We propose the Compact Clustering Attention (COCA) layer, an effective building block that introduces a hierarchical strategy for object-centric representation learning, while solving the unsupervised object discovery task on single images. COCA is an attention-based clustering module capable of extracting object-centric representations from multi-object scenes, when cascaded into a bottom-up hierarchical network architecture, referred to as COCA-Net. At its core, COCA utilizes a novel clustering algorithm that leverages the physical concept of compactness, to highlight distinct object centroids in a scene, providing a spatial inductive bias. Thanks to this strategy, COCA-Net generates high-quality segmentation masks on both the decoder side and, notably, the encoder side of its pipeline. Additionally, COCA-Net is not bound by a predetermined number of object masks that it generates and handles the segmentation of background elements better than its competitors. We demonstrate COCA-Net's segmentation performance on six widely adopted datasets, achieving superior or competitive results against the state-of-the-art models across nine different evaluation metrics.
High-quality fundus images provide essential anatomical information for clinical screening and ophthalmic disease diagnosis. Yet, due to hardware limitations, operational variability, and patient compliance, fundus images often suffer from low resolution and signal-to-noise ratio. Recent years have witnessed promising progress in fundus image enhancement. However, existing works usually focus on restoring structural details or global characteristics of fundus images, lacking a unified image enhancement framework to recover comprehensive multi-scale information. Moreover, few methods pinpoint the target of image enhancement, e.g., lesions, which is crucial for medical image-based diagnosis. To address these challenges, we propose a multi-scale target-aware representation learning framework (MTRL-FIE) for efficient fundus image enhancement. Specifically, we propose a multi-scale feature encoder (MFE) that employs wavelet decomposition to embed both low-frequency structural information and high-frequency details. Next, we design a structure-preserving hierarchical decoder (SHD) to fuse multi-scale feature embeddings for real fundus image restoration. SHD integrates hierarchical fusion and group attention mechanisms to achieve adaptive feature fusion while retaining local structural smoothness. Meanwhile, a target-aware feature aggregation (TFA) module is used to enhance pathological regions and reduce artifacts. Experimental results on multiple fundus image datasets demonstrate the effectiveness and generalizability of MTRL-FIE for fundus image enhancement. Compared to state-of-the-art methods, MTRL-FIE achieves superior enhancement performance with a more lightweight architecture. Furthermore, our approach generalizes to other ophthalmic image processing tasks without supervised fine-tuning, highlighting its potential for clinical applications.
Personalized 3D avatar editing holds significant promise due to its user-friendliness and availability to applications such as AR/VR and virtual try-ons. Previous studies have explored the feasibility of 3D editing, but often struggle to generate visually pleasing results, possibly due to the unstable representation learning under mixed optimization of geometry and texture in complicated reconstructed scenarios. In this paper, we aim to provide an accessible solution for ordinary users to create their editable 3D avatars with precise region localization, geometric adaptability, and photorealistic renderings. To tackle this challenge, we introduce a meticulously designed framework that decouples the editing process into local spatial adaptation and realistic appearance learning, utilizing a hybrid Tetrahedron-constrained Gaussian Splatting (TetGS) as the underlying representation. TetGS combines the controllable explicit structure of tetrahedral grids with the high-precision rendering capabilities of 3D Gaussian Splatting and is optimized in a progressive manner comprising three stages: 3D avatar instantiation from real-world monocular videos to provide accurate priors for TetGS initialization; localized spatial adaptation with explicitly partitioned tetrahedrons to guide the redistribution of Gaussian kernels; and geometry-based appearance generation with a coarse-to-fine activation strategy. Both qualitative and quantitative experiments demonstrate the effectiveness and superiority of our approach in generating photorealistic 3D editable avatars.
The inherent synchronization between a speaker's lip movements, voice, and the underlying linguistic content offers a rich source of information for improving speech processing tasks, especially in challenging conditions where traditional audio-only systems falter. We introduce CoGenAV, a powerful and data-efficient model designed to learn versatile audio-visual representations applicable across a wide range of speech and audio-visual tasks. CoGenAV is trained by optimizing a dual objective derived from natural audio-visual synchrony, contrastive feature alignment and generative text prediction, using only 223 hours of labeled data from the LRS2 dataset. This contrastive-generative synchronization strategy effectively captures fundamental cross-modal correlations. We showcase the effectiveness and versatility of the learned CoGenAV representations on multiple benchmarks. When utilized for Audio-Visual Speech Recognition (AVSR) on LRS2, these representations contribute to achieving a state-of-the-art Word Error Rate (WER) of 1.27. They also enable strong performance in Visual Speech Recognition (VSR) with a WER of 20.5 on LRS2, and significantly improve performance in noisy environments by over 70%. Furthermore, CoGenAV representations benefit speech reconstruction tasks, boosting performance in Speech Enhancement and Separation, and achieve competitive results in audio-visual synchronization tasks like Active Speaker Detection (ASD). Our model will be open-sourced to facilitate further development and collaboration within both academia and industry.
Vision-language models (VLMs) allow to embed texts and images in a shared representation space. However, it has been shown that these models are subject to a modality gap phenomenon meaning there exists a clear separation between the embeddings from one modality and another in the embedding space. While this misalignment is detrimental for downstream tasks such as multimodal retrieval, multimodal clustering or zero-shot classification, etc. no generic and practical methods have so far been proposed to assess it precisely and even reduce it. We therefore propose novel measures and effective techniques (spectral- and optimal transport-based methods) to achieve this goal. Extensive experiments conducted on several image-text datasets and models demonstrate their effectiveness and beneficial effects on downstream tasks. Our code is available at the URL provided in the paper's abstract.
Despite the empirical success of foundation models, we do not have a systematic characterization of the representations that these models learn. In this paper, we establish the contexture theory. It shows that a large class of representation learning methods can be characterized as learning from the association between the input and a context variable. Specifically, we show that many popular methods aim to approximate the top-d singular functions of the expectation operator induced by the context, in which case we say that the representation learns the contexture. We demonstrate the generality of the contexture theory by proving that representation learning within various learning paradigms -- supervised, self-supervised, and manifold learning -- can all be studied from such a perspective. We also prove that the representations that learn the contexture are optimal on those tasks that are compatible with the context. One important implication of the contexture theory is that once the model is large enough to approximate the top singular functions, further scaling up the model size yields diminishing returns. Therefore, scaling is not all we need, and further improvement requires better contexts. To this end, we study how to evaluate the usefulness of a context without knowing the downstream tasks. We propose a metric and show by experiments that it correlates well with the actual performance of the encoder on many real datasets.
The Limit Order Book (LOB), the mostly fundamental data of the financial market, provides a fine-grained view of market dynamics while poses significant challenges in dealing with the esteemed deep models due to its strong autocorrelation, cross-feature constrains, and feature scale disparity. Existing approaches often tightly couple representation learning with specific downstream tasks in an end-to-end manner, failed to analyze the learned representations individually and explicitly, limiting their reusability and generalization. This paper conducts the first systematic comparative study of LOB representation learning, aiming to identify the effective way of extracting transferable, compact features that capture essential LOB properties. We introduce LOBench, a standardized benchmark with real China A-share market data, offering curated datasets, unified preprocessing, consistent evaluation metrics, and strong baselines. Extensive experiments validate the sufficiency and necessity of LOB representations for various downstream tasks and highlight their advantages over both the traditional task-specific end-to-end models and the advanced representation learning models for general time series. Our work establishes a reproducible framework and provides clear guidelines for future research. Datasets and code will be publicly available at https://github.com/financial-simulation-lab/LOBench.
Recent studies have demonstrated that learning a meaningful internal representation can both accelerate generative training and enhance the generation quality of diffusion transformers. However, existing approaches necessitate to either introduce an external and complex representation training framework or rely on a large-scale, pre-trained representation foundation model to provide representation guidance during the original generative training process. In this study, we posit that the unique discriminative process inherent to diffusion transformers enables them to offer such guidance without requiring external representation components. We therefore propose Self-Representation Alignment (SRA), a simple yet straightforward method that obtains representation guidance through a self-distillation manner. Specifically, SRA aligns the output latent representation of the diffusion transformer in the earlier layer with higher noise to that in the later layer with lower noise to progressively enhance the overall representation learning during only the generative training process. Experimental results indicate that applying SRA to DiTs and SiTs yields consistent performance improvements. Moreover, SRA not only significantly outperforms approaches relying on auxiliary, complex representation training frameworks but also achieves performance comparable to methods that are heavily dependent on powerful external representation priors.
Audiovisual emotion recognition (AVER) aims to infer human emotions from nonverbal visual-audio (VA) cues, offering modality-complementary and language-agnostic advantages. However, AVER remains challenging due to the inherent ambiguity of emotional expressions, cross-modal expressive disparities, and the scarcity of reliably annotated data. Recent self-supervised AVER approaches have introduced strong multimodal representations, yet they predominantly rely on modality-specific encoders and coarse content-level alignment, limiting fine-grained emotional semantic modeling. To address these issues, we propose VAEmo, an efficient two-stage framework for emotion-centric joint VA representation learning with external knowledge injection. In Stage 1, a unified and lightweight representation network is pre-trained on large-scale speaker-centric VA corpora via masked reconstruction and contrastive objectives, mitigating the modality gap and learning expressive, complementary representations without emotion labels. In Stage 2, multimodal large language models automatically generate detailed affective descriptions according to our well-designed chain-of-thought prompting for only a small subset of VA samples; these rich textual semantics are then injected by aligning their corresponding embeddings with VA representations through dual-path contrastive learning, further bridging the emotion gap. Extensive experiments on multiple downstream AVER benchmarks show that VAEmo achieves state-of-the-art performance with a compact design, highlighting the benefit of unified cross-modal encoding and emotion-aware semantic guidance for efficient, generalizable VA emotion representations.
We propose a generalization of modern representation learning objectives by reframing them as recursive divergence alignment processes over localized conditional distributions While recent frameworks like Information Contrastive Learning I-Con unify multiple learning paradigms through KL divergence between fixed neighborhood conditionals we argue this view underplays a crucial recursive structure inherent in the learning process. We introduce Recursive KL Divergence Optimization RKDO a dynamic formalism where representation learning is framed as the evolution of KL divergences across data neighborhoods. This formulation captures contrastive clustering and dimensionality reduction methods as static slices while offering a new path to model stability and local adaptation. Our experiments demonstrate that RKDO offers dual efficiency advantages approximately 30 percent lower loss values compared to static approaches across three different datasets and 60 to 80 percent reduction in computational resources needed to achieve comparable results. This suggests that RKDOs recursive updating mechanism provides a fundamentally more efficient optimization landscape for representation learning with significant implications for resource constrained applications.
Graph auto-encoder is considered a framework for unsupervised learning on graph-structured data by representing graphs in a low dimensional space. It has been proved very powerful for graph analytics. In the real world, complex relationships in various entities can be represented by heterogeneous graphs that contain more abundant semantic information than homogeneous graphs. In general, graph auto-encoders based on homogeneous graphs are not applicable to heterogeneous graphs. In addition, little work has been done to evaluate the effect of different semantics on node embedding in heterogeneous graphs for unsupervised graph representation learning. In this work, we propose a novel Heterogeneous Graph Attention Auto-Encoders (HGATE) for unsupervised representation learning on heterogeneous graph-structured data. Based on the consideration of semantic information, our architecture of HGATE reconstructs not only the edges of the heterogeneous graph but also node attributes, through stacked encoder/decoder layers. Hierarchical attention is used to learn the relevance between a node and its meta-path based neighbors, and the relevance among different meta-paths. HGATE is applicable to transductive learning as well as inductive learning. Node classification and link prediction experiments on real-world heterogeneous graph datasets demonstrate the effectiveness of HGATE for both transductive and inductive tasks.
Multimodal representation learning has garnered significant attention in the AI community, largely due to the success of large pre-trained multimodal foundation models like LLaMA, GPT, Mistral, and CLIP. These models have achieved remarkable performance across various tasks of multimodal information retrieval (MIR), including web search, cross-modal retrieval, and recommender systems, etc. However, due to their enormous parameter sizes, significant efficiency challenges emerge across training, deployment, and inference stages when adapting these models' representation for IR tasks. These challenges present substantial obstacles to the practical adaptation of foundation models for representation learning in information retrieval tasks. To address these pressing issues, we propose organizing the first EReL@MIR workshop at the Web Conference 2025, inviting participants to explore novel solutions, emerging problems, challenges, efficiency evaluation metrics and benchmarks. This workshop aims to provide a platform for both academic and industry researchers to engage in discussions, share insights, and foster collaboration toward achieving efficient and effective representation learning for multimodal information retrieval in the era of large foundation models.
Matrix factorization (MF), a cornerstone of recommender systems, decomposes user-item interaction matrices into latent representations. Traditional MF approaches, however, employ a two-stage, non-end-to-end paradigm, sequentially performing recommendation and clustering, resulting in prohibitive computational costs for large-scale applications like e-commerce and IoT, where billions of users interact with trillions of items. To address this, we propose Matrix Factorization with Dynamic Multi-view Clustering (MFDMC), a unified framework that balances efficient end-to-end training with comprehensive utilization of web-scale data and enhances interpretability. MFDMC leverages dynamic multi-view clustering to learn user and item representations, adaptively pruning poorly formed clusters. Each entity's representation is modeled as a weighted projection of robust clusters, capturing its diverse roles across views. This design maximizes representation space utilization, improves interpretability, and ensures resilience for downstream tasks. Extensive experiments demonstrate MFDMC's superior performance in recommender systems and other representation learning domains, such as computer vision, highlighting its scalability and versatility.
Most current MKGC approaches are predominantly based on discriminative models that maximize conditional likelihood. These approaches struggle to efficiently capture the complex connections in real-world knowledge graphs, thereby limiting their overall performance. To address this issue, we propose a structure-aware multimodal Diffusion model for multimodal knowledge graph Completion (DiffusionCom). DiffusionCom innovatively approaches the problem from the perspective of generative models, modeling the association between the $(head, relation)$ pair and candidate tail entities as their joint probability distribution $p((head, relation), (tail))$, and framing the MKGC task as a process of gradually generating the joint probability distribution from noise. Furthermore, to fully leverage the structural information in MKGs, we propose Structure-MKGformer, an adaptive and structure-aware multimodal knowledge representation learning method, as the encoder for DiffusionCom. Structure-MKGformer captures rich structural information through a multimodal graph attention network (MGAT) and adaptively fuses it with entity representations, thereby enhancing the structural awareness of these representations. This design effectively addresses the limitations of existing MKGC methods, particularly those based on multimodal pre-trained models, in utilizing structural information. DiffusionCom is trained using both generative and discriminative losses for the generator, while the feature extractor is optimized exclusively with discriminative loss. This dual approach allows DiffusionCom to harness the strengths of both generative and discriminative models. Extensive experiments on the FB15k-237-IMG and WN18-IMG datasets demonstrate that DiffusionCom outperforms state-of-the-art models.
Temporal Action Detection and Moment Retrieval constitute two pivotal tasks in video understanding, focusing on precisely localizing temporal segments corresponding to specific actions or events. Recent advancements introduced Moment Detection to unify these two tasks, yet existing approaches remain confined to closed-set scenarios, limiting their applicability in open-world contexts. To bridge this gap, we present Grounding-MD, an innovative, grounded video-language pre-training framework tailored for open-world moment detection. Our framework incorporates an arbitrary number of open-ended natural language queries through a structured prompt mechanism, enabling flexible and scalable moment detection. Grounding-MD leverages a Cross-Modality Fusion Encoder and a Text-Guided Fusion Decoder to facilitate comprehensive video-text alignment and enable effective cross-task collaboration. Through large-scale pre-training on temporal action detection and moment retrieval datasets, Grounding-MD demonstrates exceptional semantic representation learning capabilities, effectively handling diverse and complex query conditions. Comprehensive evaluations across four benchmark datasets including ActivityNet, THUMOS14, ActivityNet-Captions, and Charades-STA demonstrate that Grounding-MD establishes new state-of-the-art performance in zero-shot and supervised settings in open-world moment detection scenarios. All source code and trained models will be released.
The challenge of tracing the source attribution of forged faces has gained significant attention due to the rapid advancement of generative models. However, existing deepfake attribution (DFA) works primarily focus on the interaction among various domains in vision modality, and other modalities such as texts and face parsing are not fully explored. Besides, they tend to fail to assess the generalization performance of deepfake attributors to unseen generators in a fine-grained manner. In this paper, we propose a novel bi-modal guided multi-perspective representation learning (BMRL) framework for zero-shot deepfake attribution (ZS-DFA), which facilitates effective traceability to unseen generators. Specifically, we design a multi-perspective visual encoder (MPVE) to explore general deepfake attribution visual characteristics across three views (i.e., image, noise, and edge). We devise a novel parsing encoder to focus on global face attribute embeddings, enabling parsing-guided DFA representation learning via vision-parsing matching. A language encoder is proposed to capture fine-grained language embeddings, facilitating language-guided general visual forgery representation learning through vision-language alignment. Additionally, we present a novel deepfake attribution contrastive center (DFACC) loss, to pull relevant generators closer and push irrelevant ones away, which can be introduced into DFA models to enhance traceability. Experimental results demonstrate that our method outperforms the state-of-the-art on the ZS-DFA task through various protocols evaluation.
Spatial Transcriptomics (ST) reveals the spatial distribution of gene expression in tissues, offering critical insights into biological processes and disease mechanisms. However, the high cost, limited coverage, and technical complexity of current ST technologies restrict their widespread use in clinical and research settings, making obtaining high-resolution transcriptomic profiles across large tissue areas challenging. Predicting ST from H\&E-stained histology images has emerged as a promising alternative to address these limitations but remains challenging due to the heterogeneous relationship between histomorphology and gene expression, which is affected by substantial variability across patients and tissue sections. In response, we propose PH2ST, a prompt-guided hypergraph learning framework, which leverages limited ST signals to guide multi-scale histological representation learning for accurate and robust spatial gene expression prediction. Extensive evaluations on two public ST datasets and multiple prompt sampling strategies simulating real-world scenarios demonstrate that PH2ST not only outperforms existing state-of-the-art methods, but also shows strong potential for practical applications such as imputing missing spots, ST super-resolution, and local-to-global prediction, highlighting its value for scalable and cost-effective spatial gene expression mapping in biomedical contexts.
Time series anomaly detection holds notable importance for risk identification and fault detection across diverse application domains. Unsupervised learning methods have become popular because they have no requirement for labels. However, due to the challenges posed by the multiplicity of abnormal patterns, the sparsity of anomalies, and the growth of data scale and complexity, these methods often fail to capture robust and representative dependencies within the time series for identifying anomalies. To enhance the ability of models to capture normal patterns of time series and avoid the retrogression of modeling ability triggered by the dependencies on high-quality prior knowledge, we propose a differencing-based contrastive representation learning framework for time series anomaly detection (DConAD). Specifically, DConAD generates differential data to provide additional information about time series and utilizes transformer-based architecture to capture spatiotemporal dependencies, which enhances the robustness of unbiased representation learning ability. Furthermore, DConAD implements a novel KL divergence-based contrastive learning paradigm that only uses positive samples to avoid deviation from reconstruction and deploys the stop-gradient strategy to compel convergence. Extensive experiments on five public datasets show the superiority and effectiveness of DConAD compared with nine baselines. The code is available at https://github.com/shaieesss/DConAD.
Graph Neural Networks (GNNs) update node representations through message passing, which is primarily based on the homophily principle, assuming that adjacent nodes share similar features. However, in real-world graphs with long-tailed degree distributions, high-degree nodes dominate message passing, causing a degree bias where low-degree nodes remain under-represented due to inadequate messages. The main challenge in addressing degree bias is how to discover non-adjacent nodes to provide additional messages to low-degree nodes while reducing excessive messages for high-degree nodes. Nevertheless, exploiting non-adjacent nodes to provide valuable messages is challenging, as it could generate noisy information and disrupt the original graph structures. To solve it, we propose a novel Degree Fairness Graph Transformer, named DegFairGT, to mitigate degree bias by discovering structural similarities between non-adjacent nodes through learnable structural augmentation and structural self-attention. Our key idea is to exploit non-adjacent nodes with similar roles in the same community to generate informative edges under our augmentation, which could provide informative messages between nodes with similar roles while ensuring that the homophily principle is maintained within the community. To enable DegFairGT to learn such structural similarities, we then propose a structural self-attention to capture the similarities between node pairs. To preserve global graph structures and prevent graph augmentation from hindering graph structure, we propose a Self-Supervised Learning task to preserve p-step transition probability and regularize graph augmentation. Extensive experiments on six datasets showed that DegFairGT outperformed state-of-the-art baselines in degree fairness analysis, node classification, and node clustering tasks.
Lifelong Person Re-identification (LReID) suffers from a key challenge in preserving old knowledge while adapting to new information. The existing solutions include rehearsal-based and rehearsal-free methods to address this challenge. Rehearsal-based approaches rely on knowledge distillation, continuously accumulating forgetting during the distillation process. Rehearsal-free methods insufficiently learn the distribution of each domain, leading to forgetfulness over time. To solve these issues, we propose a novel Distribution-aware Forgetting Compensation (DAFC) model that explores cross-domain shared representation learning and domain-specific distribution integration without using old exemplars or knowledge distillation. We propose a Text-driven Prompt Aggregation (TPA) that utilizes text features to enrich prompt elements and guide the prompt model to learn fine-grained representations for each instance. This can enhance the differentiation of identity information and establish the foundation for domain distribution awareness. Then, Distribution-based Awareness and Integration (DAI) is designed to capture each domain-specific distribution by a dedicated expert network and adaptively consolidate them into a shared region in high-dimensional space. In this manner, DAI can consolidate and enhance cross-domain shared representation learning while alleviating catastrophic forgetting. Furthermore, we develop a Knowledge Consolidation Mechanism (KCM) that comprises instance-level discrimination and cross-domain consistency alignment strategies to facilitate model adaptive learning of new knowledge from the current domain and promote knowledge consolidation learning between acquired domain-specific distributions, respectively. Experimental results show that our DAFC outperform state-of-the-art methods by at least 9.8\%/6.6\% and 6.4\%/6.2\% of average mAP/R@1 on two training orders.