The recent rise of generative artificial intelligence (AI), powered by Transformer networks, has achieved remarkable success in natural language processing, computer vision, and graphics. However, the application of Transformers in computer-aided design (CAD), particularly for processing boundary representation (B-rep) models, remains largely unexplored. To bridge this gap, this paper introduces Boundary Representation Transformer (BRT), a novel method adapting Transformer for B-rep learning. B-rep models pose unique challenges due to their irregular topology and continuous geometric definitions, which are fundamentally different from the structured and discrete data Transformers are designed for. To address this, BRT proposes a continuous geometric embedding method that encodes B-rep surfaces (trimmed and untrimmed) into B\'ezier triangles, preserving their shape and continuity without discretization. Additionally, BRT employs a topology-aware embedding method that organizes these geometric embeddings into a sequence of discrete tokens suitable for Transformers, capturing both geometric and topological characteristics within B-rep models. This enables the Transformer's attention mechanism to effectively learn shape patterns and contextual semantics of boundary elements in a B-rep model. Extensive experiments demonstrate that BRT achieves state-of-the-art performance in part classification and feature recognition tasks.
Visual emotion analysis or recognition has gained considerable attention due to the growing interest in understanding how images can convey rich semantics and evoke emotions in human perception. However, visual emotion analysis poses distinctive challenges compared to traditional vision tasks, especially due to the intricate relationship between general visual features and the different affective states they evoke, known as the affective gap. Researchers have used deep representation learning methods to address this challenge of extracting generalized features from entire images. However, most existing methods overlook the importance of specific emotional attributes such as brightness, colorfulness, scene understanding, and facial expressions. Through this paper, we introduce A4Net, a deep representation network to bridge the affective gap by leveraging four key attributes: brightness (Attribute 1), colorfulness (Attribute 2), scene context (Attribute 3), and facial expressions (Attribute 4). By fusing and jointly training all aspects of attribute recognition and visual emotion analysis, A4Net aims to provide a better insight into emotional content in images. Experimental results show the effectiveness of A4Net, showcasing competitive performance compared to state-of-the-art methods across diverse visual emotion datasets. Furthermore, visualizations of activation maps generated by A4Net offer insights into its ability to generalize across different visual emotion datasets.
Watermarking has emerged as a promising technique for detecting texts generated by LLMs. Current research has primarily focused on three design criteria: high quality of the watermarked text, high detectability, and robustness against removal attack. However, the security against spoofing attacks remains relatively understudied. For example, a piggyback attack can maliciously alter the meaning of watermarked text-transforming it into hate speech-while preserving the original watermark, thereby damaging the reputation of the LLM provider. We identify two core challenges that make defending against spoofing difficult: (1) the need for watermarks to be both sensitive to semantic-distorting changes and insensitive to semantic-preserving edits, and (2) the contradiction between the need to detect global semantic shifts and the local, auto-regressive nature of most watermarking schemes. To address these challenges, we propose a semantic-aware watermarking algorithm that post-hoc embeds watermarks into a given target text while preserving its original meaning. Our method introduces a semantic mapping model, which guides the generation of a green-red token list, contrastively trained to be sensitive to semantic-distorting changes and insensitive to semantic-preserving changes. Experiments on two standard benchmarks demonstrate strong robustness against removal attacks and security against spoofing attacks, including sentiment reversal and toxic content insertion, while maintaining high watermark detectability. Our approach offers a significant step toward more secure and semantically aware watermarking for LLMs. Our code is available at https://github.com/UCSB-NLP-Chang/contrastive-watermark.
Postoperative delirium (POD), a severe neuropsychiatric complication affecting nearly 50% of high-risk surgical patients, is defined as an acute disorder of attention and cognition, It remains significantly underdiagnosed in the intensive care units (ICUs) due to subjective monitoring methods. Early and accurate diagnosis of POD is critical and achievable. Here, we propose a POD prediction framework comprising a Transformer representation model followed by traditional machine learning algorithms. Our approaches utilizes multi-modal physiological data, including amplitude-integrated electroencephalography (aEEG), vital signs, electrocardiographic monitor data as well as hemodynamic parameters. We curated the first multi-modal POD dataset encompassing two patient types and evaluated the various Transformer architectures for representation learning. Empirical results indicate a consistent improvements of sensitivity and Youden index in patient TYPE I using Transformer representations, particularly our fusion adaptation of Pathformer. By enabling effective delirium diagnosis from postoperative day 1 to 3, our extensive experimental findings emphasize the potential of multi-modal physiological data and highlight the necessity of representation learning via multi-modal Transformer architecture in clinical diagnosis.
The rapid advancement of foundation models has revolutionized visual representation learning in a self-supervised manner. However, their application in remote sensing (RS) remains constrained by a fundamental gap: existing models predominantly handle single or limited modalities, overlooking the inherently multi-modal nature of RS observations. Optical, synthetic aperture radar (SAR), and multi-spectral data offer complementary insights that significantly reduce the inherent ambiguity and uncertainty in single-source analysis. To bridge this gap, we introduce RingMoE, a unified multi-modal RS foundation model with 14.7 billion parameters, pre-trained on 400 million multi-modal RS images from nine satellites. RingMoE incorporates three key innovations: (1) A hierarchical Mixture-of-Experts (MoE) architecture comprising modal-specialized, collaborative, and shared experts, effectively modeling intra-modal knowledge while capturing cross-modal dependencies to mitigate conflicts between modal representations; (2) Physics-informed self-supervised learning, explicitly embedding sensor-specific radiometric characteristics into the pre-training objectives; (3) Dynamic expert pruning, enabling adaptive model compression from 14.7B to 1B parameters while maintaining performance, facilitating efficient deployment in Earth observation applications. Evaluated across 23 benchmarks spanning six key RS tasks (i.e., classification, detection, segmentation, tracking, change detection, and depth estimation), RingMoE outperforms existing foundation models and sets new SOTAs, demonstrating remarkable adaptability from single-modal to multi-modal scenarios. Beyond theoretical progress, it has been deployed and trialed in multiple sectors, including emergency response, land management, marine sciences, and urban planning.
Existing programming knowledge tracing methods have not deeply explored the relationships between knowledge concepts and programming questions or students’ codes, which leads to insufficient representation of programming questions and students’ codes and insufficient tracing of students’ knowledge states. To address these problems, this study proposes a programming knowledge tracing method based on a heterogeneous graph representation. We constructed a heterogeneous graph, where nodes represent the knowledge concept, programming question, and student code, and edges represent the relationships among knowledge concepts, programming questions, and student code. Subsequently, we leverage a graph structure learning algorithm to exploit the potential connections among them, aiming to enhance the graph structure information. Based on the graph, we further fully integrated the knowledge concept information into the representations of programming questions and student codes through the graph convolutional representation learning algorithm. Moreover, an attention mechanism is utilized to compute the effect of programming questions and student codes on the knowledge state and update the student’s knowledge state in conjunction with the corresponding answer results. Furthermore, the proposed method provides an assessment of the mastery level for each knowledge concept rather than only assessing each question using traditional methods. Experiments on two datasets, CodeWorkout and CodeNet, show that the proposed method achieves an average improvement of 4.28% in ACC and 6.48% in AUC compared with the baseline models, validating the effectiveness of the proposed method.
Attribute skew in federated learning leads local models to focus on learning non-causal associations, guiding them towards inconsistent optimization directions, which inevitably results in performance degradation and unstable convergence. Existing methods typically leverage data augmentation to enhance sample diversity or employ knowledge distillation to learn invariant representations. However, the instability in the quality of generated data and the lack of domain information limit their performance on unseen samples. To address these issues, this paper presents a global intervention and distillation method, termed FedGID, which utilizes diverse attribute features for backdoor adjustment to break the spurious association between background and label. It includes two main modules, where the global intervention module adaptively decouples objects and backgrounds in images, injects background information into random samples to intervene in the sample distribution, which links backgrounds to all categories to prevent the model from treating background-label associations as causal. The global distillation module leverages a unified knowledge base to guide the representation learning of client models, preventing local models from overfitting to client-specific attributes. Experimental results on three datasets demonstrate that FedGID enhances the model's ability to focus on the main subjects in unseen data and outperforms existing methods in collaborative modeling.
Self-supervised learning (SSL) has revolutionized representation learning in Remote Sensing (RS), advancing Geospatial Foundation Models (GFMs) to leverage vast unlabeled satellite imagery for diverse downstream tasks. Currently, GFMs primarily focus on discriminative objectives, such as contrastive learning or masked image modeling, owing to their proven success in learning transferable representations. However, generative diffusion models--which demonstrate the potential to capture multi-grained semantics essential for RS tasks during image generation--remain underexplored for discriminative applications. This prompts the question: can generative diffusion models also excel and serve as GFMs with sufficient discriminative power? In this work, we answer this question with SatDiFuser, a framework that transforms a diffusion-based generative geospatial foundation model into a powerful pretraining tool for discriminative RS. By systematically analyzing multi-stage, noise-dependent diffusion features, we develop three fusion strategies to effectively leverage these diverse representations. Extensive experiments on remote sensing benchmarks show that SatDiFuser outperforms state-of-the-art GFMs, achieving gains of up to +5.7% mIoU in semantic segmentation and +7.9% F1-score in classification, demonstrating the capacity of diffusion-based generative foundation models to rival or exceed discriminative GFMs. Code will be released.
In this study, Disentanglement in Difference(DiD) is proposed to address the inherent inconsistency between the statistical independence of latent variables and the goal of semantic disentanglement in disentanglement representation learning. Conventional disentanglement methods achieve disentanglement representation by improving statistical independence among latent variables. However, the statistical independence of latent variables does not necessarily imply that they are semantically unrelated, thus, improving statistical independence does not always enhance disentanglement performance. To address the above issue, DiD is proposed to directly learn semantic differences rather than the statistical independence of latent variables. In the DiD, a Difference Encoder is designed to measure the semantic differences; a contrastive loss function is established to facilitate inter-dimensional comparison. Both of them allow the model to directly differentiate and disentangle distinct semantic factors, thereby resolving the inconsistency between statistical independence and semantic disentanglement. Experimental results on the dSprites and 3DShapes datasets demonstrate that the proposed DiD outperforms existing mainstream methods across various disentanglement metrics.
Video-text retrieval, the task of retrieving videos based on a textual query or vice versa, is of paramount importance for video understanding and multimodal information retrieval. Recent methods in this area rely primarily on visual and textual features and often ignore audio, although it helps enhance overall comprehension of video content. Moreover, traditional models that incorporate audio blindly utilize the audio input regardless of whether it is useful or not, resulting in suboptimal video representation. To address these limitations, we propose a novel video-text retrieval framework, Audio-guided VIdeo representation learning with GATEd attention (AVIGATE), that effectively leverages audio cues through a gated attention mechanism that selectively filters out uninformative audio signals. In addition, we propose an adaptive margin-based contrastive loss to deal with the inherently unclear positive-negative relationship between video and text, which facilitates learning better video-text alignment. Our extensive experiments demonstrate that AVIGATE achieves state-of-the-art performance on all the public benchmarks.
Visual Language Models (VLMs) have emerged as pivotal tools for robotic systems, enabling cross-task generalization, dynamic environmental interaction, and long-horizon planning through multimodal perception and semantic reasoning. However, existing open-source VLMs predominantly trained for generic vision-language alignment tasks fail to model temporally correlated action semantics that are crucial for robotic manipulation effectively. While current image-based fine-tuning methods partially adapt VLMs to robotic applications, they fundamentally disregard temporal evolution patterns in video sequences and suffer from visual feature entanglement between robotic agents, manipulated objects, and environmental contexts, thereby limiting semantic decoupling capability for atomic actions and compromising model generalizability.To overcome these challenges, this work presents RoboAct-CLIP with dual technical contributions: 1) A dataset reconstruction framework that performs semantic-constrained action unit segmentation and re-annotation on open-source robotic videos, constructing purified training sets containing singular atomic actions (e.g., "grasp"); 2) A temporal-decoupling fine-tuning strategy based on Contrastive Language-Image Pretraining (CLIP) architecture, which disentangles temporal action features across video frames from object-centric characteristics to achieve hierarchical representation learning of robotic atomic actions.Experimental results in simulated environments demonstrate that the RoboAct-CLIP pretrained model achieves a 12% higher success rate than baseline VLMs, along with superior generalization in multi-object manipulation tasks.
We address the challenge of representation learning from a continuous stream of video as input, in a self-supervised manner. This differs from the standard approaches to video learning where videos are chopped and shuffled during training in order to create a non-redundant batch that satisfies the independently and identically distributed (IID) sample assumption expected by conventional training paradigms. When videos are only available as a continuous stream of input, the IID assumption is evidently broken, leading to poor performance. We demonstrate the drop in performance when moving from shuffled to sequential learning on three tasks: the one-video representation learning method DoRA, standard VideoMAE on multi-video datasets, and the task of future video prediction. To address this drop, we propose a geometric modification to standard optimizers, to decorrelate batches by utilising orthogonal gradients during training. The proposed modification can be applied to any optimizer -- we demonstrate it with Stochastic Gradient Descent (SGD) and AdamW. Our proposed orthogonal optimizer allows models trained from streaming videos to alleviate the drop in representation learning performance, as evaluated on downstream tasks. On three scenarios (DoRA, VideoMAE, future prediction), we show our orthogonal optimizer outperforms the strong AdamW in all three scenarios.
This paper introduces a novel approach to monocular 3D human pose estimation using contextualized representation learning with the Transformer-GCN dual-stream model. Monocular 3D human pose estimation is challenged by depth ambiguity, limited 3D-labeled training data, imbalanced modeling, and restricted model generalization. To address these limitations, our work introduces a groundbreaking motion pre-training method based on contextualized representation learning. Specifically, our method involves masking 2D pose features and utilizing a Transformer-GCN dual-stream model to learn high-dimensional representations through a self-distillation setup. By focusing on contextualized representation learning and spatial-temporal modeling, our approach enhances the model's ability to understand spatial-temporal relationships between postures, resulting in superior generalization. Furthermore, leveraging the Transformer-GCN dual-stream model, our approach effectively balances global and local interactions in video pose estimation. The model adaptively integrates information from both the Transformer and GCN streams, where the GCN stream effectively learns local relationships between adjacent key points and frames, while the Transformer stream captures comprehensive global spatial and temporal features. Our model achieves state-of-the-art performance on two benchmark datasets, with an MPJPE of 38.0mm and P-MPJPE of 31.9mm on Human3.6M, and an MPJPE of 15.9mm on MPI-INF-3DHP. Furthermore, visual experiments on public datasets and in-the-wild videos demonstrate the robustness and generalization capabilities of our approach.
Understanding the complex myocardial architecture is critical for diagnosing and treating heart disease. However, existing methods often struggle to accurately capture this intricate structure from Diffusion Tensor Imaging (DTI) data, particularly due to the lack of ground truth labels and the ambiguous, intertwined nature of fiber trajectories. We present a novel deep learning framework for unsupervised clustering of myocardial fibers, providing a data-driven approach to identifying distinct fiber bundles. We uniquely combine a Bidirectional Long Short-Term Memory network to capture local sequential information along fibers, with a Transformer autoencoder to learn global shape features, with pointwise incorporation of essential anatomical context. Clustering these representations using a density-based algorithm identifies 33 to 62 robust clusters, successfully capturing the subtle distinctions in fiber trajectories with varying levels of granularity. Our framework offers a new, flexible, and quantitative way to analyze myocardial structure, achieving a level of delineation that, to our knowledge, has not been previously achieved, with potential applications in improving surgical planning, characterizing disease-related remodeling, and ultimately, advancing personalized cardiac care.
Most model-based 3D hand pose and shape estimation methods directly regress the parametric model parameters from an image to obtain 3D joints under weak supervision. However, these methods involve solving a complex optimization problem with many local minima, making training difficult. To address this challenge, we propose learning direction-aware hybrid features (DaHyF) that fuse implicit image features and explicit 2D joint coordinate features. This fusion is enhanced by the pixel direction information in the camera coordinate system to estimate pose, shape, and camera viewpoint. Our method directly predicts 3D hand poses with DaHyF representation and reduces jittering during motion capture using prediction confidence based on contrastive learning. We evaluate our method on the FreiHAND dataset and show that it outperforms existing state-of-the-art methods by more than 33% in accuracy. DaHyF also achieves the top ranking on both the HO3Dv2 and HO3Dv3 leaderboards for the metric of Mean Joint Error (after scale and translation alignment). Compared to the second-best results, the largest improvement observed is 10%. We also demonstrate its effectiveness in real-time motion capture scenarios with hand position variability, occlusion, and motion blur.
The main objective of fair statistical modeling and machine learning is to minimize or eliminate biases that may arise from the data or the model itself, ensuring that predictions and decisions are not unjustly influenced by sensitive attributes such as race, gender, age, or other protected characteristics. In this paper, we introduce a Fair Sufficient Representation Learning (FSRL) method that balances sufficiency and fairness. Sufficiency ensures that the representation should capture all necessary information about the target variables, while fairness requires that the learned representation remains independent of sensitive attributes. FSRL is based on a convex combination of an objective function for learning a sufficient representation and an objective function that ensures fairness. Our approach manages fairness and sufficiency at the representation level, offering a novel perspective on fair representation learning. We implement this method using distance covariance, which is effective for characterizing independence between random variables. We establish the convergence properties of the learned representations. Experiments conducted on healthcase and text datasets with diverse structures demonstrate that FSRL achieves a superior trade-off between fairness and accuracy compared to existing approaches.
Reproducing realistic collective behaviors presents a captivating yet formidable challenge. Traditional rule-based methods rely on hand-crafted principles, limiting motion diversity and realism in generated collective behaviors. Recent imitation learning methods learn from data but often require ground truth motion trajectories and struggle with authenticity, especially in high-density groups with erratic movements. In this paper, we present a scalable approach, Collective Behavior Imitation Learning (CBIL), for learning fish schooling behavior directly from videos, without relying on captured motion trajectories. Our method first leverages Video Representation Learning, where a Masked Video AutoEncoder (MVAE) extracts implicit states from video inputs in a self-supervised manner. The MVAE effectively maps 2D observations to implicit states that are compact and expressive for following the imitation learning stage. Then, we propose a novel adversarial imitation learning method to effectively capture complex movements of the schools of fish, allowing for efficient imitation of the distribution for motion patterns measured in the latent space. It also incorporates bio-inspired rewards alongside priors to regularize and stabilize training. Once trained, CBIL can be used for various animation tasks with the learned collective motion priors. We further show its effectiveness across different species. Finally, we demonstrate the application of our system in detecting abnormal fish behavior from in-the-wild videos.
Current video-based Masked Autoencoders (MAEs) primarily learn general spatial-temporal patterns from a visual perspective but often overlook nuanced semantic attributes like specific interactions or sequences that define actions align more closely with human cognition for space-time correspondence. This can limit the model’s ability to capture the essence of certain actions that are contextually rich and continuous. Humans can map visual concepts, object view invariance, and semantic attributes available in static instances to comprehend natural dynamic scenes or videos. Existing MAEs for videos and static images rely on frame-by-frame continuity or separate datasets for videos and images, which may lack the rich semantic attributes necessary for fully understanding the learned concepts, especially when compared to using video and corresponding sampled frame images together. To this end, we propose CrossVideoMAE, a self-supervised image-video contrastive MAE pre-training framework that effectively learns both video-level and frame-level spatiotemporal representations and semantic attributes. Our method integrates mutual spatiotemporal information from videos with spatial information from sampled frames within a feature-invariant space while encouraging invariance to augmentations within the video domain. This objective is achieved through jointly embedding features of visible tokens and combining feature correspondence within and across modalities, which is critical for acquiring rich, label-free guiding signals from both video and frame image modalities in a self-supervised manner. Extensive experiments demonstrate that our approach surpasses previous SOTA methods and ablation studies validate the effectiveness of our approach.
While graph neural networks (GNNs) operating in hyperbolic spaces have shown promise for modeling hierarchical and complex relational data, a critical limitation often overlooked is their potentially limited discriminative power compared to their Euclidean counterparts or fundamental graph isomorphism tests like the Weisfeiler-Lehman (WL) hierarchy. Existing hyperbolic aggregation schemes, while curvature-aware, may not sufficiently capture the intricate structural differences required to robustly distinguish non-isomorphic graphs owing to non-injective aggregation functions. To address this expressiveness gap in hyperbolic graph learning, we introduce the Lorentzian Graph Isomorphic Network (LGIN), a novel GNN designed to achieve enhanced discriminative capabilities within the Lorentzian model of hyperbolic space. LGIN proposes a new update rule that effectively combines local neighborhood information with a richer representation of graph structure designed to preserve the Lorentzian metric tensor. This represents a significant step towards building more expressive GNNs in non-Euclidean geometries, overcoming a common bottleneck in current hyperbolic methods. We conduct extensive evaluations across nine diverse benchmark datasets, including molecular and protein structures. LGIN consistently outperforms or matches state-of-the-art hyperbolic and Euclidean GNNs, showcasing its practical efficacy and validating its superior ability to capture complex graph structures and distinguish between different graphs. To the best of our knowledge, LGIN is the first work to study the framework behind a powerful GNN on the hyperbolic space. The code for our paper can be found at https://github.com/Deceptrax123/LGIN
Expressive representation of pose sequences is crucial for accurate motion modeling in human motion prediction (HMP). While recent deep learning-based methods have shown promise in learning motion representations, these methods tend to overlook the varying relevance and dependencies between historical information and future moments, with a stronger correlation for short-term predictions and weaker for distant future predictions. This limits the learning of motion representation and then hampers prediction performance. In this paper, we propose a novel approach called multi-range decoupling decoding with gating-adjusting aggregation ($MD2GA$), which leverages the temporal correlations to refine motion representation learning. This approach employs a two-stage strategy for HMP. In the first stage, a multi-range decoupling decoding adeptly adjusts feature learning by decoding the shared features into distinct future lengths, where different decoders offer diverse insights into motion patterns. In the second stage, a gating-adjusting aggregation dynamically combines the diverse insights guided by input motion data. Extensive experiments demonstrate that the proposed method can be easily integrated into other motion prediction methods and enhance their prediction performance.