Pu Wang

Ekkasit Pinyoanuntapong, Ayman Ali, Pu Wang et al.

Most existing gait recognition methods are appearance-based, which rely on the silhouettes extracted from the video data of human walking activities. The less-investigated skeleton-based gait recognition methods directly learn the gait dynamics from 2D/3D human skeleton sequences, which are theoretically more robust solutions in the presence of appearance changes caused by clothes, hairstyles, and carrying objects. However, the performance of skeleton-based solutions is still largely behind the appearance-based ones. This paper aims to close such performance gap by proposing a novel network model, GaitMixer, to learn more discriminative gait representation from skeleton sequence data. In particular, GaitMixer follows a heterogeneous multi-axial mixer architecture, which exploits the spatial self-attention mixer followed by the temporal large-kernel convolution mixer to learn rich multi-frequency signals in the gait feature maps. Experiments on the widely used gait database, CASIA-B, demonstrate that GaitMixer outperforms the previous SOTA skeleton-based methods by a large margin while achieving a competitive performance compared with the representative appearance-based solutions. Code will be available at https://github.com/exitudio/gaitmixer

Pu Wang, Yao-Xiang Ding

We study $N$-armed stochastic dueling bandits under the Condorcet-winner assumption, where three widely adopted objectives are considered: best-arm identification (BAI), weak regret, and strong regret. We propose Tree-Guided Identify-Then-Exploit (TG-ITE), the first unified framework to tackle all these objectives to our knowledge. Without requiring stronger assumptions, we propose a shared tree-guided identification approach to find a high-confidence incumbent within $O(N)$ comparisons. We further propose varied exploitation strategies to utilize this warm-start stage to optimize the specific objectives at hand. This methodology enables our approach to (1) achieve $O(N)$ sample complexity in BAI without commonly adopted stronger assumptions; (2) build the first winner-stays-style algorithm to achieve $O(N)$ weak regret; (3) enjoy the same $O(N \log T)$ guarantee as specialized strong-regret approaches; (4) realize the joint optimization of BAI and weak regret with $O(N)$ guarantees for both, eliminating the sub-optimal gap of $O(\log N)$ in the existing approach. Our results provide evidence that the trade-off between BAI and regret minimization is relatively benign in dueling bandits.

Peizhao Li, Pu Wang, Karl Berntorp et al.

We consider the object recognition problem in autonomous driving using automotive radar sensors. Comparing to Lidar sensors, radar is cost-effective and robust in all-weather conditions for perception in autonomous driving. However, radar signals suffer from low angular resolution and precision in recognizing surrounding objects. To enhance the capacity of automotive radar, in this work, we exploit the temporal information from successive ego-centric bird-eye-view radar image frames for radar object recognition. We leverage the consistency of an object's existence and attributes (size, orientation, etc.), and propose a temporal relational layer to explicitly model the relations between objects within successive radar images. In both object detection and multiple object tracking, we show the superiority of our method compared to several baseline approaches.

Jianyi Zhang, Hao Frank Yang, Ang Li et al.

Previous studies on federated learning (FL) often encounter performance degradation due to data heterogeneity among different clients. In light of the recent advances in multimodal large language models (MLLMs), such as GPT-4v and LLaVA, which demonstrate their exceptional proficiency in multimodal tasks, such as image captioning and multimodal question answering. We introduce a novel federated learning framework, named Multimodal Large Language Model Assisted Federated Learning (MLLM-LLaVA-FL), which employs powerful MLLMs at the server end to address the heterogeneous and long-tailed challenges. Owing to the advanced cross-modality representation capabilities and the extensive open-vocabulary prior knowledge of MLLMs, our framework is adept at harnessing the extensive, yet previously underexploited, open-source data accessible from websites and powerful server-side computational resources. Hence, the MLLM-LLaVA-FL not only enhances the performance but also avoids increasing the risk of privacy leakage and the computational burden on local devices, distinguishing it from prior methodologies. Our framework has three key stages. Initially, we conduct global visual-text pretraining of the model. This pretraining is facilitated by utilizing the extensive open-source data available online, with the assistance of MLLMs. Subsequently, the pretrained model is distributed among various clients for local training. Finally, once the locally trained models are transmitted back to the server, a global alignment is carried out under the supervision of MLLMs to further enhance the performance. Experimental evaluations on established benchmarks, show that our framework delivers promising performance in the typical scenarios with data heterogeneity and long-tail distribution across different clients in FL.

Toshiaki Koike-Akino, Pu Wang, Ye Wang

Commercial Wi-Fi devices can be used for integrated sensing and communications (ISAC) to jointly exchange data and monitor indoor environment. In this paper, we investigate a proof-of-concept approach using automated quantum machine learning (AutoQML) framework called AutoAnsatz to recognize human gesture. We address how to efficiently design quantum circuits to configure quantum neural networks (QNN). The effectiveness of AutoQML is validated by an in-house experiment for human pose recognition, achieving state-of-the-art performance greater than 80% accuracy for a limited data size with a significantly small number of trainable parameters.

Toshiaki Koike-Akino, Pu Wang, Ye Wang

Beyond data communications, commercial-off-the-shelf Wi-Fi devices can be used to monitor human activities, track device locomotion, and sense the ambient environment. In particular, spatial beam attributes that are inherently available in the 60-GHz IEEE 802.11ad/ay standards have shown to be effective in terms of overhead and channel measurement granularity for these indoor sensing tasks. In this paper, we investigate transfer learning to mitigate domain shift in human monitoring tasks when Wi-Fi settings and environments change over time. As a proof-of-concept study, we consider quantum neural networks (QNN) as well as classical deep neural networks (DNN) for the future quantum-ready society. The effectiveness of both DNN and QNN is validated by an in-house experiment for human pose recognition, achieving greater than 90% accuracy with a limited data size.

Pu Wang, Hugo Van hamme

End-to-end spoken language understanding (SLU) systems benefit from pretraining on large corpora, followed by fine-tuning on application-specific data. The resulting models are too large for on-edge applications. For instance, BERT-based systems contain over 110M parameters. Observing the model is overparameterized, we propose lean transformer structure where the dimension of the attention mechanism is automatically reduced using group sparsity. We propose a variant where the learned attention subspace is transferred to an attention bottleneck layer. In a low-resource setting and without pre-training, the resulting compact SLU model achieves accuracies competitive with pre-trained large models.

Ekkasit Pinyoanuntapong, Ayman Ali, Kalvik Jakkala et al.

mmWave radar-based gait recognition is a novel user identification method that captures human gait biometrics from mmWave radar return signals. This technology offers privacy protection and is resilient to weather and lighting conditions. However, its generalization performance is yet unknown and limits its practical deployment. To address this problem, in this paper, a non-synthetic dataset is collected and analyzed to reveal the presence of spatial and temporal domain shifts in mmWave gait biometric data, which significantly impacts identification accuracy. To mitigate this issue, a novel self-aligned domain adaptation method called GaitSADA is proposed. GaitSADA improves system generalization performance by using a two-stage semi-supervised model training approach. The first stage employs semi-supervised contrastive learning to learn a compact gait representation from both source and target domain data, aligning source-target domain distributions implicitly. The second stage uses semi-supervised consistency training with centroid alignment to further close source-target domain gap by pseudo-labelling the target-domain samples, clustering together the samples belonging to the same class but from different domains, and pushing the class centroid close to the weight vector of each class. Experiments show that GaitSADA outperforms representative domain adaptation methods with an improvement ranging from 15.41\% to 26.32\% on average accuracy in low data regimes. Code and dataset will be available at https://exitudio.github.io/GaitSADA

Guangyu Sun, Umar Khalid, Matias Mendieta et al.

Federated learning (FL) has emerged as a promising paradigm for enabling the collaborative training of models without centralized access to the raw data on local devices. In the typical FL paradigm (e.g., FedAvg), model weights are sent to and from the server each round to participating clients. Recently, the use of small pre-trained models has been shown to be effective in federated learning optimization and improving convergence. However, recent state-of-the-art pre-trained models are getting more capable but also have more parameters, known as the "Foundation Models." In conventional FL, sharing the enormous model weights can quickly put a massive communication burden on the system, especially if more capable models are employed. Can we find a solution to enable those strong and readily available pre-trained models in FL to achieve excellent performance while simultaneously reducing the communication burden? To this end, we investigate the use of parameter-efficient fine-tuning in federated learning and thus introduce a new framework: FedPEFT. Specifically, we systemically evaluate the performance of FedPEFT across a variety of client stability, data distribution, and differential privacy settings. By only locally tuning and globally sharing a small portion of the model weights, significant reductions in the total communication overhead can be achieved while maintaining competitive or even better performance in a wide range of federated learning scenarios, providing insight into a new paradigm for practical and effective federated systems.

Toshiaki Koike-Akino, Pu Wang, Genki Yamashita et al.

A learning-based THz multi-layer imaging has been recently used for contactless three-dimensional (3D) positioning and encoding. We show a proof-of-concept demonstration of an emerging quantum machine learning (QML) framework to deal with depth variation, shadow effect, and double-sided content recognition, through an experimental validation.

Ayman Ali, Ekkasit Pinyoanuntapong, Pu Wang et al.

Recently, there has been a remarkable increase in the interest towards skeleton-based action recognition within the research community, owing to its various advantageous features, including computational efficiency, representative features, and illumination invariance. Despite this, researchers continue to explore and investigate the most optimal way to represent human actions through skeleton representation and the extracted features. As a result, the growth and availability of human action recognition datasets have risen substantially. In addition, deep learning-based algorithms have gained widespread popularity due to the remarkable advancements in various computer vision tasks. Most state-of-the-art contributions in skeleton-based action recognition incorporate a Graph Neural Network (GCN) architecture for representing the human body and extracting features. Our research demonstrates that Convolutional Neural Networks (CNNs) can attain comparable results to GCN, provided that the proper training techniques, augmentations, and optimizers are applied. Our approach has been rigorously validated, and we have achieved a score of 95% on the NTU-60 dataset

Ce Zheng, Xianpeng Liu, Qucheng Peng et al.

Human mesh recovery (HMR) provides rich human body information for various real-world applications. While image-based HMR methods have achieved impressive results, they often struggle to recover humans in dynamic scenarios, leading to temporal inconsistencies and non-smooth 3D motion predictions due to the absence of human motion. In contrast, video-based approaches leverage temporal information to mitigate this issue. In this paper, we present DiffMesh, an innovative motion-aware Diffusion-like framework for video-based HMR. DiffMesh establishes a bridge between diffusion models and human motion, efficiently generating accurate and smooth output mesh sequences by incorporating human motion within the forward process and reverse process in the diffusion model. Extensive experiments are conducted on the widely used datasets (Human3.6M \cite{h36m_pami} and 3DPW \cite{pw3d2018}), which demonstrate the effectiveness and efficiency of our DiffMesh. Visual comparisons in real-world scenarios further highlight DiffMesh's suitability for practical applications.

Ziyan Wang, Enmao Diao, Qi Le et al.

Reasoning LLMs (RLMs) such as OpenAI o1, DeepSeek-R1, and Qwen3 deliver strong multi-step reasoning through chain-of-thought generation, but their large model sizes and lengthy decode-time outputs make them costly to deploy and unsuitable for resource-constrained settings. To reduce computing and memory cost, pruning offers a promising solution by removing unimportant parameters. However, despite their success on standard LLMs, existing pruning methods severely damage RLMs, as even moderate sparsity (e.g., 20%) can collapse accuracy and completely disrupt the model's reasoning coherence. We begin by analyzing why existing pruning pipelines fail on reasoning LLMs and find that their brittleness largely stems from a mismatch between the calibration data, the pruning objective, and the model's decode-time reasoning behavior. Our study further shows that the most reliable calibration signal comes not from human-written labels but from the model's own self-generated reasoning traces, which more accurately reflect its inference distribution. Guided by these insights, we introduce RESP, a self-reflective structured pruning framework that aligns pruning decisions with the model's reasoning dynamics through self-generated calibration, decode-only gradient-based importance estimation, and progressive regeneration that maintains calibration fidelity as sparsity increases. Experiments on Qwen3-8B demonstrate that RESP markedly outperforms existing structured pruning methods on both GSM8K and MathQA, preserving near-dense accuracy at 20-30% sparsity and substantially mitigating performance collapse at higher sparsity levels. At 40% sparsity, RESP attains 81.3% accuracy on GSM8K and 59.6% on MathQA, surpassing the strongest baselines by 66.87% and 47%, respectively.

Kunqi Wang, Daolin Si, Pu Wang et al.

Matching the rail cross-section profiles measured on site with the designed profile is a must to evaluate the wear of the rail, which is very important for track maintenance and rail safety. So far, the measured rail profiles to be matched usually have four features, that is, large amount of data, diverse section shapes, hardware made errors, and human experience needs to be introduced to solve the complex situation on site during matching process. However, traditional matching methods based on feature points or feature lines could no longer meet the requirements. To this end, we first establish the rail profiles matching dataset composed of 46386 pairs of professional manual matched data, then propose a general high-precision method for rail profiles matching using pre-trained convolutional neural network (CNN). This new method based on deep learning is promising to be the dominant approach for this issue. Source code is at https://github.com/Kunqi1994/Deep-learning-on-rail-profile-matching.

Ayman Ali, Ekkasit Pinyoanuntapong, Pu Wang et al.

In this research, we address the challenge faced by existing deep learning-based human mesh reconstruction methods in balancing accuracy and computational efficiency. These methods typically prioritize accuracy, resulting in large network sizes and excessive computational complexity, which may hinder their practical application in real-world scenarios, such as virtual reality systems. To address this issue, we introduce a modular multi-stage lightweight graph-based transformer network for human pose and shape estimation from 2D human pose, a pose-based human mesh reconstruction approach that prioritizes computational efficiency without sacrificing reconstruction accuracy. Our method consists of a 2D-to-3D lifter module that utilizes graph transformers to analyze structured and implicit joint correlations in 2D human poses, and a mesh regression module that combines the extracted pose features with a mesh template to produce the final human mesh parameters.

Muhammad Usama Saleem, Tejasvi Ravi, Tianyu Xu et al.

Multimodal music creation requires models that can both generate audio from high-level cues and edit existing mixtures in a targeted manner. Yet most multimodal music systems are built for a single task and a fixed prompting interface, making their conditioning brittle when guidance is ambiguous, temporally misaligned, or partially missing. Common additive fusion or feature concatenation further weakens cross-modal grounding, often causing prompt drift and spurious musical content during generation and editing. We propose MAGE, a modality-agnostic framework that unifies multimodal music generation and mixture-grounded editing within a single continuous latent formulation. At its core, MAGE uses a Controlled Multimodal FluxFormer, a flow-based Transformer that learns controllable latent trajectories for synthesis and editing under any available subset of conditions. To improve grounding, we introduce Audio-Visual Nexus Alignment to select temporally consistent visual evidence for the audio timeline, and a cross-gated modulation mechanism that applies multiplicative control from aligned visual and textual cues to the audio latents, suppressing unsupported components rather than injecting them. Finally, we train with a dynamic modality-masking curriculum that exposes the model to text-only, visual-only, joint multimodal, and mixture-guided settings, enabling robust inference under missing modalities without training separate models. Experiments on the MUSIC benchmark show that MAGE supports effective multimodal-guided music generation and targeted editing, achieving competitive quality while offering a lightweight and flexible interface tailored to practical music workflows.

Sheng Xu, Enshu Wang, Hongfei Xue et al.

Collaborative perception allows connected vehicles to overcome occlusions and limited viewpoints by sharing sensory information. However, existing approaches struggle to achieve high accuracy under strict bandwidth constraints and remain highly vulnerable to random transmission packet loss. We introduce QPoint2Comm, a quantized point-cloud communication framework that dramatically reduces bandwidth while preserving high-fidelity 3D information. Instead of transmitting intermediate features, QPoint2Comm directly communicates quantized point-cloud indices using a shared codebook, enabling efficient reconstruction with lower bandwidth than feature-based methods. To ensure robustness to possible communication packet loss, we employ a masked training strategy that simulates random packet loss, allowing the model to maintain strong performance even under severe transmission failures. In addition, a cascade attention fusion module is proposed to enhance multi-vehicle information integration. Extensive experiments on both simulated and real-world datasets demonstrate that QPoint2Comm sets a new state of the art in accuracy, communication efficiency, and resilience to packet loss.

Muhammad Usama Saleem, Mayur Jagdishbhai Patel, Ekkasit Pinyoanuntapong et al.

We propose LiveGesture, the first fully streamable, speech-driven full-body gesture generation framework that operates with zero look-ahead and supports arbitrary sequence length. Unlike existing co-speech gesture methods, which are designed for offline generation and either treat body regions independently or entangle all joints within a single model, LiveGesture is built from the ground up for causal, region-coordinated motion generation. LiveGesture consists of two main modules: the Streamable Vector Quantized Motion Tokenizer (SVQ) and the Hierarchical Autoregressive Transformer (HAR). The SVQ tokenizer converts the motion sequence of each body region into causal, discrete motion tokens, enabling real-time, streamable token decoding. On top of SVQ, HAR employs region-expert autoregressive (xAR) transformers to model expressive, fine-grained motion dynamics for each body region. A causal spatio-temporal fusion module (xAR Fusion) then captures and integrates correlated motion dynamics across regions. Both xAR and xAR Fusion are conditioned on live, continuously arriving audio signals encoded by a streamable causal audio encoder. To enhance robustness under streaming noise and prediction errors, we introduce autoregressive masking training, which leverages uncertainty-guided token masking and random region masking to expose the model to imperfect, partially erroneous histories during training. Experiments on the BEAT2 dataset demonstrate that LiveGesture produces coherent, diverse, and beat-synchronous full-body gestures in real time, matching or surpassing state-of-the-art offline methods under true zero look-ahead conditions.

Yucheng Xin, Dawei Zhao, Xiang Chen et al.

Ultra-high-definition (UHD) image deblurring poses significant challenges for UHD restoration methods, which must balance fine-grained detail recovery and practical inference efficiency. Although prominent discriminative and generative methods have achieved remarkable results, a trade-off persists between computational cost and the ability to generate fine-grained detail for UHD image deblurring tasks. To further alleviate these issues, we propose a novel autoregressive flow method for UHD image deblurring with an ill-conditioned constraint. Our core idea is to decompose UHD restoration into a progressive, coarse-to-fine process: at each scale, the sharp estimate is formed by upsampling the previous-scale result and adding a current-scale residual, enabling stable, stage-wise refinement from low to high resolution. We further introduce Flow Matching to model residual generation as a conditional vector field and perform few-step ODE sampling with efficient Euler/Heun solvers, enriching details while keeping inference affordable. Since multi-step generation at UHD can be numerically unstable, we propose an ill-conditioning suppression scheme by imposing condition-number regularization on a feature-induced attention matrix, improving convergence and cross-scale consistency. Our method demonstrates promising performance on blurred images at 4K (3840$\times$2160) or higher resolutions.

Junhui Li, Pu Wang, Jialu Li et al.

Recent high-performance transformer-based speech enhancement models demonstrate that time domain methods could achieve similar performance as time-frequency domain methods. However, time-domain speech enhancement systems typically receive input audio sequences consisting of a large number of time steps, making it challenging to model extremely long sequences and train models to perform adequately. In this paper, we utilize smaller audio chunks as input to achieve efficient utilization of audio information to address the above challenges. We propose a dual-phase audio transformer for denoising (DPATD), a novel model to organize transformer layers in a deep structure to learn clean audio sequences for denoising. DPATD splits the audio input into smaller chunks, where the input length can be proportional to the square root of the original sequence length. Our memory-compressed explainable attention is efficient and converges faster compared to the frequently used self-attention module. Extensive experiments demonstrate that our model outperforms state-of-the-art methods.

Pu Wang, Zhixuan Mao, Jialu Li et al.

Automatic diagnosis of canine pneumothorax is challenged by data scarcity and the need for trustworthy models. To address this, we first introduce a public, pixel-level annotated dataset to facilitate research. We then propose a novel diagnostic paradigm that reframes the task as a synergistic process of signal localization and spectral detection. For localization, our method employs a Vision-Language Model (VLM) to guide an iterative Flow Matching process, which progressively refines segmentation masks to achieve superior boundary accuracy. For detection, the segmented mask is used to isolate features from the suspected lesion. We then apply Random Matrix Theory (RMT), a departure from traditional classifiers, to analyze these features. This approach models healthy tissue as predictable random noise and identifies pneumothorax by detecting statistically significant outlier eigenvalues that represent a non-random pathological signal. The high-fidelity localization from Flow Matching is crucial for purifying the signal, thus maximizing the sensitivity of our RMT detector. This synergy of generative segmentation and first-principles statistical analysis yields a highly accurate and interpretable diagnostic system (source code is available at: https://github.com/Pu-Wang-alt/Canine-pneumothorax).

Qucheng Peng, Hongfei Xue, Pu Wang et al.

3D Human Pose Estimation (3D HPE) is vital in various applications, from person re-identification and action recognition to virtual reality. However, the reliance on annotated 3D data collected in controlled environments poses challenges for generalization to diverse in-the-wild scenarios. Existing domain adaptation (DA) paradigms like general DA and source-free DA for 3D HPE overlook the issues of non-stationary target pose datasets. To address these challenges, we propose a novel task named lifelong domain adaptive 3D HPE. To our knowledge, we are the first to introduce the lifelong domain adaptation to the 3D HPE task. In this lifelong DA setting, the pose estimator is pretrained on the source domain and subsequently adapted to distinct target domains. Moreover, during adaptation to the current target domain, the pose estimator cannot access the source and all the previous target domains. The lifelong DA for 3D HPE involves overcoming challenges in adapting to current domain poses and preserving knowledge from previous domains, particularly combating catastrophic forgetting. We present an innovative Generative Adversarial Network (GAN) framework, which incorporates 3D pose generators, a 2D pose discriminator, and a 3D pose estimator. This framework effectively mitigates domain shifts and aligns original and augmented poses. Moreover, we construct a novel 3D pose generator paradigm, integrating pose-aware, temporal-aware, and domain-aware knowledge to enhance the current domain's adaptation and alleviate catastrophic forgetting on previous domains. Our method demonstrates superior performance through extensive experiments on diverse domain adaptive 3D HPE datasets.

Pu Wang, Pengwen Dai, Chen Wu et al.

In this paper, we propose an efficient visual transformer framework for ultra-high-definition (UHD) image dehazing that addresses the key challenges of slow training speed and high memory consumption for existing methods. Our approach introduces two key innovations: 1) an \textbf{a}daptive \textbf{n}ormalization mechanism inspired by the nGPT architecture that enables ultra-fast and stable training with a network with a restricted range of parameter expressions; and 2) we devise an atmospheric scattering-aware KV caching mechanism that dynamically optimizes feature preservation based on the physical haze formation model. The proposed architecture improves the training convergence speed by \textbf{5 $\times$} while reducing memory overhead, enabling real-time processing of 50 high-resolution images per second on an RTX4090 GPU. Experimental results show that our approach maintains state-of-the-art dehazing quality while significantly improving computational efficiency for 4K/8K image restoration tasks. Furthermore, we provide a new dehazing image interpretable method with the help of an integrated gradient attribution map. Our code can be found here: https://anonymous.4open.science/r/anDehazeFormer-632E/README.md.

Qucheng Peng, Ce Zheng, Zhengming Ding et al.

Human pose estimation (HPE) has received increasing attention recently due to its wide application in motion analysis, virtual reality, healthcare, etc. However, it suffers from the lack of labeled diverse real-world datasets due to the time- and labor-intensive annotation. To cope with the label deficiency issue, one common solution is to train the HPE models with easily available synthetic datasets (source) and apply them to real-world data (target) through domain adaptation (DA). Unfortunately, prevailing domain adaptation techniques within the HPE domain remain predominantly fixated on effecting alignment and aggregation between source and target features, often sidestepping the crucial task of excluding domain-specific representations. To rectify this, we introduce a novel framework that capitalizes on both representation aggregation and segregation for domain adaptive human pose estimation. Within this framework, we address the network architecture aspect by disentangling representations into distinct domain-invariant and domain-specific components, facilitating aggregation of domain-invariant features while simultaneously segregating domain-specific ones. Moreover, we tackle the discrepancy measurement facet by delving into various keypoint relationships and applying separate aggregation or segregation mechanisms to enhance alignment. Extensive experiments on various benchmarks, e.g., Human3.6M, LSP, H3D, and FreiHand, show that our method consistently achieves state-of-the-art performance. The project is available at \url{https://github.com/davidpengucf/EPIC}.

Manish Kumar Govind, Dominick Reilly, Pu Wang et al.

Latent action representations learned from unlabeled videos have recently emerged as a promising paradigm for pretraining vision-language-action (VLA) models without explicit robot action supervision. However, latent actions derived solely from RGB observations primarily encode appearance-driven dynamics and lack explicit 3D geometric structure, which is essential for precise and contact-rich manipulation. To address this limitation, we introduce UniLACT, a transformer-based VLA model that incorporates geometric structure through depth-aware latent pretraining, enabling downstream policies to inherit stronger spatial priors. To facilitate this process, we propose UniLARN, a unified latent action learning framework based on inverse and forward dynamics objectives that learns a shared embedding space for RGB and depth while explicitly modeling their cross-modal interactions. This formulation produces modality-specific and unified latent action representations that serve as pseudo-labels for the depth-aware pretraining of UniLACT. Extensive experiments in both simulation and real-world settings demonstrate the effectiveness of depth-aware unified latent action representations. UniLACT consistently outperforms RGB-based latent action baselines under in-domain and out-of-domain pretraining regimes, as well as on both seen and unseen manipulation tasks.

Xingchi Chen, Pu Wang, Xuerui Li et al.

Ultra-High-Definition (UHD) image dehazing faces challenges such as limited scene adaptability in prior-based methods and high computational complexity with color distortion in deep learning approaches. To address these issues, we propose 4KDehazeFlow, a novel method based on Flow Matching and the Haze-Aware vector field. This method models the dehazing process as a progressive optimization of continuous vector field flow, providing efficient data-driven adaptive nonlinear color transformation for high-quality dehazing. Specifically, our method has the following advantages: 1) 4KDehazeFlow is a general method compatible with various deep learning networks, without relying on any specific network architecture. 2) We propose a learnable 3D lookup table (LUT) that encodes haze transformation parameters into a compact 3D mapping matrix, enabling efficient inference through precomputed mappings. 3) We utilize a fourth-order Runge-Kutta (RK4) ordinary differential equation (ODE) solver to stably solve the dehazing flow field through an accurate step-by-step iterative method, effectively suppressing artifacts. Extensive experiments show that 4KDehazeFlow exceeds seven state-of-the-art methods. It delivers a 2dB PSNR increase and better performance in dense haze and color fidelity.

Matias Mendieta, Taojiannan Yang, Pu Wang et al.

Federated learning (FL) is a promising strategy for performing privacy-preserving, distributed learning with a network of clients (i.e., edge devices). However, the data distribution among clients is often non-IID in nature, making efficient optimization difficult. To alleviate this issue, many FL algorithms focus on mitigating the effects of data heterogeneity across clients by introducing a variety of proximal terms, some incurring considerable compute and/or memory overheads, to restrain local updates with respect to the global model. Instead, we consider rethinking solutions to data heterogeneity in FL with a focus on local learning generality rather than proximal restriction. To this end, we first present a systematic study informed by second-order indicators to better understand algorithm effectiveness in FL. Interestingly, we find that standard regularization methods are surprisingly strong performers in mitigating data heterogeneity effects. Based on our findings, we further propose a simple and effective method, FedAlign, to overcome data heterogeneity and the pitfalls of previous methods. FedAlign achieves competitive accuracy with state-of-the-art FL methods across a variety of settings while minimizing computation and memory overhead. Code is available at https://github.com/mmendiet/FedAlign

Taojiannan Yang, Sijie Zhu, Matias Mendieta et al.

Most existing deep neural networks are static, which means they can only do inference at a fixed complexity. But the resource budget can vary substantially across different devices. Even on a single device, the affordable budget can change with different scenarios, and repeatedly training networks for each required budget would be incredibly expensive. Therefore, in this work, we propose a general method called MutualNet to train a single network that can run at a diverse set of resource constraints. Our method trains a cohort of model configurations with various network widths and input resolutions. This mutual learning scheme not only allows the model to run at different width-resolution configurations but also transfers the unique knowledge among these configurations, helping the model to learn stronger representations overall. MutualNet is a general training methodology that can be applied to various network structures (e.g., 2D networks: MobileNets, ResNet, 3D networks: SlowFast, X3D) and various tasks (e.g., image classification, object detection, segmentation, and action recognition), and is demonstrated to achieve consistent improvements on a variety of datasets. Since we only train the model once, it also greatly reduces the training cost compared to independently training several models. Surprisingly, MutualNet can also be used to significantly boost the performance of a single network, if dynamic resource constraint is not a concern. In summary, MutualNet is a unified method for both static and adaptive, 2D and 3D networks. Codes and pre-trained models are available at \url{https://github.com/taoyang1122/MutualNet}.

Jiawei Zhu, Xin Han, Hanhan Deng et al.

While considering the spatial and temporal features of traffic, capturing the impacts of various external factors on travel is an essential step towards achieving accurate traffic forecasting. However, existing studies seldom consider external factors or neglect the effect of the complex correlations among external factors on traffic. Intuitively, knowledge graphs can naturally describe these correlations. Since knowledge graphs and traffic networks are essentially heterogeneous networks, it is challenging to integrate the information in both networks. On this background, this study presents a knowledge representation-driven traffic forecasting method based on spatial-temporal graph convolutional networks. We first construct a knowledge graph for traffic forecasting and derive knowledge representations by a knowledge representation learning method named KR-EAR. Then, we propose the Knowledge Fusion Cell (KF-Cell) to combine the knowledge and traffic features as the input of a spatial-temporal graph convolutional backbone network. Experimental results on the real-world dataset show that our strategy enhances the forecasting performances of backbones at various prediction horizons. The ablation and perturbation analysis further verify the effectiveness and robustness of the proposed method. To the best of our knowledge, this is the first study that constructs and utilizes a knowledge graph to facilitate traffic forecasting; it also offers a promising direction to integrate external information and spatial-temporal information for traffic forecasting. The source code is available at https://github.com/lehaifeng/T-GCN/tree/master/KST-GCN.

Ling Zhao, Yujiao Song, Chao Zhang et al.

Accurate and real-time traffic forecasting plays an important role in the Intelligent Traffic System and is of great significance for urban traffic planning, traffic management, and traffic control. However, traffic forecasting has always been considered an open scientific issue, owing to the constraints of urban road network topological structure and the law of dynamic change with time, namely, spatial dependence and temporal dependence. To capture the spatial and temporal dependence simultaneously, we propose a novel neural network-based traffic forecasting method, the temporal graph convolutional network (T-GCN) model, which is in combination with the graph convolutional network (GCN) and gated recurrent unit (GRU). Specifically, the GCN is used to learn complex topological structures to capture spatial dependence and the gated recurrent unit is used to learn dynamic changes of traffic data to capture temporal dependence. Then, the T-GCN model is employed to traffic forecasting based on the urban road network. Experiments demonstrate that our T-GCN model can obtain the spatio-temporal correlation from traffic data and the predictions outperform state-of-art baselines on real-world traffic datasets. Our tensorflow implementation of the T-GCN is available at https://github.com/lehaifeng/T-GCN.

Wenhan Wu, Zhishuai Guo, Chen Chen et al.

Stochastic human motion prediction aims to generate diverse, plausible futures from observed sequences. Despite advances in generative modeling, existing methods often produce predictions corrupted by high-frequency jitter and temporal discontinuities. To address these challenges, we introduce KHMP, a novel framework featuring an adaptiveKalman filter applied in the DCT domain to generate high-fidelity human motion predictions. By treating high-frequency DCT coefficients as a frequency-indexed noisy signal, the Kalman filter recursively suppresses noise while preserving motion details. Notably, its noise parameters are dynamically adjusted based on estimated Signal-to-Noise Ratio (SNR), enabling aggressive denoising for jittery predictions and conservative filtering for clean motions. This refinement is complemented by training-time physical constraints (temporal smoothness and joint angle limits) that encode biomechanical principles into the generative model. Together, these innovations establish a new paradigm integrating adaptive signal processing with physics-informed learning. Experiments on the Human3.6M and HumanEva-I datasets demonstrate that KHMP achieves state-of-the-art accuracy, effectively mitigating jitter artifacts to produce smooth and physically plausible motions.

Haoyu Xie, Shengkai Xu, Cheng Guo et al.

Multi-view human mesh recovery (HMR) is broadly deployed in diverse domains where high accuracy and strong generalization are essential. Existing approaches can be broadly grouped into geometry-based and learning-based methods. However, geometry-based methods (e.g., triangulation) rely on cumbersome camera calibration, while learning-based approaches often generalize poorly to unseen camera configurations due to the lack of multi-view training data, limiting their performance in real-world scenarios. To enable calibration-free reconstruction that generalizes to arbitrary camera setups, we propose a training-free framework that leverages pretrained single-view HMR models as strong priors, eliminating the need for multi-view training data. Our method first constructs a robust and consistent multi-view initialization from single-view predictions, and then refines it via test-time optimization guided by multi-view consistency and anatomical constraints. Extensive experiments demonstrate state-of-the-art performance on standard benchmarks, surpassing multi-view models trained with explicit multi-view supervision.

Ekkasit Pinyoanuntapong, Pu Wang, Minwoo Lee et al.

Recent advances in text-to-motion generation using diffusion and autoregressive models have shown promising results. However, these models often suffer from a trade-off between real-time performance, high fidelity, and motion editability. To address this gap, we introduce MMM, a novel yet simple motion generation paradigm based on Masked Motion Model. MMM consists of two key components: (1) a motion tokenizer that transforms 3D human motion into a sequence of discrete tokens in latent space, and (2) a conditional masked motion transformer that learns to predict randomly masked motion tokens, conditioned on the pre-computed text tokens. By attending to motion and text tokens in all directions, MMM explicitly captures inherent dependency among motion tokens and semantic mapping between motion and text tokens. During inference, this allows parallel and iterative decoding of multiple motion tokens that are highly consistent with fine-grained text descriptions, therefore simultaneously achieving high-fidelity and high-speed motion generation. In addition, MMM has innate motion editability. By simply placing mask tokens in the place that needs editing, MMM automatically fills the gaps while guaranteeing smooth transitions between editing and non-editing parts. Extensive experiments on the HumanML3D and KIT-ML datasets demonstrate that MMM surpasses current leading methods in generating high-quality motion (evidenced by superior FID scores of 0.08 and 0.429), while offering advanced editing features such as body-part modification, motion in-betweening, and the synthesis of long motion sequences. In addition, MMM is two orders of magnitude faster on a single mid-range GPU than editable motion diffusion models. Our project page is available at \url{https://exitudio.github.io/MMM-page}.

Jialang Lu, Shuning Sun, Pu Wang et al.

Purple fringing, a persistent artifact caused by Longitudinal Chromatic Aberration (LCA) in camera lenses, has long degraded the clarity and realism of digital imaging. Traditional solutions rely on complex and expensive apochromatic (APO) lens hardware and the extraction of handcrafted features, ignoring the data-driven approach. To fill this gap, we introduce DCA-LUT, the first deep learning framework for purple fringing removal. Inspired by the physical root of the problem, the spatial misalignment of RGB color channels due to lens dispersion, we introduce a novel Chromatic-Aware Coordinate Transformation (CA-CT) module, learning an image-adaptive color space to decouple and isolate fringing into a dedicated dimension. This targeted separation allows the network to learn a precise ``purple fringe channel", which then guides the accurate restoration of the luminance channel. The final color correction is performed by a learned 5D Look-Up Table (5D LUT), enabling efficient and powerful% non-linear color mapping. To enable robust training and fair evaluation, we constructed a large-scale synthetic purple fringing dataset (PF-Synth). Extensive experiments in synthetic and real-world datasets demonstrate that our method achieves state-of-the-art performance in purple fringing removal.

Pu Wang, Shuning Sun, Jialang Lu et al.

Purple flare, a diffuse chromatic aberration artifact commonly found around highlight areas, severely degrades the tone transition and color of the image. Existing traditional methods are based on hand-crafted features, which lack flexibility and rely entirely on fixed priors, while the scarcity of paired training data critically hampers deep learning. To address this issue, we propose a novel network built upon decoupled HSV Look-Up Tables (LUTs). The method aims to simplify color correction by adjusting the Hue (H), Saturation (S), and Value (V) components independently. This approach resolves the inherent color coupling problems in traditional methods. Our model adopts a two-stage architecture: First, a Chroma-Aware Spectral Tokenizer (CAST) converts the input image from RGB space to HSV space and independently encodes the Hue (H) and Value (V) channels into a set of semantic tokens describing the Purple flare status; second, the HSV-LUT module takes these tokens as input and dynamically generates independent correction curves (1D-LUTs) for the three channels H, S, and V. To effectively train and validate our model, we built the first large-scale purple flare dataset with diverse scenes. We also proposed new metrics and a loss function specifically designed for this task. Extensive experiments demonstrate that our model not only significantly outperforms existing methods in visual effects but also achieves state-of-the-art performance on all quantitative metrics.

Ekkasit Pinyoanuntapong, Muhammad Usama Saleem, Pu Wang et al.

Generating human motion from text has been dominated by denoising motion models either through diffusion or generative masking process. However, these models face great limitations in usability by requiring prior knowledge of the motion length. Conversely, autoregressive motion models address this limitation by adaptively predicting motion endpoints, at the cost of degraded generation quality and editing capabilities. To address these challenges, we propose Bidirectional Autoregressive Motion Model (BAMM), a novel text-to-motion generation framework. BAMM consists of two key components: (1) a motion tokenizer that transforms 3D human motion into discrete tokens in latent space, and (2) a masked self-attention transformer that autoregressively predicts randomly masked tokens via a hybrid attention masking strategy. By unifying generative masked modeling and autoregressive modeling, BAMM captures rich and bidirectional dependencies among motion tokens, while learning the probabilistic mapping from textual inputs to motion outputs with dynamically-adjusted motion sequence length. This feature enables BAMM to simultaneously achieving high-quality motion generation with enhanced usability and built-in motion editability. Extensive experiments on HumanML3D and KIT-ML datasets demonstrate that BAMM surpasses current state-of-the-art methods in both qualitative and quantitative measures. Our project page is available at https://exitudio.github.io/BAMM-page

Bingyi Liu, Chuanhui Zhu, Hongfei Xue et al.

Accurate 3D object detection is critical for autonomous driving, necessitating reliable, cost-effective sensors capable of operating in adverse weather conditions. Camera and millimeter-wave radar fusion has emerged as a promising solution; however, these methods often rely on finely annotated radar data, which is scarce and labor-intensive to produce. To address this challenge, we present CLLAP, a Contrastive Learning-based LiDAR-Augmented Pretraining framework that enhances the performance of existing radar-camera fusion methods for 3D object detection. CLLAP leverages abundant LiDAR data to generate pseudo-radar data using the proposed L2R (LiDAR-to-Radar) Sampling method. Then, it incorporates this data into a novel dual-stage, dual-modality contrastive learning strategy, enabling effective self-supervised learning from paired pseudo-radar and image data. This approach facilitates effective pretraining of existing radar-camera fusion models in a plug-and-play manner, enhancing their feature extraction capabilities and improving detection accuracy and robustness. Experimental results using NuScenes and Lyft Level 5 datasets demonstrate significant performance improvements across three baseline models, highlighting CLLAP's effectiveness in advancing radar-camera fusion for autonomous driving applications.

Xin Wang, Chi Ma, Shaobin Chen et al.

Generative recommendation (GR) offers superior modeling capabilities but suffers from prohibitive inference costs due to the repeated encoding of long user histories. While cross-request Key-Value (KV) cache reuse presents a significant optimization opportunity, the massive scale of individual user states creates a storage explosion that far exceeds physical GPU limits. We propose MTServe, a hierarchical cache management system that virtualizes GPU memory by leveraging host RAM as a scalable backup store. To bridge the I/O gap between tiers, MTServe introduces a suite of system-level optimizations, including a hybrid storage layout, an asynchronous data transfer pipeline, and a locality-driven replacement policy. On both public and production datasets, MTServe delivers up to 3.1* speedup while maintaining near-perfect hit ratios (>98.5%).

Ekkasit Pinyoanuntapong, Muhammad Usama Saleem, Korrawe Karunratanakul et al.

Recent advances in motion diffusion models have enabled spatially controllable text-to-motion generation. However, these models struggle to achieve high-precision control while maintaining high-quality motion generation. To address these challenges, we propose MaskControl, the first approach to introduce controllability to the generative masked motion model. Our approach introduces two key innovations. First, \textit{Logits Regularizer} implicitly perturbs logits at training time to align the distribution of motion tokens with the controlled joint positions, while regularizing the categorical token prediction to ensure high-fidelity generation. Second, \textit{Logit Optimization} explicitly optimizes the predicted logits during inference time, directly reshaping the token distribution that forces the generated motion to accurately align with the controlled joint positions. Moreover, we introduce \textit{Differentiable Expectation Sampling (DES)} to combat the non-differential distribution sampling process encountered by logits regularizer and optimization. Extensive experiments demonstrate that MaskControl outperforms state-of-the-art methods, achieving superior motion quality (FID decreases by ~77\%) and higher control precision (average error 0.91 vs. 1.08). Additionally, MaskControl enables diverse applications, including any-joint-any-frame control, body-part timeline control, and zero-shot objective control. Video visualization can be found at https://www.ekkasit.com/ControlMM-page/

Farnoosh Koleini, Muhammad Usama Saleem, Pu Wang et al.

Recent advancements in 3D human pose estimation from single-camera images and videos have relied on parametric models, like SMPL. However, these models oversimplify anatomical structures, limiting their accuracy in capturing true joint locations and movements, which reduces their applicability in biomechanics, healthcare, and robotics. Biomechanically accurate pose estimation, on the other hand, typically requires costly marker-based motion capture systems and optimization techniques in specialized labs. To bridge this gap, we propose BioPose, a novel learning-based framework for predicting biomechanically accurate 3D human pose directly from monocular videos. BioPose includes three key components: a Multi-Query Human Mesh Recovery model (MQ-HMR), a Neural Inverse Kinematics (NeurIK) model, and a 2D-informed pose refinement technique. MQ-HMR leverages a multi-query deformable transformer to extract multi-scale fine-grained image features, enabling precise human mesh recovery. NeurIK treats the mesh vertices as virtual markers, applying a spatial-temporal network to regress biomechanically accurate 3D poses under anatomical constraints. To further improve 3D pose estimations, a 2D-informed refinement step optimizes the query tokens during inference by aligning the 3D structure with 2D pose observations. Experiments on benchmark datasets demonstrate that BioPose significantly outperforms state-of-the-art methods. Project website: \url{https://m-usamasaleem.github.io/publication/BioPose/BioPose.html}.

Arkaprava Sinha, Monish Soundar Raj, Pu Wang et al.

Temporal Action Detection (TAD) in untrimmed videos requires models that can efficiently (1) process long-duration videos, (2) capture temporal variations within action classes, and (3) handle dense, overlapping actions, all while remaining suitable for resource-constrained edge deployment. While Transformer-based methods achieve high accuracy, their quadratic complexity hinders deployment in such scenarios. Given the recent popularity of linear complexity Mamba-based models, leveraging them for TAD is a natural choice. However, naively adapting Mamba from language or vision tasks fails to provide an optimal solution and does not address the challenges of long, untrimmed videos. Therefore, we propose Multi-Scale Temporal Mamba (MS-Temba), the first Mamba-based architecture specifically designed for densely labeled TAD tasks. MS-Temba features Temporal Mamba Blocks (Temba Blocks), consisting of Temporal Convolutional Module (TCM) and Dilated SSM (D-SSM). TCM captures short-term dependencies using dilated convolutions, while D-SSM introduces a novel dilated state-space mechanism to model long-range temporal relationships effectively at each temporal scale. These multi-scale representations are aggregated by Scale-Aware State Fuser, which learns a unified representation for detecting densely overlapping actions. Experiments show that MS-Temba achieves state-of-the-art performance on long-duration videos, remains competitive on shorter segments, and reduces model complexity by 88%. Its efficiency and effectiveness make MS-Temba well-suited for real-world edge deployment.

Raju Challagundla, Mohsen Dorodchi, Pu Wang et al.

As privacy regulations become more stringent and access to real-world data becomes increasingly constrained, synthetic data generation has emerged as a vital solution, especially for tabular datasets, which are central to domains like finance, healthcare and the social sciences. This survey presents a comprehensive and focused review of recent advances in synthetic tabular data generation, emphasizing methods that preserve complex feature relationships, maintain statistical fidelity, and satisfy privacy requirements. A key contribution of this work is the introduction of a novel taxonomy based on practical generation objectives, including intended downstream applications, privacy guarantees, and data utility, directly informing methodological design and evaluation strategies. Therefore, this review prioritizes the actionable goals that drive synthetic data creation, including conditional generation and risk-sensitive modeling. Additionally, the survey proposes a benchmark framework to align technical innovation with real-world demands. By bridging theoretical foundations with practical deployment, this work serves as both a roadmap for future research and a guide for implementing synthetic tabular data in privacy-critical environments.

Yuan Gao, Shaobo Xia, Pu Wang et al.

Light detection and ranging (LiDAR) remote sensing encompasses two major directions: data interpretation and parameter inversion. However, both directions rely heavily on costly and labor-intensive labeled data and field measurements, which constrains their scalability and spatiotemporal adaptability. Weakly Supervised Learning (WSL) provides a unified framework to address these limitations. This paper departs from the traditional view that treats interpretation and inversion as separate tasks and offers a systematic review of recent advances in LiDAR remote sensing from a unified WSL perspective. We cover typical WSL settings including incomplete supervision(e.g., sparse point labels), inexact supervision (e.g., scene-level tags), inaccurate supervision (e.g., noisy labels), and cross-domain supervision (e.g., domain adaptation/generalization) and corresponding techniques such as pseudo-labeling, consistency regularization, self-training, and label refinement, which collectively enable robust learning from limited and weak annotations.We further analyze LiDAR-specific challenges (e.g., irregular geometry, data sparsity, domain heterogeneity) that require tailored weak supervision, and examine how sparse LiDAR observations can guide joint learning with other remote-sensing data for continuous surface-parameter retrieval. Finally, we highlight future directions where WSL acts as a bridge between LiDAR and foundation models to leverage large-scale multimodal datasets and reduce labeling costs, while also enabling broader WSL-driven advances in generalization, open-world adaptation, and scalable LiDAR remote sensing.

Foram N Shah, Parshwa Shah, Muhammad Usama Saleem et al.

Recent advances in dance generation have enabled the automatic synthesis of 3D dance motions. However, existing methods still face significant challenges in simultaneously achieving high realism, precise dance-music synchronization, diverse motion expression, and physical plausibility. To address these limitations, we propose a novel approach that leverages a generative masked text-to-motion model as a distribution prior to learn a probabilistic mapping from diverse guidance signals, including music, genre, and pose, into high-quality dance motion sequences. Our framework also supports semantic motion editing, such as motion inpainting and body part modification. Specifically, we introduce a multi-tower masked motion model that integrates a text-conditioned masked motion backbone with two parallel, modality-specific branches: a music-guidance tower and a pose-guidance tower. The model is trained using synchronized and progressive masked training, which allows effective infusion of the pretrained text-to-motion prior into the dance synthesis process while enabling each guidance branch to optimize independently through its own loss function, mitigating gradient interference. During inference, we introduce classifier-free logits guidance and pose-guided token optimization to strengthen the influence of music, genre, and pose signals. Extensive experiments demonstrate that our method sets a new state of the art in dance generation, significantly advancing both the quality and editability over existing approaches. Project Page available at https://foram-s1.github.io/DanceMosaic/

Arkaprava Sinha, Dominick Reilly, Francois Bremond et al.

The introduction of vision-language models like CLIP has enabled the development of foundational video models capable of generalizing to unseen videos and human actions. However, these models are typically trained on web videos, which often fail to capture the challenges present in Activities of Daily Living (ADL) videos. Existing works address ADL-specific challenges, such as similar appearances, subtle motion patterns, and multiple viewpoints, by combining 3D skeletons and RGB videos. However, these approaches are not integrated with language, limiting their ability to generalize to unseen action classes. In this paper, we introduce SKI models, which integrate 3D skeletons into the vision-language embedding space. SKI models leverage a skeleton-language model, SkeletonCLIP, to infuse skeleton information into Vision Language Models (VLMs) and Large Vision Language Models (LVLMs) through collaborative training. Notably, SKI models do not require skeleton data during inference, enhancing their robustness for real-world applications. The effectiveness of SKI models is validated on three popular ADL datasets for zero-shot action recognition and video caption generation tasks.

Min Lu, Yuanfeng He, Anthony Chen et al.

Artistic styles often embed abstraction beyond surface appearance, involving deliberate reinterpretation of structure rather than mere changes in texture or color. Conventional style transfer methods typically preserve the input geometry and therefore struggle to capture this deeper abstraction behavior, especially for illustrative and nonphotorealistic styles. In this work, we introduce Abstraction in Style (AiS), a generative framework that separates structural abstraction from visual stylization. Given a target image and a small set of style exemplars, AiS first derives an intermediate abstraction proxy that reinterprets the target's structure in accordance with the abstraction logic exhibited by the style. The proxy captures semantic structure while relaxing geometric fidelity, enabling subsequent stylization to operate on an abstracted representation rather than the original image. In a second stage, the abstraction proxy is rendered to produce the final stylized output, preserving visual coherence with the reference style. Both stages are implemented using a shared image space analogy, enabling transformations to be learned from visual exemplars without explicit geometric supervision. By decoupling abstraction from appearance and treating abstraction as an explicit, transferable process, AiS supports a wider range of stylistic transformations, improves controllability, and enables more expressive stylization.

Farnoosh Koleini, Hongfei Xue, Ahmed Helmy et al.

Reconstructing biomechanically realistic 3D human motion - recovering both kinematics (motion) and kinetics (forces) - is a critical challenge. While marker-based systems are lab-bound and slow, popular monocular methods use oversimplified, anatomically inaccurate models (e.g., SMPL) and ignore physics, fundamentally limiting their biomechanical fidelity. In this work, we introduce MonoMSK, a hybrid framework that bridges data-driven learning and physics-based simulation for biomechanically realistic 3D human motion estimation from monocular video. MonoMSK jointly recovers both kinematics (motions) and kinetics (forces and torques) through an anatomically accurate musculoskeletal model. By integrating transformer-based inverse dynamics with differentiable forward kinematics and dynamics layers governed by ODE-based simulation, MonoMSK establishes a physics-regulated inverse-forward loop that enforces biomechanical causality and physical plausibility. A novel forward-inverse consistency loss further aligns motion reconstruction with the underlying kinetic reasoning. Experiments on BML-MoVi, BEDLAM, and OpenCap show that MonoMSK significantly outperforms state-of-the-art methods in kinematic accuracy, while for the first time enabling precise monocular kinetics estimation.

Guanlin Wu, Boyan Su, Yang Zhao et al.

How to integrate and verify spatial intelligence in foundation models remains an open challenge. Current practice often proxies Visual-Spatial Intelligence (VSI) with purely textual prompts and VQA-style scoring, which obscures geometry, invites linguistic shortcuts, and weakens attribution to genuinely spatial skills. We introduce Spatial Intelligence Grid (SIG): a structured, grid-based schema that explicitly encodes object layouts, inter-object relations, and physically grounded priors. As a complementary channel to text, SIG provides a faithful, compositional representation of scene structure for foundation-model reasoning. Building on SIG, we derive SIG-informed evaluation metrics that quantify a model's intrinsic VSI, which separates spatial capability from language priors. In few-shot in-context learning with state-of-the-art multimodal LLMs (e.g. GPT- and Gemini-family models), SIG yields consistently larger, more stable, and more comprehensive gains across all VSI metrics compared to VQA-only representations, indicating its promise as a data-labeling and training schema for learning VSI. We also release SIGBench, a benchmark of 1.4K driving frames annotated with ground-truth SIG labels and human gaze traces, supporting both grid-based machine VSI tasks and attention-driven, human-like VSI tasks in autonomous-driving scenarios.

Yang Zhao, Pu Wang, Hao Frank Yang

Designing optimal prompts and reasoning processes for large language models (LLMs) on domain-specific tasks is both necessary and challenging in real-world applications. Determining how to integrate domain knowledge, enhance reasoning efficiency, and even provide domain experts with refined knowledge integration hints are particularly crucial yet unresolved tasks. In this research, we propose Evolutionary Graph Optimization for Prompting (EGO-Prompt), an automated framework to designing better prompts, efficient reasoning processes and providing enhanced causal-informed process. EGO-Prompt begins with a general prompt and fault-tolerant initial Semantic Causal Graph (SCG) descriptions, constructed by human experts, which is then automatically refined and optimized to guide LLM reasoning. Recognizing that expert-defined SCGs may be partial or imperfect and that their optimal integration varies across LLMs, EGO-Prompt integrates a novel causal-guided textual gradient process in two steps: first, generating nearly deterministic reasoning guidance from the SCG for each instance, and second, adapting the LLM to effectively utilize the guidance alongside the original input. The iterative optimization algorithm further refines both the SCG and the reasoning mechanism using textual gradients with ground-truth. We tested the framework on real-world public health, transportation and human behavior tasks. EGO-Prompt achieves 7.32%-12.61% higher F1 than cutting-edge methods, and allows small models to reach the performence of larger models at under 20% of the original cost. It also outputs a refined, domain-specific SCG that improves interpretability.

Ziyan Wang, Enmao Diao, Qi Le et al.

Structured pruning is a practical approach to deploying large language models (LLMs) efficiently, as it yields compact, hardware-friendly architectures. However, the dominant local paradigm is task-agnostic: by optimizing layer-wise reconstruction rather than task objectives, it tends to preserve perplexity or generic zero-shot behavior but fails to capitalize on modest task-specific calibration signals, often yielding limited downstream gains. We revisit global structured pruning and present GISP-Global Iterative Structured Pruning-a post-training method that removes attention heads and MLP channels using first-order, loss-based important weights aggregated at the structure level with block-wise normalization. An iterative schedule, rather than one-shot pruning, stabilizes accuracy at higher sparsity and mitigates perplexity collapse without requiring intermediate fine-tuning; the pruning trajectory also forms nested subnetworks that support a "prune-once, deploy-many" workflow. Furthermore, because importance is defined by a model-level loss, GISP naturally supports task-specific objectives; we instantiate perplexity for language modeling and a margin-based objective for decision-style tasks. Extensive experiments show that across Llama2-7B/13B, Llama3-8B, and Mistral-0.3-7B, GISP consistently lowers WikiText-2 perplexity and improves downstream accuracy, with especially strong gains at 40-50% sparsity; on DeepSeek-R1-Distill-Llama-3-8B with GSM8K, task-aligned calibration substantially boosts exact-match accuracy.