Xiao-Ping Zhang

Ting Yao, Yehao Li, Yingwei Pan et al.

Prior works have proposed several strategies to reduce the computational cost of self-attention mechanism. Many of these works consider decomposing the self-attention procedure into regional and local feature extraction procedures that each incurs a much smaller computational complexity. However, regional information is typically only achieved at the expense of undesirable information lost owing to down-sampling. In this paper, we propose a novel Transformer architecture that aims to mitigate the cost issue, named Dual Vision Transformer (Dual-ViT). The new architecture incorporates a critical semantic pathway that can more efficiently compress token vectors into global semantics with reduced order of complexity. Such compressed global semantics then serve as useful prior information in learning finer pixel level details, through another constructed pixel pathway. The semantic pathway and pixel pathway are then integrated together and are jointly trained, spreading the enhanced self-attention information in parallel through both of the pathways. Dual-ViT is henceforth able to reduce the computational complexity without compromising much accuracy. We empirically demonstrate that Dual-ViT provides superior accuracy than SOTA Transformer architectures with reduced training complexity. Source code is available at \url{https://github.com/YehLi/ImageNetModel}.

Haoran Zhou, Honghua Chen, Yingkui Zhang et al.

Point normal, as an intrinsic geometric property of 3D objects, not only serves conventional geometric tasks such as surface consolidation and reconstruction, but also facilitates cutting-edge learning-based techniques for shape analysis and generation. In this paper, we propose a normal refinement network, called Refine-Net, to predict accurate normals for noisy point clouds. Traditional normal estimation wisdom heavily depends on priors such as surface shapes or noise distributions, while learning-based solutions settle for single types of hand-crafted features. Differently, our network is designed to refine the initial normal of each point by extracting additional information from multiple feature representations. To this end, several feature modules are developed and incorporated into Refine-Net by a novel connection module. Besides the overall network architecture of Refine-Net, we propose a new multi-scale fitting patch selection scheme for the initial normal estimation, by absorbing geometry domain knowledge. Also, Refine-Net is a generic normal estimation framework: 1) point normals obtained from other methods can be further refined, and 2) any feature module related to the surface geometric structures can be potentially integrated into the framework. Qualitative and quantitative evaluations demonstrate the clear superiority of Refine-Net over the state-of-the-arts on both synthetic and real-scanned datasets. Our code is available at https://github.com/hrzhou2/refinenet.

Yiyang Shen, Mingqiang Wei, Sen Deng et al.

The intricacy of rainy image contents often leads cutting-edge deraining models to image degradation including remnant rain, wrongly-removed details, and distorted appearance. Such degradation is further exacerbated when applying the models trained on synthetic data to real-world rainy images. We observe two types of domain gaps between synthetic and real-world rainy images: one exists in rain streak patterns; the other is the pixel-level appearance of rain-free images. To bridge the two domain gaps, we propose a semi-supervised detail-recovery image deraining network (Semi-DRDNet) with dual sample-augmented contrastive learning. Semi-DRDNet consists of three sub-networks:i) for removing rain streaks without remnants, we present a squeeze-and-excitation based rain residual network; ii) for encouraging the lost details to return, we construct a structure detail context aggregation based detail repair network; to our knowledge, this is the first time; and iii) for building efficient contrastive constraints for both rain streaks and clean backgrounds, we exploit a novel dual sample-augmented contrastive regularization network.Semi-DRDNet operates smoothly on both synthetic and real-world rainy data in terms of deraining robustness and detail accuracy. Comparisons on four datasets including our established Real200 show clear improvements of Semi-DRDNet over fifteen state-of-the-art methods. Code and dataset are available at https://github.com/syy-whu/DRD-Net.

Yuzhu Mao, Zihao Zhao, Meilin Yang et al.

Federated learning (FL) enables edge devices to collaboratively learn a model in a distributed fashion. Many existing researches have focused on improving communication efficiency of high-dimensional models and addressing bias caused by local updates. However, most of FL algorithms are either based on reliable communications or assume fixed and known unreliability characteristics. In practice, networks could suffer from dynamic channel conditions and non-deterministic disruptions, with time-varying and unknown characteristics. To this end, in this paper we propose a sparsity enabled FL framework with both communication efficiency and bias reduction, termed as SAFARI. It makes novel use of a similarity among client models to rectify and compensate for bias that is resulted from unreliable communications. More precisely, sparse learning is implemented on local clients to mitigate communication overhead, while to cope with unreliable communications, a similarity-based compensation method is proposed to provide surrogates for missing model updates. We analyze SAFARI under bounded dissimilarity and with respect to sparse models. It is demonstrated that SAFARI under unreliable communications is guaranteed to converge at the same rate as the standard FedAvg with perfect communications. Implementations and evaluations on CIFAR-10 dataset validate the effectiveness of SAFARI by showing that it can achieve the same convergence speed and accuracy as FedAvg with perfect communications, with up to 80% of the model weights being pruned and a high percentage of client updates missing in each round.

Zihao Zhao, Yuzhu Mao, Zhenpeng Shi et al.

The widespread adoption of Federated Learning (FL), a privacy-preserving distributed learning methodology, has been impeded by the challenge of high communication overheads, typically arising from the transmission of large-scale models. Existing adaptive quantization methods, designed to mitigate these overheads, operate under the impractical assumption of uniform device participation in every training round. Additionally, these methods are limited in their adaptability due to the necessity of manual quantization level selection and often overlook biases inherent in local devices' data, thereby affecting the robustness of the global model. In response, this paper introduces AQUILA (adaptive quantization in device selection strategy), a novel adaptive framework devised to effectively handle these issues, enhancing the efficiency and robustness of FL. AQUILA integrates a sophisticated device selection method that prioritizes the quality and usefulness of device updates. Utilizing the exact global model stored by devices, it enables a more precise device selection criterion, reduces model deviation, and limits the need for hyperparameter adjustments. Furthermore, AQUILA presents an innovative quantization criterion, optimized to improve communication efficiency while assuring model convergence. Our experiments demonstrate that AQUILA significantly decreases communication costs compared to existing methods, while maintaining comparable model performance across diverse non-homogeneous FL settings, such as Non-IID data and heterogeneous model architectures.

Yixuan Gao, Xiongkuo Min, Wenhan Zhu et al.

Recently, image quality has been generally describedby a mean opinion score (MOS). However, we observe that thequality scores of an image given by a group of subjects are verysubjective and diverse. Thus it is not enough to use a MOS todescribe the image quality. In this paper, we propose to describeimage quality using a parameterized distribution rather thana MOS, and an objective method is also proposed to predictthe image quality score distribution (IQSD). At first, the LIVEdatabase is re-recorded. Specifically, we have invited a largegroup of subjects to evaluate the quality of all images in theLIVE database, and each image is evaluated by a large numberof subjects (187 valid subjects), whose scores can form a reliableIQSD. By analyzing the obtained subjective quality scores, wefind that the IQSD can be well modeled by an alpha stable model,and it can reflect much more information than a single MOS, suchas the skewness of opinion score, the subject diversity and themaximum probability score for an image. Therefore, we proposeto model the IQSD using the alpha stable model. Moreover, wepropose a framework and an algorithm to predict the alphastable model based IQSD, where quality features are extractedfrom each image based on structural information and statisticalinformation, and support vector regressors are trained to predictthe alpha stable model parameters. Experimental results verifythe feasibility of using alpha stable model to describe the IQSD,and prove the effectiveness of objective alpha stable model basedIQSD prediction method.

Zihao Zhao, Yang Liu, Wenbo Ding et al.

Existing research has either adapted the Probably Approximately Correct (PAC) Bayesian framework for federated learning (FL) or used information-theoretic PAC-Bayesian bounds while introducing their theorems, but few considering the non-IID challenges in FL. Our work presents the first non-vacuous federated PAC-Bayesian bound tailored for non-IID local data. This bound assumes unique prior knowledge for each client and variable aggregation weights. We also introduce an objective function and an innovative Gibbs-based algorithm for the optimization of the derived bound. The results are validated on real-world datasets.

Guangyi Chen, Xiao Liu, Guangrun Wang et al.

Video-language pre-trained models have shown remarkable success in guiding video question-answering (VideoQA) tasks. However, due to the length of video sequences, training large-scale video-based models incurs considerably higher costs than training image-based ones. This motivates us to leverage the knowledge from image-based pretraining, despite the obvious gaps between image and video domains. To bridge these gaps, in this paper, we propose Tem-Adapter, which enables the learning of temporal dynamics and complex semantics by a visual Temporal Aligner and a textual Semantic Aligner. Unlike conventional pretrained knowledge adaptation methods that only concentrate on the downstream task objective, the Temporal Aligner introduces an extra language-guided autoregressive task aimed at facilitating the learning of temporal dependencies, with the objective of predicting future states based on historical clues and language guidance that describes event progression. Besides, to reduce the semantic gap and adapt the textual representation for better event description, we introduce a Semantic Aligner that first designs a template to fuse question and answer pairs as event descriptions and then learns a Transformer decoder with the whole video sequence as guidance for refinement. We evaluate Tem-Adapter and different pre-train transferring methods on two VideoQA benchmarks, and the significant performance improvement demonstrates the effectiveness of our method.

Shoujie Li, Haixin Yu, Wenbo Ding et al.

The accurate detection and grasping of transparent objects are challenging but of significance to robots. Here, a visual-tactile fusion framework for transparent object grasping under complex backgrounds and variant light conditions is proposed, including the grasping position detection, tactile calibration, and visual-tactile fusion based classification. First, a multi-scene synthetic grasping dataset generation method with a Gaussian distribution based data annotation is proposed. Besides, a novel grasping network named TGCNN is proposed for grasping position detection, showing good results in both synthetic and real scenes. In tactile calibration, inspired by human grasping, a fully convolutional network based tactile feature extraction method and a central location based adaptive grasping strategy are designed, improving the success rate by 36.7% compared to direct grasping. Furthermore, a visual-tactile fusion method is proposed for transparent objects classification, which improves the classification accuracy by 34%. The proposed framework synergizes the advantages of vision and touch, and greatly improves the grasping efficiency of transparent objects.

Hengxi Zhang, Zhendong Shi, Yuanquan Hu et al.

Quantitative markets are characterized by swift dynamics and abundant uncertainties, making the pursuit of profit-driven stock trading actions inherently challenging. Within this context, reinforcement learning (RL), which operates on a reward-centric mechanism for optimal control, has surfaced as a potentially effective solution to the intricate financial decision-making conundrums presented. This paper delves into the fusion of two established financial trading strategies, namely the constant proportion portfolio insurance (CPPI) and the time-invariant portfolio protection (TIPP), with the multi-agent deep deterministic policy gradient (MADDPG) framework. As a result, we introduce two novel multi-agent RL (MARL) methods, CPPI-MADDPG and TIPP-MADDPG, tailored for probing strategic trading within quantitative markets. To validate these innovations, we implemented them on a diverse selection of 100 real-market shares. Our empirical findings reveal that the CPPI-MADDPG and TIPP-MADDPG strategies consistently outpace their traditional counterparts, affirming their efficacy in the realm of quantitative trading.

Yang Tan, Enming Zhang, Yang Li et al.

We propose two novel transferability metrics F-OTCE (Fast Optimal Transport based Conditional Entropy) and JC-OTCE (Joint Correspondence OTCE) to evaluate how much the source model (task) can benefit the learning of the target task and to learn more transferable representations for cross-domain cross-task transfer learning. Unlike the existing metric that requires evaluating the empirical transferability on auxiliary tasks, our metrics are auxiliary-free such that they can be computed much more efficiently. Specifically, F-OTCE estimates transferability by first solving an Optimal Transport (OT) problem between source and target distributions, and then uses the optimal coupling to compute the Negative Conditional Entropy between source and target labels. It can also serve as a loss function to maximize the transferability of the source model before finetuning on the target task. Meanwhile, JC-OTCE improves the transferability robustness of F-OTCE by including label distances in the OT problem, though it may incur additional computation cost. Extensive experiments demonstrate that F-OTCE and JC-OTCE outperform state-of-the-art auxiliary-free metrics by 18.85% and 28.88%, respectively in correlation coefficient with the ground-truth transfer accuracy. By eliminating the training cost of auxiliary tasks, the two metrics reduces the total computation time of the previous method from 43 minutes to 9.32s and 10.78s, respectively, for a pair of tasks. When used as a loss function, F-OTCE shows consistent improvements on the transfer accuracy of the source model in few-shot classification experiments, with up to 4.41% accuracy gain.

Chengyu Zheng, Peng Li, Xiao-Ping Zhang et al.

Recall one time when we were in an unfamiliar mall. We might mistakenly think that there exists or does not exist a piece of glass in front of us. Such mistakes will remind us to walk more safely and freely at the same or a similar place next time. To absorb the human mistake correction wisdom, we propose a novel glass segmentation network to detect transparent glass, dubbed GlassSegNet. Motivated by this human behavior, GlassSegNet utilizes two key stages: the identification stage (IS) and the correction stage (CS). The IS is designed to simulate the detection procedure of human recognition for identifying transparent glass by global context and edge information. The CS then progressively refines the coarse prediction by correcting mistake regions based on gained experience. Extensive experiments show clear improvements of our GlassSegNet over thirty-four state-of-the-art methods on three benchmark datasets.

Ruiwen Gu, Qitai Tan, Yahao Liu et al.

Accurate traffic forecasting is essential for intelligent transportation systems, supporting a wide range of real-world applications. However, it remains challenging due to two key factors:~(1) Traffic series contain heterogeneous temporal patterns, where stable periodic regularities coexist with event-driven fluctuations. Existing methods often treat them within a unified representation, limiting their ability to capture fine-grained temporal dynamics.~(2)Spatial dependencies among nodes are inherently dynamic and sparse, while dense all-pairs attention often introduces redundant interactions and amplifies noise. To address these issues, we propose ADMFormer, an Adaptive-Decomposition Transformer with Time-Varying Masked Spatial Attention. Specifically, ADMFormer first employs a time-node adaptive gating mechanism to decouple traffic signals into dominant regularities and residual fluctuations that vary across time and nodes. A dual-branch temporal module is then designed to separately capture global periodic dependencies and high-frequency irregular variations from these two decomposed components. Furthermore, ADMFormer introduces a time-varying masked spatial attention that sparsifies spatial interactions based on real-time traffic states, thereby effectively preserving dynamic and informative dependencies. Extensive experiments on four real-world datasets demonstrate that ADMFormer achieves state-of-the-art performance.

Ziyi Ding, Xiao-Ping Zhang

We find a mismatch between what large language models encode about a causal question and what they answer. On anti-commonsense CLadder items, a fixed linear probe recovers the evidence-supported answer from the model's hidden state (accuracy approximately 0.97), while the spoken Yes/No reverts to the commonsense one (accuracy approximately 0.5). We call this approximately +0.5 gap Causal Tongue-Tie: a wrong Yes/No decomposes into two separable failure modes: no internal signal versus a signal the verbal interface cannot say. The implication cuts both ways for output-only causal benchmarks: a benchmark "correct" need not mean the model has understood, and a benchmark "wrong" need not mean it cannot. Sweeping claims about whether LLMs can do causal reasoning, drawn from a single accuracy number, deserve a second look.

Yinghao Wu, Zhuoyan Luo, Yiyao Yu et al.

Despite the remarkable progress achieved by recent efficient methods in accelerating multimodal understanding, they still suffer from noticeable performance degradation. Their emphasis on the high compression ratio of a single visual clue and reliance on the heuristic pruning strategy with coarse attention alignment incurs a bottleneck on the information capacity and density of visual tokens. Addressing this limitation, we propose VEN-VL, a visual ensemble MoE framework for effective and efficient perception following the enrich then compact principle. Specifically, we first enrich the information capacity by unifying the visual representations of different perspectives, and then progressively compact it with adaptive routers in specialized visual experts to enhance the information density. Furthermore, we incorporate the reconstruction ability of vanilla structure via explicit visual supervision, facilitating crucial information preservation. Experimental results demonstrate our superiority in complex visual tasks with few information-condensed tokens, which effectively bridges the gap between performance and efficiency.

Ruiwen Gu, Yahao Liu, Zhenyu Liu et al.

As a core task in intelligent transportation systems, traffic forecasting plays a critical role in urban traffic management. Accurate traffic forecasting relies on modeling complex spatiotemporal dependencies, which is inherently challenging due to spatial heterogeneity in traffic systems.Despite significant progress, most existing methods are still limited to pairwise spatial dependency modeling, making it difficult to capture dynamic high-order interactions among nodes with similar traffic patterns. To address this issue, we propose PHGNet, a novel spatiotemporal forecasting framework based on prototype-guided hypergraph construction. At the core of PHGNet, a prototype learning mechanism is designed to adaptively assign pattern-similar nodes to hyperedges, thereby capturing high-order interactions with time-varying structures. To improve the reliability of dynamic hypergraph construction, we further develop a global-local node representation module to extract time-consistent features. For forecasting, iterative residual refinement and Temporal Query Attention are introduced to improve forecasting accuracy while supporting efficient parallel decoding. Extensive experiments on multiple real-world datasets demonstrate that PHGNet achieves superior predictive performance compared with state-of-the-art methods.

Xiao Liu, Guangyi Chen, Yansong Tang et al.

Composing simple elements into complex concepts is crucial yet challenging, especially for 3D action generation. Existing methods largely rely on extensive neural language annotations to discern composable latent semantics, a process that is often costly and labor-intensive. In this study, we introduce a novel framework to generate compositional actions without reliance on language auxiliaries. Our approach consists of three main components: Action Coupling, Conditional Action Generation, and Decoupling Refinement. Action Coupling utilizes an energy model to extract the attention masks of each sub-action, subsequently integrating two actions using these attentions to generate pseudo-training examples. Then, we employ a conditional generative model, CVAE, to learn a latent space, facilitating the diverse generation. Finally, we propose Decoupling Refinement, which leverages a self-supervised pre-trained model MAE to ensure semantic consistency between the sub-actions and compositional actions. This refinement process involves rendering generated 3D actions into 2D space, decoupling these images into two sub-segments, using the MAE model to restore the complete image from sub-segments, and constraining the recovered images to match images rendered from raw sub-actions. Due to the lack of existing datasets containing both sub-actions and compositional actions, we created two new datasets, named HumanAct-C and UESTC-C, and present a corresponding evaluation metric. Both qualitative and quantitative assessments are conducted to show our efficacy.

Yoga Suhas Kuruba Manjunath, Sihao Zhao, Xiao-Ping Zhang et al.

Deep learning-based network traffic classification (NTC) techniques, including conventional and class-of-service (CoS) classifiers, are a popular tool that aids in the quality of service (QoS) and radio resource management for the Internet of Things (IoT) network. Holistic temporal features consist of inter-, intra-, and pseudo-temporal features within packets, between packets, and among flows, providing the maximum information on network services without depending on defined classes in a problem. Conventional spatio-temporal features in the current solutions extract only space and time information between packets and flows, ignoring the information within packets and flow for IoT traffic. Therefore, we propose a new, efficient, holistic feature extraction method for deep-learning-based NTC using time-distributed feature learning to maximize the accuracy of the NTC. We apply a time-distributed wrapper on deep-learning layers to help extract pseudo-temporal features and spatio-temporal features. Pseudo-temporal features are mathematically complex to explain since, in deep learning, a black box extracts them. However, the features are temporal because of the time-distributed wrapper; therefore, we call them pseudo-temporal features. Since our method is efficient in learning holistic-temporal features, we can extend our method to both conventional and CoS NTC. Our solution proves that pseudo-temporal and spatial-temporal features can significantly improve the robustness and performance of any NTC. We analyze the solution theoretically and experimentally on different real-world datasets. The experimental results show that the holistic-temporal time-distributed feature learning method, on average, is 13.5% more accurate than the state-of-the-art conventional and CoS classifiers.

Nan Zhou, Weijie Hong, Huandong Wang et al.

Fine-grained air pollution forecasting is crucial for urban management and the development of healthy buildings. Deploying portable sensors on mobile platforms such as cars and buses offers a low-cost, easy-to-maintain, and wide-coverage data collection solution. However, due to the random and uncontrollable movement patterns of these non-dedicated mobile platforms, the resulting sensor data are often incomplete and temporally inconsistent. By exploring potential training patterns in the reverse process of diffusion models, we propose Spatio-Temporal Physics-Informed Diffusion Models (STeP-Diff). STeP-Diff leverages DeepONet to model the spatial sequence of measurements along with a PDE-informed diffusion model to forecast the spatio-temporal field from incomplete and time-varying data. Through a PDE-constrained regularization framework, the denoising process asymptotically converges to the convection-diffusion dynamics, ensuring that predictions are both grounded in real-world measurements and aligned with the fundamental physics governing pollution dispersion. To assess the performance of the system, we deployed 59 self-designed portable sensing devices in two cities, operating for 14 days to collect air pollution data. Compared to the second-best performing algorithm, our model achieved improvements of up to 89.12% in MAE, 82.30% in RMSE, and 25.00% in MAPE, with extensive evaluations demonstrating that STeP-Diff effectively captures the spatio-temporal dependencies in air pollution fields.

Nan Zhou, Huandong Wang, Jiahao Li et al.

Fine-grained fire prediction plays a crucial role in emergency response. Infrared images and fire masks provide complementary thermal and boundary information, yet current methods are predominantly limited to binary mask modeling with inherent signal sparsity, failing to capture the complex dynamics of fire. While world models show promise in video generation, their physical inconsistencies pose significant challenges for fire forecasting. This paper introduces PhysFire-WM, a Physics-informed World Model for emulating Fire spread dynamics. Our approach internalizes combustion dynamics by encoding structured priors from a Physical Simulator to rectify physical discrepancies, coupled with a Cross-task Collaborative Training strategy (CC-Train) that alleviates the issue of limited information in mask-based modeling. Through parameter sharing and gradient coordination, CC-Train effectively integrates thermal radiation dynamics and spatial boundary delineation, enhancing both physical realism and geometric accuracy. Extensive experiments on a fine-grained multimodal fire dataset demonstrate the superior accuracy of PhysFire-WM in fire spread prediction. Validation underscores the importance of physical priors and cross-task collaboration, providing new insights for applying physics-informed world models to disaster prediction.

Zhaojian Yu, Kaiyue Feng, Yilun Zhao et al.

Large language models have made significant progress in complex but easy-to-verify problems, yet they still struggle with discovering the unknown. In this paper, we present \textbf{AlphaResearch}, an autonomous research agent designed to discover new algorithms on open-ended problems. To synergize the feasibility and innovation of the discovery process, we construct a novel dual research environment by combining the execution-based verify and simulated real-world peer review environment. AlphaResearch discovers new algorithm by iteratively running the following steps: (1) propose new ideas (2) verify the ideas in the dual research environment (3) optimize the research proposals for better performance. To promote a transparent evaluation process, we construct \textbf{AlphaResearchComp}, a new evaluation benchmark that includes an eight open-ended algorithmic problems competition, with each problem carefully curated and verified through executable pipelines, objective metrics, and reproducibility checks. AlphaResearch gets a 2/8 win rate in head-to-head comparison with human researchers, demonstrate the possibility of accelerating algorithm discovery with LLMs. Notably, the algorithm discovered by AlphaResearch on the \emph{``packing circles''} problem achieves the best-of-known performance, surpassing the results of human researchers and strong baselines from recent work (e.g., AlphaEvolve). Additionally, we conduct a comprehensive analysis of the remaining challenges of the 6/8 failure cases, providing valuable insights for future research.

Xinchun Yu, Shuangqing Wei, Shao-Lun Huang et al.

This paper investigates the asymptotics of the maximal throughput of communication over AWGN channels by $n$ channel uses under a covert constraint in terms of an upper bound $δ$ of Kullback-Leibler divergence (KL divergence). It is shown that the first and second order asymptotics of the maximal throughput are $\sqrt{nδ\log e}$ and $(2)^{1/2}(nδ)^{1/4}(\log e)^{3/4}\cdot Q^{-1}(ε)$, respectively. The technique we use in the achievability is quasi-$\varepsilon$-neighborhood notion from information geometry. For finite blocklength $n$, the generating distributions are chosen to be a family of truncated Gaussian distributions with decreasing variances. The law of decreasing is carefully designed so that it maximizes the throughput at the main channel in the asymptotic sense under the condition that the output distributions satisfy the covert constraint. For the converse, the optimality of Gaussian distribution for minimizing KL divergence under the second order moment constraint is extended from dimension $1$ to dimension $n$. Based on that, we establish an upper bound on the average power of the code to satisfy the covert constraint, which further leads to the direct converse bound in terms of covert metric.

Shiwei Wang, Yongzhen Wang, Bingwen Hu et al.

While Transformer-based architectures have dominated recent advances in all-in-one image restoration, they remain fundamentally reactive: propagating degradations rather than proactively suppressing them. In the absence of explicit suppression mechanisms, degraded signals interfere with feature learning, compelling the decoder to balance artifact removal and detail preservation, thereby increasing model complexity and limiting adaptability. To address these challenges, we propose M2IR, a novel restoration framework that proactively regulates degradation propagation during the encoding stage and efficiently eliminates residual degradations during decoding. Specifically, the Mamba-Style Transformer (MST) block performs pixel-wise selective state modulation to mitigate degradations while preserving structural integrity. In parallel, the Adaptive Degradation Expert Collaboration (ADEC) module utilizes degradation-specific experts guided by a DA-CLIP-driven router and complemented by a shared expert to eliminate residual degradations through targeted and cooperative restoration. By integrating the MST block and ADEC module, M2IR transitions from passive reaction to active degradation control, effectively harnessing learned representations to achieve superior generalization, enhanced adaptability, and refined recovery of fine-grained details across diverse all-in-one image restoration benchmarks. Our source codes are available at https://github.com/Im34v/M2IR.

Hengxi Zhang, Huaze Tang, Wenbo Ding et al.

The Space-Air-Ground Integrated Network (SAGIN), integrating heterogeneous devices including low earth orbit (LEO) satellites, unmanned aerial vehicles (UAVs), and ground users (GUs), holds significant promise for advancing smart city applications. However, resource management of the SAGIN is a challenge requiring urgent study in that inappropriate resource management will cause poor data transmission, and hence affect the services in smart cities. In this paper, we develop a comprehensive SAGIN system that encompasses five distinct communication links and propose an efficient cooperative multi-type multi-agent deep reinforcement learning (CMT-MARL) method to address the resource management issue. The experimental results highlight the efficacy of the proposed CMT-MARL, as evidenced by key performance indicators such as the overall transmission rate and transmission success rate. These results underscore the potential value and feasibility of future implementation of the SAGIN.

Hongyi Chen, Jingtao Ding, Yong Li et al.

Crowd simulation holds crucial applications in various domains, such as urban planning, architectural design, and traffic arrangement. In recent years, physics-informed machine learning methods have achieved state-of-the-art performance in crowd simulation but fail to model the heterogeneity and multi-modality of human movement comprehensively. In this paper, we propose a social physics-informed diffusion model named SPDiff to mitigate the above gap. SPDiff takes both the interactive and historical information of crowds in the current timeframe to reverse the diffusion process, thereby generating the distribution of pedestrian movement in the subsequent timeframe. Inspired by the well-known social physics model, i.e., Social Force, regarding crowd dynamics, we design a crowd interaction module to guide the denoising process and further enhance this module with the equivariant properties of crowd interactions. To mitigate error accumulation in long-term simulations, we propose a multi-frame rollout training algorithm for diffusion modeling. Experiments conducted on two real-world datasets demonstrate the superior performance of SPDiff in terms of macroscopic and microscopic evaluation metrics. Code and appendix are available at https://github.com/tsinghua-fib-lab/SPDiff.

Yingjie Zhou, Zicheng Zhang, Wei Sun et al.

In the realm of media technology, digital humans have gained prominence due to rapid advancements in computer technology. However, the manual modeling and control required for the majority of digital humans pose significant obstacles to efficient development. The speech-driven methods offer a novel avenue for manipulating the mouth shape and expressions of digital humans. Despite the proliferation of driving methods, the quality of many generated talking head (TH) videos remains a concern, impacting user visual experiences. To tackle this issue, this paper introduces the Talking Head Quality Assessment (THQA) database, featuring 800 TH videos generated through 8 diverse speech-driven methods. Extensive experiments affirm the THQA database's richness in character and speech features. Subsequent subjective quality assessment experiments analyze correlations between scoring results and speech-driven methods, ages, and genders. In addition, experimental results show that mainstream image and video quality assessment methods have limitations for the THQA database, underscoring the imperative for further research to enhance TH video quality assessment. The THQA database is publicly accessible at https://github.com/zyj-2000/THQA.

Jiamei Xiong, Xuefeng Yan, Yongzhen Wang et al.

Haze severely degrades the visual quality of remote sensing images and hampers the performance of road extraction, vehicle detection, and traffic flow monitoring. The emerging denoising diffusion probabilistic model (DDPM) exhibits the significant potential for dense haze removal with its strong generation ability. Since remote sensing images contain extensive small-scale texture structures, it is important to effectively restore image details from hazy images. However, current wisdom of DDPM fails to preserve image details and color fidelity well, limiting its dehazing capacity for remote sensing images. In this paper, we propose a novel unified Fourier-aware diffusion model for remote sensing image dehazing, termed RSHazeDiff. From a new perspective, RSHazeDiff explores the conditional DDPM to improve image quality in dense hazy scenarios, and it makes three key contributions. First, RSHazeDiff refines the training phase of diffusion process by performing noise estimation and reconstruction constraints in a coarse-to-fine fashion. Thus, it remedies the unpleasing results caused by the simple noise estimation constraint in DDPM. Second, by taking the frequency information as important prior knowledge during iterative sampling steps, RSHazeDiff can preserve more texture details and color fidelity in dehazed images. Third, we design a global compensated learning module to utilize the Fourier transform to capture the global dependency features of input images, which can effectively mitigate the effects of boundary artifacts when processing fixed-size patches. Experiments on both synthetic and real-world benchmarks validate the favorable performance of RSHazeDiff over state-of-the-art methods. Source code will be released at https://github.com/jm-xiong/RSHazeDiff.

Weichen Zhang, Zile Zhou, Xin Zeng et al.

Spatial reasoning is a fundamental capability of multimodal large language models (MLLMs), yet their performance in open aerial environments remains underexplored. In this work, we present Open3D-VQA, a novel benchmark for evaluating MLLMs' ability to reason about complex spatial relationships from an aerial perspective. The benchmark comprises 73k QA pairs spanning 7 general spatial reasoning tasks, including multiple-choice, true/false, and short-answer formats, and supports both visual and point cloud modalities. The questions are automatically generated from spatial relations extracted from both real-world and simulated aerial scenes. Evaluation on 13 popular MLLMs reveals that: 1) Models are generally better at answering questions about relative spatial relations than absolute distances, 2) 3D LLMs fail to demonstrate significant advantages over 2D LLMs, and 3) Fine-tuning solely on the simulated dataset can significantly improve the model's spatial reasoning performance in real-world scenarios. We release our benchmark, data generation pipeline, and evaluation toolkit to support further research: https://github.com/EmbodiedCity/Open3D-VQA.code.

Jingjing Fan, Yushan Liu, Shoujie Li et al.

General vision-language models increasingly support unified spatiotemporal reasoning over long video streams, yet deploying such capabilities on robots remains constrained by the prohibitive latency of processing long-horizon histories and generating high-dimensional future predictions. To bridge this gap, we present FUTURE-VLA, a unified architecture that reformulates long-horizon control and future forecasting as a monolithic sequence-generation task. Adopting a dual-sided efficiency paradigm, FUTURE-VLA leverages a temporally adaptive compression strategy to maximize spatiotemporal information density, enabling the ingestion of extensive multi-view histories while maintaining constant inference latency. Simultaneously, it performs latent-space autoregression to align actionable dynamics with reviewable visual look-aheads in a single forward pass. These real-time predictive capabilities further enable a prediction-guided Human-In-the-Loop mechanism via interactive execution gating, allowing operators to dynamically validate behaviors based on interpretable future previews. Extensive evaluations demonstrate that FUTURE-VLA establishes new state-of-the-art performance, attaining success rates of 99.2% on LIBERO, 75.4% on RoboTwin, and 78.0% on a real-world Piper platform, all with a $16\times$ extended spatiotemporal window while maintaining the inference latency of a single-frame baseline.

Yicong Li, Yang Tan, Jingyun Yang et al.

Although many studies have successfully applied transfer learning to medical image segmentation, very few of them have investigated the selection strategy when multiple source tasks are available for transfer. In this paper, we propose a prior knowledge guided and transferability based framework to select the best source tasks among a collection of brain image segmentation tasks, to improve the transfer learning performance on the given target task. The framework consists of modality analysis, RoI (region of interest) analysis, and transferability estimation, such that the source task selection can be refined step by step. Specifically, we adapt the state-of-the-art analytical transferability estimation metrics to medical image segmentation tasks and further show that their performance can be significantly boosted by filtering candidate source tasks based on modality and RoI characteristics. Our experiments on brain matter, brain tumor, and white matter hyperintensities segmentation datasets reveal that transferring from different tasks under the same modality is often more successful than transferring from the same task under different modalities. Furthermore, within the same modality, transferring from the source task that has stronger RoI shape similarity with the target task can significantly improve the final transfer performance. And such similarity can be captured using the Structural Similarity index in the label space.

Zhuoyan Luo, Yinghao Wu, Tianheng Cheng et al.

The newly proposed Generalized Referring Expression Segmentation (GRES) amplifies the formulation of classic RES by involving complex multiple/non-target scenarios. Recent approaches address GRES by directly extending the well-adopted RES frameworks with object-existence identification. However, these approaches tend to encode multi-granularity object information into a single representation, which makes it difficult to precisely represent comprehensive objects of different granularity. Moreover, the simple binary object-existence identification across all referent scenarios fails to specify their inherent differences, incurring ambiguity in object understanding. To tackle the above issues, we propose a \textbf{Co}unting-Aware \textbf{H}ierarchical \textbf{D}ecoding framework (CoHD) for GRES. By decoupling the intricate referring semantics into different granularity with a visual-linguistic hierarchy, and dynamic aggregating it with intra- and inter-selection, CoHD boosts multi-granularity comprehension with the reciprocal benefit of the hierarchical nature. Furthermore, we incorporate the counting ability by embodying multiple/single/non-target scenarios into count- and category-level supervision, facilitating comprehensive object perception. Experimental results on gRefCOCO, Ref-ZOM, R-RefCOCO, and RefCOCO benchmarks demonstrate the effectiveness and rationality of CoHD which outperforms state-of-the-art GRES methods by a remarkable margin. Code is available at \href{https://github.com/RobertLuo1/CoHD}{here}.

Nan Zhou, Zuxin Li, Fanhang Man et al.

This paper addresses the challenge of achieving optimal Quality of Information (QoI) in non-dedicated vehicular mobile crowdsensing (NVMCS) systems. The key obstacles are the interrelated issues of sensing coverage, sensing reliability, and the dynamic participation of vehicles. To tackle these, we propose QUIDS, a QUality-informed Incentive-driven multi-agent Dispatching System, which ensures high sensing coverage and reliability under budget constraints. QUIDS introduces a novel metric, Aggregated Sensing Quality (ASQ), to quantitatively capture QoI by integrating both coverage and reliability. We also develop a Mutually Assisted Belief-aware Vehicle Dispatching algorithm that estimates sensing reliability and allocates incentives under uncertainty, further improving ASQ. Evaluation using real-world data from a metropolitan NVMCS deployment shows QUIDS improves ASQ by 38% over non-dispatching scenarios and by 10% over state-of-the-art methods. It also reduces reconstruction map errors by 39-74% across algorithms. By jointly optimizing coverage and reliability via a quality-informed incentive mechanism, QUIDS enables low-cost, high-quality urban monitoring without dedicated infrastructure, applicable to smart-city scenarios like traffic and environmental sensing.

Wenhao Jiang, Duo Li, Menghan Hu et al.

To tackle the issues of catastrophic forgetting and overfitting in few-shot class-incremental learning (FSCIL), previous work has primarily concentrated on preserving the memory of old knowledge during the incremental phase. The role of pre-trained model in shaping the effectiveness of incremental learning is frequently underestimated in these studies. Therefore, to enhance the generalization ability of the pre-trained model, we propose Learning with Prior Knowledge (LwPK) by introducing nearly free prior knowledge from a few unlabeled data of subsequent incremental classes. We cluster unlabeled incremental class samples to produce pseudo-labels, then jointly train these with labeled base class samples, effectively allocating embedding space for both old and new class data. Experimental results indicate that LwPK effectively enhances the model resilience against catastrophic forgetting, with theoretical analysis based on empirical risk minimization and class distance measurement corroborating its operational principles. The source code of LwPK is publicly available at: \url{https://github.com/StevenJ308/LwPK}.

Jie Sun, Heng Liu, Yongzhen Wang et al.

In this paper, we reveal a novel haze-specific wavelet degradation prior observed through wavelet transform analysis, which shows that haze-related information predominantly resides in low-frequency components. Exploiting this insight, we propose a novel dehazing framework, WDMamba, which decomposes the image dehazing task into two sequential stages: low-frequency restoration followed by detail enhancement. This coarse-to-fine strategy enables WDMamba to effectively capture features specific to each stage of the dehazing process, resulting in high-quality restored images. Specifically, in the low-frequency restoration stage, we integrate Mamba blocks to reconstruct global structures with linear complexity, efficiently removing overall haze and producing a coarse restored image. Thereafter, the detail enhancement stage reinstates fine-grained information that may have been overlooked during the previous phase, culminating in the final dehazed output. Furthermore, to enhance detail retention and achieve more natural dehazing, we introduce a self-guided contrastive regularization during network training. By utilizing the coarse restored output as a hard negative example, our model learns more discriminative representations, substantially boosting the overall dehazing performance. Extensive evaluations on public dehazing benchmarks demonstrate that our method surpasses state-of-the-art approaches both qualitatively and quantitatively. Code is available at https://github.com/SunJ000/WDMamba.

Huining Yuan, Zelai Xu, Huaijie Wang et al.

Reinforcement learning from verifiable rewards (RLVR) has improved the reasoning abilities of large language models (LLMs), but most existing approaches rely on sparse outcome-level feedback. This sparsity creates a credit assignment challenge in long-horizon agentic reasoning: a trajectory may fail despite containing many correct intermediate decisions, or succeed despite containing flawed ones. In this work, we study a class of densely-verifiable agentic reasoning problems, where intermediate actions can be objectively checked by symbolic or algorithmic oracles. We propose Verifiable Process Rewards (VPR), a framework that converts such oracles into dense turn-level supervision for reinforcement learning, and instantiate it in three representative settings: search-based verification for dynamic deduction, constraint-based verification for logical reasoning, and posterior-based verification for probabilistic inference. We further provide a theoretical analysis showing that dense verifier-grounded rewards can improve long-horizon credit assignment by providing more localized learning signals, with the benefit depending on the reliability of the verifier. Empirically, VPR outperforms outcome-level reward and rollout-based process reward baselines across controlled environments, and more importantly, transfers to both general and agentic reasoning benchmarks, suggesting that verifiable process supervision can foster general reasoning skills applicable beyond the training environments. Our results indicate that VPR is a promising approach for enhancing LLM agents whenever reliable intermediate verification is available, while also highlighting its dependence on oracle quality and the open challenge of extending VPR to less structured, open-ended environments.

Zixuan Zhou, Xuefei Ning, Ke Hong et al. · tsinghua

Large Language Models (LLMs) have attracted extensive attention due to their remarkable performance across various tasks. However, the substantial computational and memory requirements of LLM inference pose challenges for deployment in resource-constrained scenarios. Efforts within the field have been directed towards developing techniques aimed at enhancing the efficiency of LLM inference. This paper presents a comprehensive survey of the existing literature on efficient LLM inference. We start by analyzing the primary causes of the inefficient LLM inference, i.e., the large model size, the quadratic-complexity attention operation, and the auto-regressive decoding approach. Then, we introduce a comprehensive taxonomy that organizes the current literature into data-level, model-level, and system-level optimization. Moreover, the paper includes comparative experiments on representative methods within critical sub-fields to provide quantitative insights. Last but not least, we provide some knowledge summary and discuss future research directions.

Ziyi Ding, Chenfei Ye-Hao, Zheyuan Wang et al.

Causal discovery aims to recover ``what causes what'', but classical constraint-based methods (e.g., PC, FCI) suffer from error propagation, and recent LLM-based causal oracles often behave as opaque, confidence-free black boxes. This paper introduces Tree-Query, a tree-structured, multi-expert LLM framework that reduces pairwise causal discovery to a short sequence of queries about backdoor paths, (in)dependence, latent confounding, and causal direction, yielding interpretable judgments with robustness-aware confidence scores. Theoretical guarantees are provided for asymptotic identifiability of four pairwise relations. On data-free benchmarks derived from Mooij et al. and UCI causal graphs, Tree-Query improves structural metrics over direct LLM baselines, and a diet--weight case study illustrates confounder screening and stable, high-confidence causal conclusions. Tree-Query thus offers a principled way to obtain data-free causal priors from LLMs that can complement downstream data-driven causal discovery. Code is available at https://anonymous.4open.science/r/Repo-9B3E-4F96.

Hongyi Chen, Jianhai Shu, Jingtao Ding et al.

Langevin dynamics sampling suffers from extremely low generation speed, fundamentally limited by numerous fine-grained iterations to converge to the target distribution. We introduce PID-controlled Langevin Dynamics (PIDLD), a novel sampling acceleration algorithm that reinterprets the sampling process using control-theoretic principles. By treating energy gradients as feedback signals, PIDLD combines historical gradients (the integral term) and gradient trends (the derivative term) to efficiently traverse energy landscapes and adaptively stabilize, thereby significantly reducing the number of iterations required to produce high-quality samples. Our approach requires no additional training, datasets, or prior information, making it immediately integrable with any Langevin-based method. Extensive experiments across image generation and reasoning tasks demonstrate that PIDLD achieves higher quality with fewer steps, making Langevin-based generative models more practical for efficiency-critical applications. The implementation can be found at \href{https://github.com/tsinghua-fib-lab/PIDLD}{https://github.com/tsinghua-fib-lab/PIDLD}.

Yushan Liu, Peibo Sun, Shoujie Li et al.

World Action Models (WAMs) enhance Vision-Language-Action policies by jointly predicting scene evolution and robot actions, but existing methods usually represent the predicted world as holistic images, video tokens, or global latents. These representations are difficult for an action decoder to address when an instruction refers to a particular object, especially under scene shifts where object identity is entangled with context. We propose OA-WAM, an Object-Addressable World Action Model for robust robot manipulation. OA-WAM decomposes each frame into N+1 slot states, with one robot slot and N object slots. Each slot contains a persistent address vector and a time-varying content vector, and is fused with text, image, proprioception, and past-action tokens in a block-causal sequence. A world head predicts next-frame slot states, while a flow-matching action head decodes a 16-step continuous action chunk in the same forward pass. Addressability is enforced by routing cross-slot attention through address-only keys and resetting the address slice at every transformer layer, separating which object to act on from what that object currently is without adding extra tokens. OA-WAM matches strong VLA and WAM baselines on LIBERO (97.8%) and SimplerEnv (79.3%), reaches state-of-the-art performance on the most relevant LIBERO-Plus geometric axes, and remains competitive on the seven-axis aggregate. A causal slot-intervention test yields a swap-binding cosine of 0.87, versus at most 0.09 for holistic baselines. These results suggest that addressable object states provide an effective interface for robust world-action modeling under scene perturbations.

Qitai Tan, Yiyun Chen, Mo Li et al.

Recent advances in deep learning have driven rapid progress in time series forecasting, yet many state-of-the-art models continue to struggle with robust performance in real-world applications, even when they achieve strong results on standard benchmark datasets. This persistent gap can be attributed to the black-box nature of deep learning architectures and the inherent limitations of current evaluation frameworks, which frequently lack the capacity to provide clear, quantitative insights into the specific strengths and weaknesses of different models, thereby complicating the selection of appropriate models for particular forecasting scenarios. To address these issues, we propose a synthetic data-driven evaluation paradigm, SynTSBench, that systematically assesses fundamental modeling capabilities of time series forecasting models through programmable feature configuration. Our framework isolates confounding factors and establishes an interpretable evaluation system with three core analytical dimensions: (1) temporal feature decomposition and capability mapping, which enables systematic evaluation of model capacities to learn specific pattern types; (2) robustness analysis under data irregularities, which quantifies noise tolerance thresholds and anomaly recovery capabilities; and (3) theoretical optimum benchmarking, which establishes performance boundaries for each pattern type-enabling direct comparison between model predictions and mathematical optima. Our experiments show that current deep learning models do not universally approach optimal baselines across all types of temporal features.The code is available at https://github.com/TanQitai/SynTSBench

Huining Yuan, Zelai Xu, Zheyue Tan et al.

Developing Large Language Models (LLMs) to cooperate and compete effectively within multi-agent systems is a critical step towards more advanced intelligence. While reinforcement learning (RL) has proven effective for enhancing reasoning in single-agent tasks, its extension to multi-turn, multi-agent scenarios remains underexplored due to the challenges of long-horizon credit assignment and agent-specific advantage estimation. To address these challenges, we introduce MARS, an end-to-end RL framework that incentivizes Multi-Agent Reasoning of LLMs through Self-play in both cooperative and competitive games. MARS features a turn-level advantage estimator that aligns learning signals with each interaction for credit assignment, and an agent-specific advantage normalization to stabilize multi-agent training. By learning with self-play across cooperative and competitive games, the MARS agent trained from Qwen3-4B develops strong strategic abilities that generalize to held-out games with up to 28.7% performance improvements. More importantly, the capability acquired through self-play generalizes beyond games, yielding consistent performance gains of multi-agent systems in reasoning benchmarks. When integrated into leading multi-agent systems, our MARS agent achieves significant performance gains of 10.0% on AIME and 12.5% on GPQA-Diamond. These results establish end-to-end RL training with self-play in strategic games as a powerful approach for developing generalizable multi-agent reasoning capabilities in LLMs. Our code and models are publicly available at https://github.com/thu-nics/MARS.

Yongzhen Wang, Liangliang Chen, Bingwen Hu et al.

Recent progress in image restoration has underscored Spatial State Models (SSMs) as powerful tools for modeling long-range dependencies, owing to their appealing linear complexity and computational efficiency. However, SSM-based approaches exhibit limitations in reconstructing localized structures and tend to be less effective when handling high-dimensional data, frequently resulting in suboptimal recovery of fine image features. To tackle these challenges, we introduce Laplace-Mamba, a novel framework that integrates Laplace frequency prior with a hybrid Mamba-CNN architecture for efficient image dehazing. Leveraging the Laplace decomposition, the image is disentangled into low-frequency components capturing global texture and high-frequency components representing edges and fine details. This decomposition enables specialized processing via dual parallel pathways: the low-frequency branch employs SSMs for global context modeling, while the high-frequency branch utilizes CNNs to refine local structural details, effectively addressing diverse haze scenarios. Notably, the Laplace transformation facilitates information-preserving downsampling of low-frequency components in accordance with the Nyquist theory, thereby significantly improving computational efficiency. Extensive evaluations across multiple benchmarks demonstrate that our method outperforms state-of-the-art approaches in both restoration quality and efficiency. The source code and pretrained models are available at https://github.com/yz-wang/Laplace-Mamba.

Hongyi Chen, Jingtao Ding, Xiaojun Liang et al.

Source localization in graph information propagation is essential for mitigating network disruptions, including misinformation spread, cyber threats, and infrastructure failures. Existing deep generative approaches face significant challenges in real-world applications due to limited propagation data availability. We present SIDSL (\textbf{S}tructure-prior \textbf{I}nformed \textbf{D}iffusion model for \textbf{S}ource \textbf{L}ocalization), a generative diffusion framework that leverages topology-aware priors to enable robust source localization with limited data. SIDSL addresses three key challenges: unknown propagation patterns through structure-based source estimations via graph label propagation, complex topology-propagation relationships via a propagation-enhanced conditional denoiser with GNN-parameterized label propagation module, and class imbalance through structure-prior biased diffusion initialization. By learning pattern-invariant features from synthetic data generated by established propagation models, SIDSL enables effective knowledge transfer to real-world scenarios. Experimental evaluation on four real-world datasets demonstrates superior performance with 7.5-13.3\% F1 score improvements over baselines, including over 19\% improvement in few-shot and 40\% in zero-shot settings, validating the framework's effectiveness for practical source localization. Our code can be found \href{https://github.com/tsinghua-fib-lab/SIDSL}{here}.

Kuo Yang, Duo Li, Menghan Hu et al.

For semi-supervised learning with imbalance classes, the long-tailed distribution of data will increase the model prediction bias toward dominant classes, undermining performance on less frequent classes. Existing methods also face challenges in ensuring the selection of sufficiently reliable pseudo-labels for model training and there is a lack of mechanisms to adjust the selection of more reliable pseudo-labels based on different training stages. To mitigate this issue, we introduce uncertainty into the modeling process for pseudo-label sampling, taking into account that the model performance on the tailed classes varies over different training stages. For example, at the early stage of model training, the limited predictive accuracy of model results in a higher rate of uncertain pseudo-labels. To counter this, we propose an Uncertainty-Aware Dynamic Threshold Selection (UDTS) approach. This approach allows the model to perceive the uncertainty of pseudo-labels at different training stages, thereby adaptively adjusting the selection thresholds for different classes. Compared to other methods such as the baseline method FixMatch, UDTS achieves an increase in accuracy of at least approximately 5.26%, 1.75%, 9.96%, and 1.28% on the natural scene image datasets CIFAR10-LT, CIFAR100-LT, STL-10-LT, and the medical image dataset TissueMNIST, respectively. The source code of UDTS is publicly available at: https://github.com/yangk/UDTS.

Shen Zheng, Jiaran Cai, Yuansheng Guan et al.

Recent progress in diffusion models has significantly advanced the field of human image animation. While existing methods can generate temporally consistent results for short or regular motions, significant challenges remain, particularly in generating long-duration videos. Furthermore, the synthesis of fine-grained facial and hand details remains under-explored, limiting the applicability of current approaches in real-world, high-quality applications. To address these limitations, we propose a diffusion transformer (DiT)-based framework which focuses on generating high-fidelity and long-duration human animation videos. First, we design a set of hybrid implicit guidance signals and a sharpness guidance factor, enabling our framework to additionally incorporate detailed facial and hand features as guidance. Next, we incorporate the time-aware position shift fusion module, modify the input format within the DiT backbone, and refer to this mechanism as the Position Shift Adaptive Module, which enables video generation of arbitrary length. Finally, we introduce a novel data augmentation strategy and a skeleton alignment model to reduce the impact of human shape variations across different identities. Experimental results demonstrate that our method outperforms existing state-of-the-art approaches, achieving superior performance in both high-fidelity and long-duration human image animation.

Zhaojian Yu, Yinghao Wu, Yilun Zhao et al.

Large Language Models (LLMs) can achieve enhanced complex problem-solving through test-time computing scaling, yet this often entails longer contexts and numerous reasoning token costs. In this paper, we propose an efficient test-time scaling method that trains LLMs on code-related reasoning trajectories, facilitating their reduction of excess thinking tokens while maintaining performance. First, we create Z1-Code-Reasoning-107K, a curated dataset of simple and complex coding problems paired with their short and long solution trajectories. Second, we present a novel Shifted Thinking Window to mitigate overthinking overhead by removing context-delimiting tags (e.g., <think>. . . </think>) and capping reasoning tokens. Trained with long and short trajectory data and equipped with Shifted Thinking Window, our model, Z1-7B, demonstrates the ability to adjust its reasoning level as the complexity of problems and exhibits efficient test-time scaling across different reasoning tasks that matches R1-Distill-Qwen-7B performance with about 30% of its average thinking tokens. Notably, fine-tuned with only code trajectories, Z1-7B demonstrates generalization to broader reasoning tasks (47.5% on GPQA Diamond). Our analysis of efficient reasoning elicitation also provides valuable insights for future research.

Xubing Ye, Yukang Gan, Yixiao Ge et al. · tsinghua

Large Vision Language Models (LVLMs) have achieved significant success across multi-modal tasks. However, the computational cost of processing long visual tokens can be prohibitively expensive on resource-limited devices. Previous methods have identified redundancy in visual tokens within the Large Language Model (LLM) decoder layers and have mitigated this by pruning tokens using a pre-defined or fixed ratio, thereby reducing computational overhead. Nonetheless, we observe that the impact of pruning ratio varies across different LLM layers and instances (image-prompt pairs). Therefore, it is essential to develop a layer-wise and instance-wise vision token pruning strategy to balance computational cost and model performance effectively. We propose ATP-LLaVA, a novel approach that adaptively determines instance-specific token pruning ratios for each LLM layer. Specifically, we introduce an Adaptive Token Pruning (ATP) module, which computes the importance score and pruning threshold based on input instance adaptively. The ATP module can be seamlessly integrated between any two LLM layers with negligible computational overhead. Additionally, we develop a Spatial Augmented Pruning (SAP) strategy that prunes visual tokens with both token redundancy and spatial modeling perspectives. Our approach reduces the average token count by 75% while maintaining performance, with only a minimal 1.9% degradation across seven widely used benchmarks. The project page can be accessed via https://yxxxb.github.io/ATP-LLaVA-page/.

Zhaojian Yu, Yilun Zhao, Arman Cohan et al.

We introduce self-invoking code generation, a new task designed to evaluate the progressive reasoning and problem-solving capabilities of LLMs. In this task, models are presented with a base problem and a related, more complex problem. They must solve the base problem and then utilize its solution to address the more complex one. This work features three key contributions. First, we propose a general recipe for generating more challenging versions of existing benchmarks, resulting in three new benchmarks: HumanEval Pro, MBPP Pro, and BigCodeBench-Lite Pro, specifically designed to assess LLMs on self-invoking code generation. Second, from the analysis of experimental results over twenty LLMs on our benchmarks, we have two important observations: (i) Most LLMs excel in traditional code generation benchmarks like HumanEval and MBPP, but their performance declines on self-invoking tasks. For example, o1-mini achieves 96.2% pass@1 on HumanEval but only 76.2% on HumanEval Pro. (ii) On self-invoking code generation task, the instruction-tuned models demonstrate only marginal improvements compared to the base models. Third, we disclose the types of failure modes that exist in our evaluation results. All these results underscore the need for further advancements in self-invoking code generation tasks and provide a new direction for future research on enhancing LLMs' code reasoning capabilities.

Jingyi Wang, Lei Zhu, Tengjin Weng et al.

Reinforcement Learning with Verifiable Rewards (RLVR) has advanced the reasoning capabilities of Large Language Models (LLMs) by leveraging direct outcome verification instead of learned reward models. Building on this paradigm, Group Relative Policy Optimization (GRPO) eliminates the need for critic models but suffers from indiscriminate credit assignment for intermediate steps, which limits its ability to identify effective reasoning strategies and incurs overthinking. In this work, we introduce a model-free and verifiable process supervision via probing the model's belief in the correct answer throughout its reasoning trajectory. By segmenting the generation into discrete steps and tracking the conditional probability of the correct answer appended at each segment boundary, we efficiently compute interpretable segment-wise progress measurements to refine GRPO's trajectory-level feedback. This approach enables more targeted and sample-efficient policy updates, while avoiding the need for intermediate supervision derived from costly Monte Carlo rollouts or auxiliary models. Experiments on mathematical and general-domain benchmarks show consistent gains over GRPO across diverse models: up to 2.6-point accuracy improvements and 13.7% reasoning-length reductions on math tasks, and up to 2.4 points and 4% on general-domain tasks, demonstrating strong generalization.

Siqi Ping, Yuzhu Mao, Yang Liu et al.

Although large-scale pre-trained models hold great potential for adapting to downstream tasks through fine-tuning, the performance of such fine-tuned models is often limited by the difficulty of collecting sufficient high-quality, task-specific data. Federated Learning (FL) offers a promising solution by enabling fine-tuning across large-scale clients with a variety of task data, but it is bottlenecked by significant communication overhead due to the pre-trained models' extensive size. This paper addresses the high communication cost for fine-tuning large pre-trained models within FL frameworks through low-rank fine-tuning. Specifically, we train a low-rank adapter for each individual task on the client side, followed by server-side clustering for similar group of adapters to achieve task-specific aggregation. Extensive experiments on various language and vision tasks, such as GLUE and CIFAR-10/100, reveal the evolution of task-specific adapters throughout the FL training process and verify the effectiveness of the proposed low-rank task-specific adapter clustering (TAC) method.