different targets, so as to achieve robust data association in the tracking process. For the detections without being associated, we design a novel single-shot feature learning module to extract discriminative features of each detection, which can efficiently associate targets between adjacent frames. For the tracklets being lost several frames, we design a novel multi-shot feature learning module to extract discriminative features of each tracklet, which can accurately refind these lost targets after a long period. Once equipped with a simple data association logic, the resulting VisualTracker can perform robust MOT based on the single-shot and multi-shot feature representations. Extensive experimental results demonstrate that our method has achieved significant improvements on MOT17 and MOT20 datasets while reaching state-of-the-art performance on DanceTrack dataset.

Bin Zhu, Zhaoquan Gu, Le Wang et al.

Recent studies have found that removing the norm-bounded projection and increasing search steps in adversarial training can significantly improve robustness. However, we observe that a too large number of search steps can hurt accuracy. We aim to obtain strong robustness efficiently using fewer steps. Through a toy experiment, we find that perturbing the clean data to the decision boundary but not crossing it does not degrade the test accuracy. Inspired by this, we propose friendly adversarial data augmentation (FADA) to generate friendly adversarial data. On top of FADA, we propose geometry-aware adversarial training (GAT) to perform adversarial training on friendly adversarial data so that we can save a large number of search steps. Comprehensive experiments across two widely used datasets and three pre-trained language models demonstrate that GAT can obtain stronger robustness via fewer steps. In addition, we provide extensive empirical results and in-depth analyses on robustness to facilitate future studies.

Yonghao Dong, Le Wang, Sanpin Zhou et al.

Pedestrian trajectory prediction in a first-person view has recently attracted much attention due to its importance in autonomous driving. Recent work utilizes pedestrian character information, \textit{i.e.}, action and appearance, to improve the learned trajectory embedding and achieves state-of-the-art performance. However, it neglects the invalid and negative pedestrian character information, which is harmful to trajectory representation and thus leads to performance degradation. To address this issue, we present a two-stream sparse-character-based network~(TSNet) for pedestrian trajectory prediction. Specifically, TSNet learns the negative-removed characters in the sparse character representation stream to improve the trajectory embedding obtained in the trajectory representation stream. Moreover, to model the negative-removed characters, we propose a novel sparse character graph, including the sparse category and sparse temporal character graphs, to learn the different effects of various characters in category and temporal dimensions, respectively. Extensive experiments on two first-person view datasets, PIE and JAAD, show that our method outperforms existing state-of-the-art methods. In addition, ablation studies demonstrate different effects of various characters and prove that TSNet outperforms approaches without eliminating negative characters.

Chang Sun, Zili Wang, Shuyou Zhang et al.

Bi-layer metallic tube (BMT) plays an extremely crucial role in engineering applications, with rotary draw bending (RDB) the high-precision bending processing can be achieved, however, the product will further springback. Due to the complex structure of BMT and the high cost of dataset acquisi-tion, the existing methods based on mechanism research and machine learn-ing cannot meet the engineering requirements of springback prediction. Based on the preliminary mechanism analysis, a physical logic enhanced network (PE-NET) is proposed. The architecture includes ES-NET which equivalent the BMT to the single-layer tube, and SP-NET for the final predic-tion of springback with sufficient single-layer tube samples. Specifically, in the first stage, with the theory-driven pre-exploration and the data-driven pretraining, the ES-NET and SP-NET are constructed, respectively. In the second stage, under the physical logic, the PE-NET is assembled by ES-NET and SP-NET and then fine-tuned with the small sample BMT dataset and composite loss function. The validity and stability of the proposed method are verified by the FE simulation dataset, the small-sample dataset BMT springback angle prediction is achieved, and the method potential in inter-pretability and engineering applications are demonstrated.

Arman H Ter-Petrosyan, Jenna A Bilbrey, Christina M Doty et al.

We present a method for the unsupervised segmentation of electron microscopy images, which are powerful descriptors of materials and chemical systems. Images are oversegmented into overlapping chips, and similarity graphs are generated from embeddings extracted from a domain$\unicode{x2010}$pretrained convolutional neural network (CNN). The Louvain method for community detection is then applied to perform segmentation. The graph representation provides an intuitive way of presenting the relationship between chips and communities. We demonstrate our method to track irradiation$\unicode{x2010}$induced amorphous fronts in thin films used for catalysis and electronics. This method has potential for "on$\unicode{x2010}$the$\unicode{x2010}$fly" segmentation to guide emerging automated electron microscopes.

Sen Wang, Huaiyi Dong, Jingyi Tian et al.

Prevailing 2D-centric visuomotor policies exhibit a pronounced deficiency in novel view generalization, as their reliance on static observations hinders consistent action mapping across unseen views. In response, we introduce GenSplat, a feed-forward 3D Gaussian Splatting framework that facilitates view-generalized policy learning through novel view rendering. GenSplat employs a permutation-equivariant architecture to reconstruct high-fidelity 3D scenes from sparse, uncalibrated inputs in a single forward pass. To ensure structural integrity, we design a 3D-prior distillation strategy that regularizes the 3DGS optimization, preventing the geometric collapse typical of purely photometric supervision. By rendering diverse synthetic views from these stable 3D representations, we systematically augment the observational manifold during training. This augmentation forces the policy to ground its decisions in underlying 3D structures, thereby ensuring robust execution under severe spatial perturbations where baselines severely degrade.

Jiexia Ye, Yongzi Yu, Weiqi Zhang et al.

Time series data are ubiquitous across diverse real-world applications, making time series analysis critically important. Traditional approaches are largely task-specific, offering limited functionality and poor transferability. In recent years, foundation models have revolutionized NLP and CV with their remarkable cross-task transferability, zero-/few-shot learning capabilities, and multimodal integration capacity. This success has motivated increasing efforts to explore foundation models for addressing time series modeling challenges. Although some tutorials and surveys were published in the early stages of this field, the rapid pace of recent developments necessitates a more comprehensive and in-depth synthesis to cover the latest advances. Our survey aims to fill this gap by introducing a modality-aware, challenge-oriented perspective, which reveals how foundation models pre-trained on different modalities face distinct hurdles when adapted to time series tasks. Building on this perspective, we propose a taxonomy of existing works organized by pre-training modality (time series, language, and vision), analyze modality-specific challenges and categorize corresponding solutions, discussing their advantages and limitations. Beyond this, we review real-world applications to illustrate domain-specific advancements, provide open-source codes, and conclude with potential future research directions in this rapidly evolving field.

Zeyang Liu, Le Wang, Sanping Zhou et al.

Content-aware layout generation is a critical task in graphic design automation, focused on creating visually appealing arrangements of elements that seamlessly blend with a given background image. The variety of real-world applications makes it highly challenging to develop a single model capable of unifying the diverse range of input-constrained generation sub-tasks, such as those conditioned by element types, sizes, or their relationships. Current methods either address only a subset of these tasks or necessitate separate model parameters for different conditions, failing to offer a truly unified solution. In this paper, we propose UniLayDiff: a Unified Diffusion Transformer, that for the first time, addresses various content-aware layout generation tasks with a single, end-to-end trainable model. Specifically, we treat layout constraints as a distinct modality and employ Multi-Modal Diffusion Transformer framework to capture the complex interplay between the background image, layout elements, and diverse constraints. Moreover, we integrate relation constraints through fine-tuning the model with LoRA after pretraining the model on other tasks. Such a schema not only achieves unified conditional generation but also enhances overall layout quality. Extensive experiments demonstrate that UniLayDiff achieves state-of-the-art performance across from unconditional to various conditional generation tasks and, to the best of our knowledge, is the first model to unify the full range of content-aware layout generation tasks.

Zonglei Jing, Zonghao Ying, Le Wang et al.

The development of text-to-image (T2I) generative models, that enable the creation of high-quality synthetic images from textual prompts, has opened new frontiers in creative design and content generation. However, this paper reveals a significant and previously unrecognized ethical risk inherent in this technology and introduces a novel method, termed the Cognitive Morphing Attack (CogMorph), which manipulates T2I models to generate images that retain the original core subjects but embeds toxic or harmful contextual elements. This nuanced manipulation exploits the cognitive principle that human perception of concepts is shaped by the entire visual scene and its context, producing images that amplify emotional harm far beyond attacks that merely preserve the original semantics. To address this, we first construct an imagery toxicity taxonomy spanning 10 major and 48 sub-categories, aligned with human cognitive-perceptual dimensions, and further build a toxicity risk matrix resulting in 1,176 high-quality T2I toxic prompts. Based on this, our CogMorph first introduces Cognitive Toxicity Augmentation, which develops a cognitive toxicity knowledge base with rich external toxic representations for humans (e.g., fine-grained visual features) that can be utilized to further guide the optimization of adversarial prompts. In addition, we present Contextual Hierarchical Morphing, which hierarchically extracts critical parts of the original prompt (e.g., scenes, subjects, and body parts), and then iteratively retrieves and fuses toxic features to inject harmful contexts. Extensive experiments on multiple open-sourced T2I models and black-box commercial APIs (e.g., DALLE-3) demonstrate the efficacy of CogMorph which significantly outperforms other baselines by large margins (+20.62% on average).

Jun Wang, Xijuan Zeng, Chunyu Qiang et al.

We propose Kling-Foley, a large-scale multimodal Video-to-Audio generation model that synthesizes high-quality audio synchronized with video content. In Kling-Foley, we introduce multimodal diffusion transformers to model the interactions between video, audio, and text modalities, and combine it with a visual semantic representation module and an audio-visual synchronization module to enhance alignment capabilities. Specifically, these modules align video conditions with latent audio elements at the frame level, thereby improving semantic alignment and audio-visual synchronization. Together with text conditions, this integrated approach enables precise generation of video-matching sound effects. In addition, we propose a universal latent audio codec that can achieve high-quality modeling in various scenarios such as sound effects, speech, singing, and music. We employ a stereo rendering method that imbues synthesized audio with a spatial presence. At the same time, in order to make up for the incomplete types and annotations of the open-source benchmark, we also open-source an industrial-level benchmark Kling-Audio-Eval. Our experiments show that Kling-Foley trained with the flow matching objective achieves new audio-visual SOTA performance among public models in terms of distribution matching, semantic alignment, temporal alignment and audio quality.

Le Wang, Zonghao Ying, Xiao Yang et al.

Embodied agents powered by vision-language models (VLMs) are increasingly capable of executing complex real-world tasks, yet they remain vulnerable to hazardous instructions that may trigger unsafe behaviors. Runtime safety guardrails, which intercept hazardous actions during task execution, offer a promising solution due to their flexibility. However, existing defenses often rely on static rule filters or prompt-level control, which struggle to address implicit risks arising in dynamic, temporally dependent, and context-rich environments. To address this, we propose RoboSafe, a hybrid reasoning runtime safeguard for embodied agents through executable predicate-based safety logic. RoboSafe integrates two complementary reasoning processes on a Hybrid Long-Short Safety Memory. We first propose a Backward Reflective Reasoning module that continuously revisits recent trajectories in short-term memory to infer temporal safety predicates and proactively triggers replanning when violations are detected. We then propose a Forward Predictive Reasoning module that anticipates upcoming risks by generating context-aware safety predicates from the long-term safety memory and the agent's multimodal observations. Together, these components form an adaptive, verifiable safety logic that is both interpretable and executable as code. Extensive experiments across multiple agents demonstrate that RoboSafe substantially reduces hazardous actions (-36.8% risk occurrence) compared with leading baselines, while maintaining near-original task performance. Real-world evaluations on physical robotic arms further confirm its practicality. Code will be released upon acceptance.

Qinggang Zhang, Zhishang Xiang, Yilin Xiao et al.

Large language models (LLMs) augmented with retrieval systems have demonstrated significant potential in handling knowledge-intensive tasks. However, these models often struggle with unfaithfulness issues, generating outputs that either ignore the retrieved context or inconsistently blend it with the LLM`s parametric knowledge. This issue is particularly severe in cases of knowledge conflict, where the retrieved context conflicts with the model`s parametric knowledge. While existing faithful RAG approaches enforce strict context adherence through well-designed prompts or modified decoding strategies, our analysis reveals a critical limitation: they achieve faithfulness by forcibly suppressing the model`s parametric knowledge, which undermines the model`s internal knowledge structure and increases the risk of misinterpreting the context. To this end, this paper proposes FaithfulRAG, a novel framework that resolves knowledge conflicts by explicitly modeling discrepancies between the model`s parametric knowledge and retrieved context. Specifically, FaithfulRAG identifies conflicting knowledge at the fact level and designs a self-thinking process, allowing LLMs to reason about and integrate conflicting facts before generating responses. Extensive experiments demonstrate that our method outperforms state-of-the-art methods. The code is available at https://github.com/DeepLearnXMU/Faithful-RAG

Linfeng Gao, Baolong Bi, Zheng Yuan et al.

Retrieval-Augmented Generation (RAG) has emerged as a powerful paradigm to enhance the factuality of Large Language Models (LLMs). However, existing RAG systems often suffer from an unfaithfulness issue, where the model's response contradicts evidence from the retrieved context. Existing approaches to improving contextual faithfulness largely rely on external interventions, such as prompt engineering, decoding constraints, or reward-based fine-tuning. These works treat the LLM as a black box and overlook a crucial question: how does the LLM internally integrate retrieved evidence with its parametric memory, particularly under knowledge conflicts? To address this gap, we conduct a probing-based analysis of hidden-state representations in LLMs and observe three findings: knowledge integration occurs hierarchically, conflicts manifest as latent signals at the sentence level, and irrelevant context is often amplified when aligned with parametric knowledge. Building on these findings, we propose CLEAR (Conflict-Localized and Enhanced Attention for RAG), a framework that (i) decomposes context into fine-grained sentence-level knowledge, (ii) employs hidden-state probing to localize conflicting knowledge, and (iii) introduces conflict-aware fine-tuning to guide the model to accurately integrate retrieved evidence. Extensive experiments across three benchmarks demonstrate that CLEAR substantially improves both accuracy and contextual faithfulness, consistently outperforming strong baselines under diverse conflict conditions. The related resources are available at https://github.com/LinfengGao/CLEAR.

Zheng Qin, Le Wang, Sanping Zhou et al.

Multi-Object Tracking MOT encompasses various tracking scenarios, each characterized by unique traits. Effective trackers should demonstrate a high degree of generalizability across diverse scenarios. However, existing trackers struggle to accommodate all aspects or necessitate hypothesis and experimentation to customize the association information motion and or appearance for a given scenario, leading to narrowly tailored solutions with limited generalizability. In this paper, we investigate the factors that influence trackers generalization to different scenarios and concretize them into a set of tracking scenario attributes to guide the design of more generalizable trackers. Furthermore, we propose a point-wise to instance-wise relation framework for MOT, i.e., GeneralTrack, which can generalize across diverse scenarios while eliminating the need to balance motion and appearance. Thanks to its superior generalizability, our proposed GeneralTrack achieves state-of-the-art performance on multiple benchmarks and demonstrates the potential for domain generalization. https://github.com/qinzheng2000/GeneralTrack.git

Zhixuan Zhang, Chi Zhang, Zhenning Niu et al.

In this manuscript, we introduce a semi-automatic scene graph annotation tool for images, the GeneAnnotator. This software allows human annotators to describe the existing relationships between participators in the visual scene in the form of directed graphs, hence enabling the learning and reasoning on visual relationships, e.g., image captioning, VQA and scene graph generation, etc. The annotations for certain image datasets could either be merged in a single VG150 data-format file to support most existing models for scene graph learning or transformed into a separated annotation file for each single image to build customized datasets. Moreover, GeneAnnotator provides a rule-based relationship recommending algorithm to reduce the heavy annotation workload. With GeneAnnotator, we propose Traffic Genome, a comprehensive scene graph dataset with 1000 diverse traffic images, which in return validates the effectiveness of the proposed software for scene graph annotation. The project source code, with usage examples and sample data is available at https://github.com/Milomilo0320/A-Semi-automatic-Annotation-Software-for-Scene-Graph, under the Apache open-source license.

Zixin Zhu, Wei Tang, Le Wang et al.

Effectively tackling the problem of temporal action localization (TAL) necessitates a visual representation that jointly pursues two confounding goals, i.e., fine-grained discrimination for temporal localization and sufficient visual invariance for action classification. We address this challenge by enriching both the local and global contexts in the popular two-stage temporal localization framework, where action proposals are first generated followed by action classification and temporal boundary regression. Our proposed model, dubbed ContextLoc, can be divided into three sub-networks: L-Net, G-Net and P-Net. L-Net enriches the local context via fine-grained modeling of snippet-level features, which is formulated as a query-and-retrieval process. G-Net enriches the global context via higher-level modeling of the video-level representation. In addition, we introduce a novel context adaptation module to adapt the global context to different proposals. P-Net further models the context-aware inter-proposal relations. We explore two existing models to be the P-Net in our experiments. The efficacy of our proposed method is validated by experimental results on the THUMOS14 (54.3\% at tIoU@0.5) and ActivityNet v1.3 (56.01\% at tIoU@0.5) datasets, which outperforms recent states of the art. Code is available at https://github.com/buxiangzhiren/ContextLoc.

Le Wang, Zonghao Ying, Tianyuan Zhang et al.

The emergence of multimodal large language models has redefined the agent paradigm by integrating language and vision modalities with external data sources, enabling agents to better interpret human instructions and execute increasingly complex tasks. However, in this paper, we identify a critical yet previously overlooked security vulnerability in multimodal agents: cross-modal prompt injection attacks. To exploit this vulnerability, we propose CrossInject, a novel attack framework in which attackers embed adversarial perturbations across multiple modalities to align with target malicious content, allowing external instructions to hijack the agent's decision-making process and execute unauthorized tasks. Our approach incorporates two key coordinated components. First, we introduce Visual Latent Alignment, where we optimize adversarial features to the malicious instructions in the visual embedding space based on a text-to-image generative model, ensuring that adversarial images subtly encode cues for malicious task execution. Subsequently, we present Textual Guidance Enhancement, where a large language model is leveraged to construct the black-box defensive system prompt through adversarial meta prompting and generate an malicious textual command that steers the agent's output toward better compliance with attackers' requests. Extensive experiments demonstrate that our method outperforms state-of-the-art attacks, achieving at least a +30.1% increase in attack success rates across diverse tasks. Furthermore, we validate our attack's effectiveness in real-world multimodal autonomous agents, highlighting its potential implications for safety-critical applications.

Yonghao Dong, Le Wang, Sanping Zhou et al.

Pedestrian trajectory prediction is a crucial component in computer vision and robotics, but remains challenging due to the domain shift problem. Previous studies have tried to tackle this problem by leveraging a portion of the trajectory data from the target domain to adapt the model. However, such domain adaptation methods are impractical in real-world scenarios, as it is infeasible to collect trajectory data from all potential target domains. In this paper, we study a task named generalized pedestrian trajectory prediction, with the aim of generalizing the model to unseen domains without accessing their trajectories. To tackle this task, we introduce a Recurrent Aligned Network~(RAN) to minimize the domain gap through domain alignment. Specifically, we devise a recurrent alignment module to effectively align the trajectory feature spaces at both time-state and time-sequence levels by the recurrent alignment strategy.Furthermore, we introduce a pre-aligned representation module to combine social interactions with the recurrent alignment strategy, which aims to consider social interactions during the alignment process instead of just target trajectories. We extensively evaluate our method and compare it with state-of-the-art methods on three widely used benchmarks. The experimental results demonstrate the superior generalization capability of our method. Our work not only fills the gap in the generalization setting for practical pedestrian trajectory prediction but also sets strong baselines in this field.

Jing-Yang Wei, Hao Huang, Xin Zhang et al.

To provide a lightweight and cost-effective solution for the long-wave infrared imaging using a singlet, we develop a camera by integrating a High-Frequency-Enhancing Cycle-GAN neural network into a metalens imaging system. The High-Frequency-Enhancing Cycle-GAN improves the quality of the original metalens images by addressing inherent frequency loss introduced by the metalens. In addition to the bidirectional cyclic generative adversarial network, it incorporates a high-frequency adversarial learning module. This module utilizes wavelet transform to extract high-frequency components, and then establishes a high-frequency feedback loop. It enables the generator to enhance the camera outputs by integrating adversarial feedback from the high-frequency discriminator. This ensures that the generator adheres to the constraints imposed by the high-frequency adversarial loss, thereby effectively recovering the camera's frequency loss. This recovery guarantees high-fidelity image output from the camera, facilitating smooth video production. Our camera is capable of achieving dynamic imaging at 125 frames per second with an End Point Error value of 12.58. We also achieve 0.42 for Fréchet Inception Distance, 30.62 for Peak Signal to Noise Ratio, and 0.69 for Structural Similarity in the recorded videos.

Sen Wang, Le Wang, Sanping Zhou et al.

Robotic manipulation in high-precision tasks is essential for numerous industrial and real-world applications where accuracy and speed are required. Yet current diffusion-based policy learning methods generally suffer from low computational efficiency due to the iterative denoising process during inference. Moreover, these methods do not fully explore the potential of generative models for enhancing information exploration in 3D environments. In response, we propose FlowRAM, a novel framework that leverages generative models to achieve region-aware perception, enabling efficient multimodal information processing. Specifically, we devise a Dynamic Radius Schedule, which allows adaptive perception, facilitating transitions from global scene comprehension to fine-grained geometric details. Furthermore, we integrate state space models to integrate multimodal information, while preserving linear computational complexity. In addition, we employ conditional flow matching to learn action poses by regressing deterministic vector fields, simplifying the learning process while maintaining performance. We verify the effectiveness of the FlowRAM in the RLBench, an established manipulation benchmark, and achieve state-of-the-art performance. The results demonstrate that FlowRAM achieves a remarkable improvement, particularly in high-precision tasks, where it outperforms previous methods by 12.0% in average success rate. Additionally, FlowRAM is able to generate physically plausible actions for a variety of real-world tasks in less than 4 time steps, significantly increasing inference speed.

Zheng Qin, Ruobing Zheng, Yabing Wang et al.

In filmmaking, directors typically allow actors to perform freely based on the script before providing specific guidance on how to present key actions. AI-generated content faces similar requirements, where users not only need automatic generation of lip synchronization and basic gestures from audio input but also desire semantically accurate and expressive body movement that can be ``directly guided'' through text descriptions. Therefore, we present VersaAnimator, a versatile framework that synthesizes expressive talking human videos from arbitrary portrait images. Specifically, we design a motion generator that produces basic rhythmic movements from audio input and supports text-prompt control for specific actions. The generated whole-body 3D motion tokens can animate portraits of various scales, producing talking heads, half-body gestures and even leg movements for whole-body images. Besides, we introduce a multi-modal controlled video diffusion that generates photorealistic videos, where speech signals govern lip synchronization, facial expressions, and head motions while body movements are guided by the 2D poses. Furthermore, we introduce a token2pose translator to smoothly map 3D motion tokens to 2D pose sequences. This design mitigates the stiffness resulting from direct 3D to 2D conversion and enhances the details of the generated body movements. Extensive experiments shows that VersaAnimator synthesizes lip-synced and identity-preserving videos while generating expressive and semantically meaningful whole-body motions.

Ye Niu, Sanping Zhou, Yizhe Li et al.

In many complex scenarios, robotic manipulation relies on generative models to estimate the distribution of multiple successful actions. As the diffusion model has better training robustness than other generative models, it performs well in imitation learning through successful robot demonstrations. However, the diffusion-based policy methods typically require significant time to iteratively denoise robot actions, which hinders real-time responses in robotic manipulation. Moreover, existing diffusion policies model a time-varying action denoising process, whose temporal complexity increases the difficulty of model training and leads to suboptimal action accuracy. To generate robot actions efficiently and accurately, we present the Time-Unified Diffusion Policy (TUDP), which utilizes action recognition capabilities to build a time-unified denoising process. On the one hand, we build a time-unified velocity field in action space with additional action discrimination information. By unifying all timesteps of action denoising, our velocity field reduces the difficulty of policy learning and speeds up action generation. On the other hand, we propose an action-wise training method, which introduces an action discrimination branch to supply additional action discrimination information. Through action-wise training, the TUDP implicitly learns the ability to discern successful actions to better denoising accuracy. Our method achieves state-of-the-art performance on RLBench with the highest success rate of 82.6% on a multi-view setup and 83.8% on a single-view setup. In particular, when using fewer denoising iterations, TUDP achieves a more significant improvement in success rate. Additionally, TUDP can produce accurate actions for a wide range of real-world tasks.

Yabing Wang, Zhuotao Tian, Qingpei Guo et al.

Composed Image Retrieval (CIR) is a challenging multimodal task that retrieves a target image based on a reference image and accompanying modification text. Due to the high cost of annotating CIR triplet datasets, zero-shot (ZS) CIR has gained traction as a promising alternative. Existing studies mainly focus on projection-based methods, which map an image to a single pseudo-word token. However, these methods face three critical challenges: (1) insufficient pseudo-word token representation capacity, (2) discrepancies between training and inference phases, and (3) reliance on large-scale synthetic data. To address these issues, we propose a two-stage framework where the training is accomplished from mapping to composing. In the first stage, we enhance image-to-pseudo-word token learning by introducing a visual semantic injection module and a soft text alignment objective, enabling the token to capture richer and fine-grained image information. In the second stage, we optimize the text encoder using a small amount of synthetic triplet data, enabling it to effectively extract compositional semantics by combining pseudo-word tokens with modification text for accurate target image retrieval. The strong visual-to-pseudo mapping established in the first stage provides a solid foundation for the second stage, making our approach compatible with both high- and low-quality synthetic data, and capable of achieving significant performance gains with only a small amount of synthetic data. Extensive experiments were conducted on three public datasets, achieving superior performance compared to existing approaches.

Zihan Lan, Weixin Mao, Haosheng Li et al.

In real-world scenarios, multi-view cameras are typically employed for fine-grained manipulation tasks. Existing approaches (e.g., ACT) tend to treat multi-view features equally and directly concatenate them for policy learning. However, it will introduce redundant visual information and bring higher computational costs, leading to ineffective manipulation. For a fine-grained manipulation task, it tends to involve multiple stages while the most contributed view for different stages is varied over time. In this paper, we propose a plug-and-play best-feature-aware (BFA) fusion strategy for multi-view manipulation tasks, which is adaptable to various policies. Built upon the visual backbone of the policy network, we design a lightweight network to predict the importance score of each view. Based on the predicted importance scores, the reweighted multi-view features are subsequently fused and input into the end-to-end policy network, enabling seamless integration. Notably, our method demonstrates outstanding performance in fine-grained manipulations. Experimental results show that our approach outperforms multiple baselines by 22-46% success rate on different tasks. Our work provides new insights and inspiration for tackling key challenges in fine-grained manipulations.

Yabing Wang, Zhuotao Tian, Qingpei Guo et al.

Visual Grounding aims to localize the referring object in an image given a natural language expression. Recent advancements in DETR-based visual grounding methods have attracted considerable attention, as they directly predict the coordinates of the target object without relying on additional efforts, such as pre-generated proposal candidates or pre-defined anchor boxes. However, existing research primarily focuses on designing stronger multi-modal decoder, which typically generates learnable queries by random initialization or by using linguistic embeddings. This vanilla query generation approach inevitably increases the learning difficulty for the model, as it does not involve any target-related information at the beginning of decoding. Furthermore, they only use the deepest image feature during the query learning process, overlooking the importance of features from other levels. To address these issues, we propose a novel approach, called RefFormer. It consists of the query adaption module that can be seamlessly integrated into CLIP and generate the referential query to provide the prior context for decoder, along with a task-specific decoder. By incorporating the referential query into the decoder, we can effectively mitigate the learning difficulty of the decoder, and accurately concentrate on the target object. Additionally, our proposed query adaption module can also act as an adapter, preserving the rich knowledge within CLIP without the need to tune the parameters of the backbone network. Extensive experiments demonstrate the effectiveness and efficiency of our proposed method, outperforming state-of-the-art approaches on five visual grounding benchmarks.

Yu Wang, Sanping Zhou, Kun Xia et al.

Semi-supervised action recognition aims to improve spatio-temporal reasoning ability with a few labeled data in conjunction with a large amount of unlabeled data. Albeit recent advancements, existing powerful methods are still prone to making ambiguous predictions under scarce labeled data, embodied as the limitation of distinguishing different actions with similar spatio-temporal information. In this paper, we approach this problem by empowering the model two aspects of capability, namely discriminative spatial modeling and temporal structure modeling for learning discriminative spatio-temporal representations. Specifically, we propose an Adaptive Contrastive Learning~(ACL) strategy. It assesses the confidence of all unlabeled samples by the class prototypes of the labeled data, and adaptively selects positive-negative samples from a pseudo-labeled sample bank to construct contrastive learning. Additionally, we introduce a Multi-scale Temporal Learning~(MTL) strategy. It could highlight informative semantics from long-term clips and integrate them into the short-term clip while suppressing noisy information. Afterwards, both of these two new techniques are integrated in a unified framework to encourage the model to make accurate predictions. Extensive experiments on UCF101, HMDB51 and Kinetics400 show the superiority of our method over prior state-of-the-art approaches.

Kun Xia, Le Wang, Sanping Zhou et al.

The crux of semi-supervised temporal action localization (SS-TAL) lies in excavating valuable information from abundant unlabeled videos. However, current approaches predominantly focus on building models that are robust to the error-prone target class (i.e, the predicted class with the highest confidence) while ignoring informative semantics within non-target classes. This paper approaches SS-TAL from a novel perspective by advocating for learning from non-target classes, transcending the conventional focus solely on the target class. The proposed approach involves partitioning the label space of the predicted class distribution into distinct subspaces: target class, positive classes, negative classes, and ambiguous classes, aiming to mine both positive and negative semantics that are absent in the target class, while excluding ambiguous classes. To this end, we first devise innovative strategies to adaptively select high-quality positive and negative classes from the label space, by modeling both the confidence and rank of a class in relation to those of the target class. Then, we introduce novel positive and negative losses designed to guide the learning process, pushing predictions closer to positive classes and away from negative classes. Finally, the positive and negative processes are integrated into a hybrid positive-negative learning framework, facilitating the utilization of non-target classes in both labeled and unlabeled videos. Experimental results on THUMOS14 and ActivityNet v1.3 demonstrate the superiority of the proposed method over prior state-of-the-art approaches.

Canhui Tang, Sanping Zhou, Yizhe Li et al.

With the wide application of knowledge distillation between an ImageNet pre-trained teacher model and a learnable student model, industrial anomaly detection has witnessed a significant achievement in the past few years. The success of knowledge distillation mainly relies on how to keep the feature discrepancy between the teacher and student model, in which it assumes that: (1) the teacher model can jointly represent two different distributions for the normal and abnormal patterns, while (2) the student model can only reconstruct the normal distribution. However, it still remains a challenging issue to maintain these ideal assumptions in practice. In this paper, we propose a simple yet effective two-stage industrial anomaly detection framework, termed as AAND, which sequentially performs Anomaly Amplification and Normality Distillation to obtain robust feature discrepancy. In the first anomaly amplification stage, we propose a novel Residual Anomaly Amplification (RAA) module to advance the pre-trained teacher encoder. With the exposure of synthetic anomalies, it amplifies anomalies via residual generation while maintaining the integrity of pre-trained model. It mainly comprises a Matching-guided Residual Gate and an Attribute-scaling Residual Generator, which can determine the residuals' proportion and characteristic, respectively. In the second normality distillation stage, we further employ a reverse distillation paradigm to train a student decoder, in which a novel Hard Knowledge Distillation (HKD) loss is built to better facilitate the reconstruction of normal patterns. Comprehensive experiments on the MvTecAD, VisA, and MvTec3D-RGB datasets show that our method achieves state-of-the-art performance.

Zheng Qin, Yabing Wang, Minghui Yang et al.

Generating 3D human motions from text is a challenging yet valuable task. The key aspects of this task are ensuring text-motion consistency and achieving generation diversity. Although recent advancements have enabled the generation of precise and high-quality human motions from text, achieving diversity in the generated motions remains a significant challenge. In this paper, we aim to overcome the above challenge by designing a simple yet effective text-to-motion generation method, \textit{i.e.}, Diverse-T2M. Our method introduces uncertainty into the generation process, enabling the generation of highly diverse motions while preserving the semantic consistency of the text. Specifically, we propose a novel perspective that utilizes noise signals as carriers of diversity information in transformer-based methods, facilitating a explicit modeling of uncertainty. Moreover, we construct a latent space where text is projected into a continuous representation, instead of a rigid one-to-one mapping, and integrate a latent space sampler to introduce stochastic sampling into the generation process, thereby enhancing the diversity and uncertainty of the outputs. Our results on text-to-motion generation benchmark datasets~(HumanML3D and KIT-ML) demonstrate that our method significantly enhances diversity while maintaining state-of-the-art performance in text consistency.

Zheng Qin, Ruobing Zheng, Yabing Wang et al.

While Multimodal Large Language Models (MLLMs) show immense promise for achieving truly human-like interactions, progress is hindered by the lack of fine-grained evaluation frameworks for human-centered scenarios, encompassing both the understanding of complex human intentions and the provision of empathetic, context-aware responses. Here we introduce HumanSense, a comprehensive benchmark designed to evaluate the human-centered perception and interaction capabilities of MLLMs, with a particular focus on deep understanding of extended multimodal contexts and the formulation of rational feedback. Our evaluation reveals that leading MLLMs still have considerable room for improvement, particularly for advanced interaction-oriented tasks. Supplementing visual input with audio and text information yields substantial improvements, and Omni-modal models show advantages on these tasks.Furthermore, grounded in the observation that appropriate feedback stems from a contextual analysis of the interlocutor's needs and emotions, we posit that reasoning ability serves as the key to unlocking it. We devise a multi-stage, modality-progressive reinforcement learning approach, resulting in HumanSense-Omni-Reasoning, which substantially enhances performance on higher-level understanding and interactive tasks. Additionally, we observe that successful reasoning processes appear to exhibit consistent thought patterns. By designing corresponding prompts, we also enhance the performance of non-reasoning models in a training-free manner.Project page: \textcolor{brightpink}{https://digital-avatar.github.io/ai/HumanSense/}

Le Wang, Jun Wang, Chunyu Qiang et al.

We present AudioGen-Omni - a unified approach based on multimodal diffusion transformers (MMDit), capable of generating high-fidelity audio, speech, and song coherently synchronized with the input video. AudioGen-Omni introduces a novel joint training paradigm that seamlessly integrates large-scale video-text-audio corpora, enabling a model capable of generating semantically rich, acoustically diverse audio conditioned on multimodal inputs and adaptable to a wide range of audio generation tasks. AudioGen-Omni employs a unified lyrics-transcription encoder that encodes graphemes and phonemes from both song and spoken inputs into dense frame-level representations. Dense frame-level representations are fused using an AdaLN-based joint attention mechanism enhanced with phase-aligned anisotropic positional infusion (PAAPI), wherein RoPE is selectively applied to temporally structured modalities to ensure precise and robust cross-modal alignment. By unfreezing all modalities and masking missing inputs, AudioGen-Omni mitigates the semantic constraints of text-frozen paradigms, enabling effective cross-modal conditioning. This joint training approach enhances audio quality, semantic alignment, and lip-sync accuracy, while also achieving state-of-the-art results on Text-to-Audio/Speech/Song tasks. With an inference time of 1.91 seconds for 8 seconds of audio, it offers substantial improvements in both efficiency and generality.

Haoang Lu, Yuanqi Su, Xiaoning Zhang et al.

This paper introduces VisHall3D, a novel two-stage framework for monocular semantic scene completion that aims to address the issues of feature entanglement and geometric inconsistency prevalent in existing methods. VisHall3D decomposes the scene completion task into two stages: reconstructing the visible regions (vision) and inferring the invisible regions (hallucination). In the first stage, VisFrontierNet, a visibility-aware projection module, is introduced to accurately trace the visual frontier while preserving fine-grained details. In the second stage, OcclusionMAE, a hallucination network, is employed to generate plausible geometries for the invisible regions using a noise injection mechanism. By decoupling scene completion into these two distinct stages, VisHall3D effectively mitigates feature entanglement and geometric inconsistency, leading to significantly improved reconstruction quality. The effectiveness of VisHall3D is validated through extensive experiments on two challenging benchmarks: SemanticKITTI and SSCBench-KITTI-360. VisHall3D achieves state-of-the-art performance, outperforming previous methods by a significant margin and paves the way for more accurate and reliable scene understanding in autonomous driving and other applications.

Yuxuan Wu, Le Wang, Sanping Zhou et al.

Controllable layout generation aims to create plausible visual arrangements of element bounding boxes within a graphic design according to certain optional constraints, such as the type or position of a specific component. While recent diffusion or flow-matching models have achieved considerable advances in multifarious conditional generation tasks, there remains considerable room for generating optimal arrangements under given conditions. In this work, we propose to carry out layout generation through retrieving by conditions and reference-guided generation. Specifically, we retrieve appropriate layout templates according to given conditions as references. The references are then utilized to guide the denoising or flow-based transport process. By retrieving layouts compatible with the given conditions, we can uncover the potential information not explicitly provided in the given condition. Such an approach offers more effective guidance to the model during the generation process, in contrast to previous models that feed the condition to the model and let the model infer the unprovided layout attributes directly. Meanwhile, we design a condition-modulated attention that selectively absorbs retrieval knowledge, adapting to the difference between retrieved templates and given conditions. Extensive experiment results show that our method successfully produces high-quality layouts that meet the given conditions and outperforms existing state-of-the-art models. Code will be released upon acceptance.

Xiaolong Sun, Le Wang, Sanping Zhou et al.

Video temporal grounding is a critical video understanding task, which aims to localize moments relevant to a language description. The challenge of this task lies in distinguishing relevant and irrelevant moments. Previous methods focused on learning continuous features exhibit weak differentiation between foreground and background features. In this paper, we propose a novel Moment-Quantization based Video Temporal Grounding method (MQVTG), which quantizes the input video into various discrete vectors to enhance the discrimination between relevant and irrelevant moments. Specifically, MQVTG maintains a learnable moment codebook, where each video moment matches a codeword. Considering the visual diversity, i.e., various visual expressions for the same moment, MQVTG treats moment-codeword matching as a clustering process without using discrete vectors, avoiding the loss of useful information from direct hard quantization. Additionally, we employ effective prior-initialization and joint-projection strategies to enhance the maintained moment codebook. With its simple implementation, the proposed method can be integrated into existing temporal grounding models as a plug-and-play component. Extensive experiments on six popular benchmarks demonstrate the effectiveness and generalizability of MQVTG, significantly outperforming state-of-the-art methods. Further qualitative analysis shows that our method effectively groups relevant features and separates irrelevant ones, aligning with our goal of enhancing discrimination.

Jingyi Tian, Le Wang, Sanping Zhou et al.

Learning generalizable robotic manipulation policies remains a key challenge due to the scarcity of diverse real-world training data. While recent approaches have attempted to mitigate this through self-supervised representation learning, most either rely on 2D vision pretraining paradigms such as masked image modeling, which primarily focus on static semantics or scene geometry, or utilize large-scale video prediction models that emphasize 2D dynamics, thus failing to jointly learn the geometry, semantics, and dynamics required for effective manipulation. In this paper, we present DynaRend, a representation learning framework that learns 3D-aware and dynamics-informed triplane features via masked reconstruction and future prediction using differentiable volumetric rendering. By pretraining on multi-view RGB-D video data, DynaRend jointly captures spatial geometry, future dynamics, and task semantics in a unified triplane representation. The learned representations can be effectively transferred to downstream robotic manipulation tasks via action value map prediction. We evaluate DynaRend on two challenging benchmarks, RLBench and Colosseum, as well as in real-world robotic experiments, demonstrating substantial improvements in policy success rate, generalization to environmental perturbations, and real-world applicability across diverse manipulation tasks.

Sen Wang, Jingyi Tian, Le Wang et al.

World models allow agents to simulate the consequences of actions in imagined environments for planning, control, and long-horizon decision-making. However, existing autoregressive world models struggle with visually coherent predictions due to disrupted spatial structure, inefficient decoding, and inadequate motion modeling. In response, we propose \textbf{S}cale-wise \textbf{A}utoregression with \textbf{M}otion \textbf{P}r\textbf{O}mpt (\textbf{SAMPO}), a hybrid framework that combines visual autoregressive modeling for intra-frame generation with causal modeling for next-frame generation. Specifically, SAMPO integrates temporal causal decoding with bidirectional spatial attention, which preserves spatial locality and supports parallel decoding within each scale. This design significantly enhances both temporal consistency and rollout efficiency. To further improve dynamic scene understanding, we devise an asymmetric multi-scale tokenizer that preserves spatial details in observed frames and extracts compact dynamic representations for future frames, optimizing both memory usage and model performance. Additionally, we introduce a trajectory-aware motion prompt module that injects spatiotemporal cues about object and robot trajectories, focusing attention on dynamic regions and improving temporal consistency and physical realism. Extensive experiments show that SAMPO achieves competitive performance in action-conditioned video prediction and model-based control, improving generation quality with 4.4$\times$ faster inference. We also evaluate SAMPO's zero-shot generalization and scaling behavior, demonstrating its ability to generalize to unseen tasks and benefit from larger model sizes.

Canhui Tang, Sanping Zhou, Haoyue Shi et al.

Zero-Shot Video Anomaly Detection (ZS-VAD) requires temporally localizing anomalies without target domain training data, which is a crucial task due to various practical concerns, e.g., data privacy or new surveillance deployments. Skeleton-based approach has inherent generalizable advantages in achieving ZS-VAD as it eliminates domain disparities both in background and human appearance. However, existing methods only learn low-level skeleton representation and rely on the domain-limited normality boundary, which cannot generalize well to new scenes with different normal and abnormal behavior patterns. In this paper, we propose a novel zero-shot video anomaly detection framework, unlocking the potential of skeleton data via action typicality and uniqueness learning. Firstly, we introduce a language-guided semantic typicality modeling module that projects skeleton snippets into action semantic space and distills LLM's knowledge of typical normal and abnormal behaviors during training. Secondly, we propose a test-time context uniqueness analysis module to finely analyze the spatio-temporal differences between skeleton snippets and then derive scene-adaptive boundaries. Without using any training samples from the target domain, our method achieves state-of-the-art results against skeleton-based methods on four large-scale VAD datasets: ShanghaiTech, UBnormal, NWPU, and UCF-Crime, featuring over 100 unseen surveillance scenes.

Mang Cao, Sanping Zhou, Yizhe Li et al.

Sufficient cross-task interaction is crucial for success in multi-task dense prediction. However, sufficient interaction often results in high computational complexity, forcing existing methods to face the trade-off between interaction completeness and computational efficiency. To address this limitation, this work proposes a Bidirectional Interaction Mamba (BIM), which incorporates novel scanning mechanisms to adapt the Mamba modeling approach for multi-task dense prediction. On the one hand, we introduce a novel Bidirectional Interaction Scan (BI-Scan) mechanism, which constructs task-specific representations as bidirectional sequences during interaction. By integrating task-first and position-first scanning modes within a unified linear complexity architecture, BI-Scan efficiently preserves critical cross-task information. On the other hand, we employ a Multi-Scale Scan~(MS-Scan) mechanism to achieve multi-granularity scene modeling. This design not only meets the diverse granularity requirements of various tasks but also enhances nuanced cross-task feature interactions. Extensive experiments on two challenging benchmarks, \emph{i.e.}, NYUD-V2 and PASCAL-Context, show the superiority of our BIM vs its state-of-the-art competitors.

Jun Hou, Le Wang, Xuan Wang

Mixture-of-Experts (MoE) models have become increasingly powerful in multimodal learning by enabling modular specialization across modalities. However, their effectiveness remains unclear when additional modalities introduce more noise than complementary information. Existing approaches, such as the Partial Information Decomposition, struggle to scale beyond two modalities and lack the resolution needed for instance-level control. We propose Beyond Two-modality Weighting (BTW), a bi-level, non-parametric weighting framework that combines instance-level Kullback-Leibler (KL) divergence and modality-level mutual information (MI) to dynamically adjust modality importance during training. Our method does not require additional parameters and can be applied to an arbitrary number of modalities. Specifically, BTW computes per-example KL weights by measuring the divergence between each unimodal and the current multimodal prediction, and modality-wide MI weights by estimating global alignment between unimodal and multimodal outputs. Extensive experiments on sentiment regression and clinical classification demonstrate that our method significantly improves regression performance and multiclass classification accuracy.

Bing Liu, Le Wang, Mingming Liu et al.

Existing dehazing methods deal with real-world haze images with difficulty, especially scenes with thick haze. One of the main reasons is the lack of real-world paired data and robust priors. To avoid the costly collection of paired hazy and clear images, we propose an efficient semi-supervised image dehazing method via Expectation-Maximization and Bidirectional Brownian Bridge Diffusion Models (EM-B3DM) with a two-stage learning scheme. In the first stage, we employ the EM algorithm to decouple the joint distribution of paired hazy and clear images into two conditional distributions, which are then modeled using a unified Brownian Bridge diffusion model to directly capture the structural and content-related correlations between hazy and clear images. In the second stage, we leverage the pre-trained model and large-scale unpaired hazy and clear images to further improve the performance of image dehazing. Additionally, we introduce a detail-enhanced Residual Difference Convolution block (RDC) to capture gradient-level information, significantly enhancing the model's representation capability. Extensive experiments demonstrate that our EM-B3DM achieves superior or at least comparable performance to state-of-the-art methods on both synthetic and real-world datasets.