论文检索

Jiexia Ye, Weiqi Zhang, Ziyue Li, Jia Li, Fugee Tsung

Accurate medical time series (MedTS) classification is essential for effective clinical diagnosis, yet remains challenging due to complex multi-channel temporal dependencies, information redundancy, and label scarcity. While transformer-based models have shown promise in time series analysis, most are designed for forecasting tasks and fail to fully exploit the unique characteristics of MedTS. In this paper, we introduce MedSpaformer, a transformer-based framework tailored for MedTS classification. It incorporates a sparse token-based dual-attention mechanism that enables global context modeling and token sparsification, allowing dynamic feature refinement by focusing on informative tokens while reducing redundancy. This mechanism is integrated into a multi-granularity cross-channel encoding scheme to capture intra- and inter-granularity temporal dependencies and inter-channel correlations, enabling progressive refinement of task-relevant patterns in medical signals. The sparsification design allows our model to flexibly accommodate inputs with variable lengths and channel dimensions. We also introduce an adaptive label encoder to extract label semantics and address cross-dataset label space misalignment. Together, these components enhance the model’s transferability across heterogeneous medical datasets, which helps alleviate the challenge of label scarcity. Our model outperforms 13 baselines across 7 medical datasets under supervised learning. It also excels in few-shot learning and demonstrates zero-shot capability in both in-domain and cross-domain diagnostics. These results highlight MedSpaformer's robustness and its potential as a unified solution for MedTS classification across diverse settings.

Guanhua Ye, Jifeng He, Yan Li, Junping Du, Zhe Xue, Yingxia Shao, Meiyu Liang, Yawen Li

Questionnaire data serve as a valuable resource across numerous scientific domains, offering insights into human behavior, health, and social trends. Traditional downsampling-based representation learning methods—such as standardization and one-hot encoding—reformat these data into tabular structures that inherently discard semantic richness and obscure inter-sample and inter-feature relationships. Consequently, advanced deep learning models often underperform compared to simpler approaches like gradient-boosted decision trees (GBDT), due to their limited capacity to extract meaningful representations from semantically sparse inputs. To address this limitation, we introduce SemantiQ, a novel upsampling-based representation learning framework that embeds questionnaire responses into a unified semantic space. Leveraging Retrieval-Augmented Generation (RAG) in conjunction with large language models (LLMs), SemantiQ transforms question text, option text, and external knowledge into semantically enriched natural language statements. These statements are then encoded into semantic embeddings, which are further refined through a three-stage training mechanism and test-time training (TTT), enabling the model to capture complex sample- and feature-wise dependencies. Extensive experiments on multiple real-world datasets demonstrate that SemantiQ consistently outperforms state-of-the-art baselines.

Fei Ye, YongCheng Zhong, Qihe Liu, Adrian G. Bors, JingLing Sun, Jinyu Guo, ShiJie Zhou

Continuous learning constitutes a fundamental capability of artificial intelligence systems, enabling them to incrementally assimilate novel information without succumbing to catastrophic forgetting. Recent research has leveraged Pre-Trained Models (PTMs) to enhance continual learning efficacy. Nevertheless, prevailing methodologies typically depend on a singular pre-trained backbone and freeze all pre-trained parameters to mitigate network forgetting, thereby constraining adaptability to emerging tasks. In this study, we introduce an innovative PTM-based framework featuring a Dual-Representation Backbone Architecture (DRBA), which integrates both invariant and evolved representation networks to concurrently capture static and dynamic features. Building upon DRBA, we propose an Adaptive and Expandable Mixture Model (AEMM) that incrementally incorporates new expert modules with minimal parameter overhead to accommodate the learning of each novel task. To further augment adaptability, we develop a Dynamic Adaptive Representation Fusion Mechanism (DARFM) that processes outputs from both representation networks and autonomously generates data-driven adaptive weights, optimizing the contribution of each representation. This mechanism yields an adaptive, semantically enriched composite representation, thereby maximizing positive knowledge transfer. Additionally, we propose a Dynamic Knowledge Calibration Mechanism (DKCM), comprising prediction and representation calibration processes, to ensure consistency in both predictions and feature representations. This approach achieves a balance between stability and plasticity, even when learning complex datasets. Empirical evaluations substantiate that the proposed approach attains state-of-the-art performance.

Shibo Yao, Renshuai Tao, Xiaolong Zheng, Chao Liang, Chunjie Zhang

Recent deepfake detection studies often treat unseen sample detection as a ``zero-shot" task, training on images generated by known models but generalizing to unknown ones. A key real-world challenge arises when a model performs poorly on unknown samples, yet these samples remain available for analysis. This highlights that it should be approached as a ``few-shot" task, where effectively utilizing a small number of samples can lead to significant improvement. Unlike typical few-shot tasks focused on semantic understanding, deepfake detection prioritizes image realism, which closely mirrors real-world distributions. In this work, we propose the Few-shot Training-free Network (FTNet) for real-world few-shot deepfake detection. Simple yet effective, FTNet differs from traditional methods that rely on large-scale known data for training. Instead, FTNet uses only one fake sample from an evaluation set, mimicking the scenario where new samples emerge in the real world and can be gathered for use, without any training or parameter updates. During evaluation, each test sample is compared to the known fake and real samples, and it is classified based on the category of the nearest sample. We conduct a comprehensive analysis of AI-generated images from 29 different generative models and achieve a new SoTA performance, with an average improvement of 8.7% compared to existing methods. This work introduces a fresh perspective on real-world deepfake detection: when the model struggles to generalize on a few-shot sample, leveraging the failed samples leads to better performance.

Lingfeng Yao, Chenpei Huang, Shengyao Wang, Junpei Xue, Hanqing Guo, Jiang Liu, Phone Lin, Tomoaki Ohtsuki, Miao Pan

As generative audio models are rapidly evolving, AI-generated audios increasingly raise concerns about copyright infringement and misinformation spread. Audio watermarking, as a proactive defense, can embed secret messages into audio for copyright protection and source verification. However, current neural audio watermarking methods focus primarily on the imperceptibility and robustness of watermarking, while ignoring its vulnerability to security attacks. In this paper, we develop a simple yet powerful attack: the overwriting attack that overwrites the legitimate audio watermark with a forged one and makes the original legitimate watermark undetectable. Based on the audio watermarking information that the adversary has, we propose three categories of overwriting attacks, i.e., white-box, gray-box, and black-box attacks. We also thoroughly evaluate the proposed attacks on state-of-the-art neural audio watermarking methods. Experimental results demonstrate that the proposed overwriting attacks can effectively compromise existing watermarking schemes across various settings and achieve a nearly 100% attack success rate. The practicality and effectiveness of the proposed overwriting attacks expose security flaws in existing neural audio watermarking systems, underscoring the need to enhance security in future audio watermarking designs.

Ziduo Yang, Yi-Ming Zhao, Xian Wang, Wei Zhuo, Xiaoqing Liu, Lei Shen

Structure optimization, which yields the relaxed structure (minimum‑energy state), is essential for reliable materials property calculations, yet traditional ab initio approaches such as density‑functional theory (DFT) are computationally intensive. Machine learning (ML) has emerged to alleviate this bottleneck but suffers from two major limitations: (i) existing models operate mainly on atoms, leaving lattice vectors implicit despite their critical role in structural optimization; and (ii) they often rely on multi-stage, non-end-to-end workflows that are prone to error accumulation. Here, we present E³Relax, an end-to-end equivariant graph neural network that maps an unrelaxed crystal directly to its relaxed structure. E³Relax promotes both atoms and lattice vectors to graph nodes endowed with dual scalar–vector features, enabling unified and symmetry‑preserving modeling of atomic displacements and lattice deformations. A layer‑wise supervision strategy forces every network depth to make a physically meaningful refinement, mimicking the incremental convergence of DFT while preserving a fully end‑to‑end pipeline. We evaluate E³Relax on four benchmark datasets and demonstrate that it achieves remarkable accuracy and efficiency. Through DFT validations, we show that the structures predicted by E³Relax are energetically favorable, making them suitable as high-quality initial configurations to accelerate DFT calculations.

Zhiwen Yang, Jiehua Zhang, Chenggang Yan, Yuhan Gao, Zongpeng Li, Xichun Sheng, Liang Li

Continual forgetting task aims to continuously remove multiple target knowledge subsets from pre-trained models while maintaining the integrity of remaining knowledge. Existing methods suffer from both incomplete forgetting of target knowledge and unintended forgetting of indistinguishable remaining knowledge. To address these challenges, we propose the forgetting knowledge localization and isolation for continual forgetting in pre-trained vision models which precisely forgets target knowledge while reducing over-forgetting of remaining knowledge. To achieve precise forgetting, we first propose the forgetting knowledge layer localization to explore layers in the model which are more related to forgetting knowledge. Then, we design the forgetting knowledge parameter isolation to isolate the parameters sensitive to forgetting knowledge in these selected layers, mitigating over-forgetting of remaining knowledge. Finally, we fine-tune these isolated parameters and freeze the remaining parameters to achieve efficient forgetting while maintaining high performance on retained datasets. Extensive experimental results demonstrate that our method achieves superior performance over state-of-the-art methods across multiple continual forgetting tasks.

Zhan Yang, Yiran Liu, Youyuan Huang, Yinan Li

Unsupervised cross-modal hashing has gained significant attention for efficient retrieval between heterogeneous modalities through encoding data into the unified binary representations, offering low storage cost and fast response. However, the constraints of existing methods persist in bridging the cross-modal semantic gap and capturing fine-grained global semantic structures without explicit labels. In this paper, we propose an innovative unsupervised Stationary distribution and soft Clustering Transformer Hashing approach for cross-modal retrieval, denoted as SCTH. Initially, a Transformer-based modality fusion encoder is employed to extract abundant cross-modal semantic representations, further integrated with contrastive hashing to minimize the semantic gap. To enhance the inter-modal alignment, a pseudo-classifier clustering module with entropy-regularized contrastive loss is presented, ensuring balanced and diverse cluster assignments in unsupervised settings. Additionally, a Markovian stationary distribution strategy stabilizes the feature representations through mitigating the interference of noise and outliers. Comprehensive experiments on MIRFlickr, NUS-WIDE, and IAPR-TC12 datasets validate that SCTH outperforms state-of-the-art hashing methods in cross-modal retrieval tasks, demonstrating superior generalization performance.

Zewen Yang, Dongfa Zhang, Xiaobing Dai, Fengyi Yu, Chi Zhang, Bingkun Huang, Hamid Sadeghian, Sami Haddadin

Gaussian Processes (GPs), as a nonparametric learning method, offer flexible modeling capabilities and calibrated uncertainty quantification for function approximations. Additionally, GPs support online learning by efficiently incorporating new data with polynomial-time computation, making them well-suited for safety-critical dynamical systems that require rapid adaptation. However, the inference and online updates of exact GPs, when processing streaming data, incur cubic computation time and quadratic storage memory complexity, limiting their scalability to large datasets in real-time settings. In this paper, we propose a streaming kernel-induced progressively generated expert framework of Gaussian processes (SkyGP) that addresses both computational and memory constraints by maintaining a bounded set of experts, while inheriting the learning performance guarantees from exact Gaussian processes. Furthermore, two SkyGP variants are introduced, each tailored to a specific objective, either maximizing prediction accuracy (SkyGP-Dense) or improving computational efficiency (SkyGP-Fast). The effectiveness of SkyGP is validated through extensive benchmarks and real-time control experiments demonstrating its superior performance compared to state-of-the-art approaches.

Yunfei Yang, Xiaojun Chen, Yuexin Xuan, Zhendong Zhao, Xin Zhao, He Li

Model watermarking techniques can embed watermark information into the protected model for ownership declaration by constructing specific input-output pairs. However, existing watermarks are easily removed when facing model stealing attacks, and make it difficult for model owners to effectively verify the copyright of stolen models. In this paper, we analyze the root cause of the failure of current watermarking methods under model stealing scenarios and then explore potential solutions. Specifically, we introduce a robust watermarking framework, DeepTracer, which leverages a novel watermark samples construction method and a same-class coupling loss constraint. DeepTracer can incur a high-coupling model between watermark task and primary task that makes adversaries inevitably learn the hidden watermark task when stealing the primary task functionality. Furthermore, we propose an effective watermark samples filtering mechanism that elaborately select watermark key samples used in model ownership verification to enhance the reliability of watermarks. Extensive experiments across multiple datasets and models demonstrate that our method surpasses existing approaches in defending against various model stealing attacks, as well as watermark attacks, and achieves new state-of-the-art effectiveness and robustness.

Yiyuan Yang, Guodong Long, Qinghua Lu, Liming Zhu, Jing Jiang

Federated foundation models represent a new paradigm to jointly fine-tune pre-trained foundation models across clients. It is still a challenge to fine-tune foundation models for a small group of new users or specialized scenarios, which typically involve limited data compared to the large-scale data used in pre-training. In this context, the trade-off between personalization and federation becomes more sensitive. To tackle these, we proposed a bi-level personalization framework for federated fine-tuning on foundation models. Specifically, we conduct personalized fine-tuning on the client-level using its private data, and then conduct a personalized aggregation on the server-level using similar users measured by client-specific task vectors. Given the personalization information gained from client-level fine-tuning, the server-level personalized aggregation can gain group-wise personalization information while mitigating the disturbance of irrelevant or interest-conflict clients with non-IID data. The effectiveness of the proposed algorithm has been demonstrated by extensive experimental analysis in benchmark datasets.

Yi Yang, Yinghui Huang, Xiangyu Chang

Existing theoretical work on Bayes-optimal fair classifiers usually considers a single (binary) sensitive feature. In practice, individuals are often defined by multiple sensitive features. In this paper, we characterize the Bayes-optimal fair classifier for multiple sensitive features under general approximate fairness measures, including *mean difference* (MD) and *mean ratio* (MR). We show that these approximate measures for existing group fairness notions, including Demographic Parity, Equal Opportunity, Predictive Equality, and Accuracy Parity, are linear transformations of selection rates for specific groups defined by both labels and sensitive features. We then characterize that Bayes-optimal fair classifiers for multiple sensitive features under both MD and MR become instance-dependent thresholding rules that rely on a weighted sum of these group membership probabilities. Our framework applies to both attribute-aware and attribute-blind settings and can accommodate composite fairness notions like Equalized Odds. Building on this, we propose two practical algorithms for Bayes-optimal fair classification via in-processing and post-processing. We show empirically that our methods compare favorably to existing methods.

Xu Yang, Moqi Li, Kun Wei

Continual Test-Time Domain Adaptation (CTTA) aims to adapt a pre-trained source model to a dynamically evolving target domain without requiring additional data collection or labeling efforts. A key challenge in this setting is to achieve rapid performance improvement in the current domain using unlabeled data, while avoiding impairing generalization to future domains in complex scenarios. To enhance the discriminative capability of the inference models, we propose a novel framework that integrates an external auxiliary generative model with a test-time adaptive method, leveraging cross-validation to identify reliable supervisory signals. Specifically, for each test instance, we utilize a diffusion module to generate a calibrated instance under the textual description of its predicted category. Based on the generated one, we design a learning strategy with the following components: (1) the calibrated instance and its category are used to form a supervisory signal; (2) the predicted category of the calibrated instance is compared with the test instance for selecting reliable signals. For these generated and selected instances, adaptive weighting is applied during optimization to stabilize the category distribution and preserve prediction diversity. Finally, based on the inverse process of diffusion, we construct the negative instance of the generated instance and introduce a robust contrastive learning to further calibrate model optimization. Extensive experiments demonstrate that our method achieves state-of-the-art performance across multiple benchmarks. Ablation studies further validate the effectiveness of each proposed component.

Xu Yang, Jiapeng Zhang, Dongyang Zhao, Guo Chen, Zhuo Tang

The KV cache in self-attention has emerged as a major bottleneck in long-context and large-batch inference for LLMs. Existing approaches often treat sparsity prediction and compression as separate modules—relying on auxiliary index structures to select relevant tokens, and on complex quantization schemes to reduce memory usage. This fragmented design introduces redundant overhead and limits scalability. In this paper, we propose a novel paradigm: treating the compressed key representation not merely as storage, but as a self-indexing structure that directly enables efficient sparse attention. By designing a sign-based 1-bit vector quantization (VQ) scheme, our method unifies compression and retrieval in a single, hardware-friendly format. This approach eliminates the need for external indices or learning-based predictors, offering a lightweight yet robust solution for memory-constrained inference. All components are designed to be hardware-efficient and easy to implement. By implementing custom CUDA kernels, our method integrates seamlessly with FlashAttention, minimizing additional runtime and memory overhead. Experimental results demonstrate that our approach delivers both effectiveness and efficiency.

Xinye Yang, Junhao Wang, RuiLi, Haosen Sun, Xuesheng Zhang, Zebang Liu, Gaochao Xu, Yiwei Chen

Traditional knowledge distillation relies on simple MSE or KL divergence losses that fail to capture the complex distributional relationships between teacher and student model representations. We propose FlowDistill, a novel distillation framework that employs normalizing flows to model and transfer the intricate knowledge distributions from teacher to student models. Our approach introduces three key innovations: (1) Invertible Knowledge Mapping using continuous normalizing flows (CNFs) to learn bijective transformations between teacher and student representation spaces, enabling precise knowledge transfer without information loss, (2) Flow-Guided Progressive Distillation that gradually increases the complexity of knowledge transfer by learning hierarchical flow transformations from simple to complex distributions, and (3) Conditional Flow Networks that adapt knowledge transfer based on input context and task requirements. Unlike previous diffusion-based distillation methods such as DiffKD that suffer from computational overhead due to iterative denoising processes and information loss during noise addition, our flow-based approach provides exact invertible transformations with significantly reduced computational cost. Extensive experiments on ImageNet classification, COCO object detection, and Cityscapes semantic segmentation demonstrate that FlowDistill achieves superior performance with 2.1% accuracy improvement over DiffKD on ResNet-34 to ResNet-18 distillation while reducing inference time by 3.5×. Our method establishes new state-of-the-art results across multiple distillation benchmarks and provides theoretical guarantees for lossless knowledge transfer through invertible flow transformations.

Xiao-Wen Yang, Xuan-Yi Zhu, Ding-Chu Zhang, Wen-Da Wei, Jie-Jing Shao, Zhi Zhou, Lan-Zhe Guo, Yu-Feng Li

Although autoregressive language models demonstrated remarkable performance across various tasks, their effectiveness in symbolic reasoning and decision-making scenarios remains constrained. Recent research indicates that training language models to emulate symbolic search algorithms (e.g. depth-first search or A* algorithm) can yield strong improvements in their symbolic reasoning and planning capabilities. However, existing methods only achieve superficial imitation of symbolic search trajectories, as their generation processes lack explicit backtracking mechanisms. This limitation prevents models from truly mastering symbolic search, often resulting in rigid and redundant outputs with poor solution quality. To address this issue, we propose a self-backtracking mechanism that enables LLMs to autonomously determine when to backtrack through specialized training, effectively utilizing this capability to scale during inference. By introducing a self-improvement strategy, the model can further refine its search process into optimal solution generation, improving problem-solving efficiency. Empirical evaluations demonstrate that our method boosts LLMs' reasoning on the Countdown task by 40% over optimal-path supervised fine-tuning (SFT) and improves both performance and efficiency on the Maze Navigation task.

Xianling Yang, Zhiwen Yu, Sun Song, Kaixiang Yang

Semi-supervised learning (SSL) based on pseudo-label and consistency has achieved significant success. The core idea behind these methods is to assign sample weights based on pseudo-label probabilities, thereby guiding the model toward biased learning. However, existing research still faces two major challenges in guiding learning: (1) how to evaluate learning states across different classes in the absence of labels, and (2) how to construct an effective sample weight space that provides precise guidance throughout training. To address these challenges, we propose the Bi-Dimensional Sample Weight Guidance algorithm, BidMatch. BidMatch introduces Class Information Entropy (CIE), which captures the learning relationships between classes and reflects the model’s learning state for each class. Additionally, Pseudo-label Probability Redistribution (PPR) is proposed to maintain distribution invariance and sparsity during training, thereby emphasizing differences in sample importance. By leveraging CIE and PPR, BidMatch generates sample weights that account for both class and instance dimensions, effectively guiding the model toward balanced and efficient learning across classes. BidMatch has demonstrated state-of-the-art performance on various SSL datasets. Notably, it achieved a 6.45% error rate on CIFAR-10 with only one label per class, significantly outperforming baseline methods.

Xi Yang, Han Zhang, Zhijian Lin, Yibiao Hu, Hong Han

Out-of-context misinformation (OOC) is a low-cost form of misinformation in news reports, which refers to place authentic images into out-of-context or fabricated image-text pairings. This problem has attracted significant attention from researchers in recent years. Current methods focus on assessing image-text consistency or generating explanations. However, these approaches assume that the training and test data are drawn from the same distribution. When encountering novel news domains, models tend to perform poorly due to the lack of prior knowledge. To address this challenge, we propose Variational Domain-Invariant Learning with Test-Time Training (VDT) framework to enhance the domain adaptation capability for OOC misinformation detection. Domain-Invariant Variational Align module is employed to jointly encodes source and target domain data to learn a separable distributional space and domain-invariant features. For preserving semantic integrity, we utilize domain consistency constraint module to reconstruct the source and target domain latent distribution. During testing phase, we adopt the test-time training strategy and confidence-variance filtering module to dynamically updating the VAE encoder and classifier, facilitating the model's adaptation to the target domain distribution. Extensive experiments conducted on the benchmark dataset NewsCLIPpings demonstrate that our method outperforms state-of-the-art baselines under most domain adaptation settings.

Wenjie Yang, Shengzhong Zhang, Chen Ye, Jiaxing Guo, Tongshan Xu, Zengfeng Huang

Graph neural networks (GNNs) have demonstrated strong performance in various graph mining tasks but rely heavily on extensively labeled nodes. To improve training efficiency, graph active learning (GAL) has emerged as a solution for selecting the most informative nodes for labeling. However, existing GAL methods are primarily designed for homophilic graphs, where nodes with the same labels are more likely to be connected. In this work, we systematically study active learning on heterophilic graphs, a setting that has received limited attention. Surprisingly, we observe that existing GAL methods fail to consistently outperform random sampling on heterophilic graphs. Through an in-depth investigation, we reveal that these methods implicitly assume homophily even on heterophilic graphs, leading to suboptimal performance. To address this issue, we introduce the principle of "Know Your Neighbors" and propose an active learning algorithm KyN specifically for heterophilic graphs. The core idea of KyN is to provide GNNs with accurate estimations of homophily distribution by labeling nodes together with their neighbors. We implement KyN based on subgraph sampling with probabilities proportional to l1 Lewis weights, which is supported by solid theoretical guarantees. Extensive experiments on diverse real-world datasets, including a large heterophilic graph with over 2 million nodes, demonstrate the effectiveness and scalability of KyN.

Tao Yang, Weihao Wu, Tingzhu Huang

Recently, continuous transform-based tensor representation has emerged as a promising tool for multi-dimensional data recovery. However, the existing continuous transforms are essentially single-layer linear mappings, which limits their ability to capture the complex relationships inherent in multi-dimensional data. To overcome this limitation, we propose a Hierarchical Nonlinear Continuous Transform-based Tensor Representation (HiNCoT) for multi-dimensional data recovery. By leveraging the hierarchical nonlinear continuous transform, HiNCoT constructs the recovered tensor from a latent tensor, which is generated by the deep representation module with a low-rank core tensor as input. Compared with the existing continuous transform-based methods, HiNCoT can more effectively capture the complex nonlinear relationships inherent in multi-dimensional data along the third dimension. To evaluate the effectiveness of the proposed HiNCoT, we suggest an HiNCoT-based multi-dimensional data recovery model. Extensive experiments on diverse degeneration scenarios demonstrate the superiority of our hierarchical nonlinear transform-based method over existing single-layer linear transform-based methods.