IJCAI 2025 Papers — Page 7
International Joint Conference on Artificial Intelligence · 1014 papers
Mamba-Based Graph Convolutional Networks: Tackling Over-smoothing with Selective State Space
Xin He (Jilin University), Xin Wang (Jilin University)
Graph Neural NetworkGraphBenchmark
🎯 What it does: Introduce a selective state space mechanism based on Mamba into graph neural networks, and propose the MbaGCN framework.
Mask Does Not Matter: A Unified Latent Diffusion-Enhanced Framework for Mask-Free Virtual Try-On
Chenghu Du (Wuhan University of Technology), Yi Rong (Wuhan University of Technology)
GenerationDiffusion modelAuto EncoderImage
🎯 What it does: Proposes a mask-free virtual try-on framework called MFTON, achieving high-quality garment reconstruction by directly utilizing portrait images and temporary try-on results as denoising conditions.
MaskDGNN: Self-Supervised Dynamic Graph Neural Networks with Activeness-aware Temporal Masking
Yiming He (Ocean University of China), Yanwei Yu (Ocean University of China)
Computational EfficiencyRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: Proposed a self-supervised dynamic graph neural network called MaskDGNN, which performs edge masking along the timeline by evaluating node activity levels, and enhances the representation capability of dynamic graphs through a frequency domain enhancement module that adaptively models distribution drift.
MASTER: A Multi-granularity Invariant Structure Clustering Scheme for Multi-view Clustering
Suixue Wang (Hainan University), Weiliang Huo (Hainan University)
OptimizationAuto EncoderContrastive LearningImageTextMultimodality
🎯 What it does: Proposed a multi-granularity invariant structure clustering approach called MASTER, which extracts low-level, mid-level, and high-level invariant structures from samples, neighborhoods, and class hierarchies respectively through a bottom-up process to achieve multi-view clustering.
MATCH: Modality-Calibrated Hypergraph Fusion Network for Conversational Emotion Recognition
Jiandong Shi (Zhejiang Normal University), Jiye Liang (Shanxi University)
ClassificationRecognitionRecurrent Neural NetworkGraph Neural NetworkContrastive LearningMultimodality
🎯 What it does: Propose the MATCH model, which achieves multimodal dialogue emotion recognition by performing fine-grained calibration of dialogue entities (speakers and utterances) at both modal and contextual levels, and then inputting the calibrated features into hypergraph and line graph structures.
Maximin Share Guarantees for Few Agents with Subadditive Valuations
George Christodoulou (Athena Research Center), Alkmini Sgouritsa (Athens University of Economics and Business)
Optimization
🎯 What it does: This paper studies the fair division problem under subadditive value functions with a small number of agents (up to four), proving that a 1/2 approximation of the maximin share (MMS) allocation can be achieved.
Maximum Entropy Softmax Policy Gradient via Entropy Advantage Estimation
Jean Seong Bjorn Choe (Korea University), Jong-kook Kim (Korea University)
Reinforcement LearningBenchmark
🎯 What it does: Propose Entropy Advantage Policy Optimisation (EAPO), improving direct Softmax policy gradients in Softmax max-entropy RL by separately estimating task rewards and entropy rewards;
MC3D-AD: A Unified Geometry-aware Reconstruction Model for Multi-category 3D Anomaly Detection
Jiayi Cheng (Shenzhen University), Jinbao Wang (Shenzhen University)
Anomaly DetectionTransformerPoint Cloud
🎯 What it does: Proposed a unified geometry-aware reconstruction model called MC3D-AD for multi-class 3D anomaly detection.
MCD-CLIP: Multi-view Chest Disease Diagnosis with Disentangled CLIP
Songyue Cai (University of Electronic Science and Technology of China), Xiaofeng Zhu (University of Electronic Science and Technology of China)
ClassificationRepresentation LearningPrompt EngineeringVision Language ModelContrastive LearningBiomedical Data
🎯 What it does: Propose a multi-view chest X-ray disease diagnosis framework MCD-CLIP based on CLIP, achieving efficient transfer learning through visual prompt alignment, representation decoupling, and text enhancement.
MCF-Spouse: A Multi-Label Causal Feature Selection Method with Optimal Spouses Discovery
Lin Ma (Jilin University), Juncheng Hu (Jilin University)
OptimizationData-Centric LearningTabularBiomedical Data
🎯 What it does: Propose a multi-label causal feature selection method called MCF-Spouse, which can identify optimal spouse variables on target labels to improve the quality of feature subsets.
MedualTime: A Dual-Adapter Language Model for Medical Time Series-Text Multimodal Learning
Jiexia Ye (Hong Kong University of Science and Technology), Fugee Tsung (Hong Kong University of Science and Technology)
TransformerLarge Language ModelSupervised Fine-TuningTextMultimodalityTime SeriesBiomedical Data
🎯 What it does: Propose a dual-adapter language model MedualTime for medical time series-text multimodal learning, supporting both time series and text modalities as primary modalities simultaneously.
MEGAD: A Memory-Efficient Framework for Large-Scale Attributed Graph Anomaly Detection
Yifan Zhang (Nanjing University of Information Science and Technology), Fei Dai (Southwest Forestry University)
Anomaly DetectionAuto EncoderContrastive LearningGraph
🎯 What it does: Proposes MEGAD, a memory-friendly large-scale attributed graph anomaly detection framework
Meta Label Correction with Generalization Regularizer
Tao Tong (University of Electronic Science and Technology of China), Xiaofeng Zhu (University of Electronic Science and Technology of China)
ClassificationMeta LearningImage
🎯 What it does: Propose a meta-learning based label correction method MLCGR, which first filters noisy labels through gradient scores, then introduces generalization regularization to enhance the model's generalization capability.
Metapath and Hypergraph Structure-based Multi-Channel Graph Contrastive Learning for Student Performance Prediction
Lingyun Song (Northwestern Polytechnical University), Xuequn Shang (Northwestern Polytechnical University)
Recommendation SystemGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposed a student performance prediction framework named MCGCL based on multi-channel graph contrastive learning, enhancing student and question feature learning by constructing high-order hypergraphs and multiple meta-paths.
METOR: A Unified Framework for Mutual Enhancement of Objects and Relationships in Open-vocabulary Video Visual Relationship Detection
Yongqi Wang (Beijing Institute of Technology), Shuo Yang (Shenzhen MSU-BIT University)
Object DetectionTransformerVision Language ModelContrastive LearningVideo
🎯 What it does: Proposed the METOR framework, which adopts a query-based unified model, combining CLIP-based context refinement encoding and iterative enhancement to achieve mutual improvement between objects and relationships in video visual relationship detection.
MGCA-Net: Multi-Graph Contextual Attention Network for Two-View Correspondence Learning
Shuyuan Lin (Jinan University), Qiangqiang Wu (City University of Hong Kong)
Pose EstimationGraph Neural NetworkContrastive LearningImage
🎯 What it does: This paper proposes a two-view correspondence learning network named MGCA-Net, aiming to enhance the robustness of outlier removal and camera pose estimation.
MHANet: Multi-scale Hybrid Attention Network for Auditory Attention Detection
Lu Li (Anhui University), Zhao Lv (Anhui University)
ClassificationConvolutional Neural NetworkTransformerTime SeriesBiomedical Data
🎯 What it does: Proposed a multi-scale hybrid attention network (MHANet) for detecting auditory attention from electroencephalogram (EEG) signals.
MiniMal: Hard-Label Adversarial Attack Against Static Malware Detection with Minimal Perturbation
Chengyi Li (National University of Defense Technology), Yuhang Mao
Adversarial Attack
🎯 What it does: This paper proposes MiniMal, which implements black-box hard-label adversarial attacks on Windows PE files, significantly reducing the perturbation rate while maintaining malicious functionality.
Minimizing Polarization and Disagreement in the Friedkin–Johnsen Model with Unknown Innate Opinions
Federico Cinus (Sapienza University), Francesco Bonchi (CENTAI Institute)
OptimizationGraph Neural NetworkGraph
🎯 What it does: The study addresses the opinion optimization problem under the Friedkin-Johnsen model, where intrinsic opinions are unknown and only limited nodes can be queried. It proposes a three-step framework encompassing node selection, opinion reconstruction, and objective function optimization.
MIRROR: Multi-agent Intra- and Inter-Reflection for Optimized Reasoning in Tool Learning
Zikang Guo (University of Science and Technology of China), Zhendong Mao (University of Science and Technology of China)
OptimizationLarge Language ModelAgentic AIPrompt EngineeringBenchmark
🎯 What it does: Studied a multi-agent tool learning framework called MIRROR, combining pre-reflection and post-reflection to optimize LLM's reasoning and execution.
Misclassification-driven Fingerprinting for DNNs Using Frequency-aware GANs
Weixing Liu (Shenzhen University), Shenghua Zhong (Shenzhen University)
Safty and PrivacyGenerative Adversarial NetworkImage
🎯 What it does: Propose a fingerprint generation framework based on frequency-domain aware GAN, utilizing generated misclassified samples as fingerprints for model ownership verification.
Mitigating Message Imbalance in Fraud Detection with Dual-View Graph Representation Learning
Yudan Song (Guangxi Normal University), Xingcheng Fu (Guangxi Normal University)
Anomaly DetectionRepresentation LearningGraph Neural NetworkGraphFinance Related
🎯 What it does: Proposed the MimbFD dual-view graph representation learning framework to address the information propagation imbalance problem in GNNs for fraud detection.
Mitigating Over-Smoothing in Graph Neural Networks via Separation Coefficient-Guided Adaptive Graph Structure Adjustment
Hanyang Meng (Jiangnan University), Li Peng (Jiangnan University)
ClassificationRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: Proposes an adaptive graph structure adjustment method based on pseudo-labels, enhancing inter-class separation and alleviating the over-smoothing issue in GNNs by adding/removing edges.
Mixture-of-Queries Transformer: Camouflaged Instance Segmentation via Queries Cooperation and Frequency Enhancement
Weiwei Feng (Zhejiang University), Weiqiang Wang (Ant Group)
SegmentationTransformerMixture of ExpertsImage
🎯 What it does: Propose a Mixture-of-Queries Transformer (MoQT) that achieves stealthy object instance segmentation through frequency-domain enhanced feature extractors and multi-expert query decoders.
MMET: A Multi-Input and Multi-Scale Transformer for Efficient PDEs Solving
Yichen Luo (KTH Royal Institute of Technology), Zhibo Pang (KTH Royal Institute of Technology)
TransformerBenchmarkPhysics Related
🎯 What it does: Proposed a multi-input multi-scale efficient Transformer (MMET) for real-time solving of large-scale, complex multi-physics PDEs.
MMGIA: Gradient Inversion Attack Against Multimodal Federated Learning via Intermodal Correlation
Lele Zheng (Xidian University), Xiaochun Cao (Sun Yat-sen University)
Federated LearningSafty and PrivacyAdversarial AttackConvolutional Neural NetworkTransformerLarge Language ModelVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Propose MMGIA, a two-stage gradient inversion attack that leverages cross-modal correlations to recover image and text data from shared gradients in multi-modal federated learning.
MMNet: Missing-Aware and Memory-Enhanced Network for Multivariate Time Series Imputation
Xiaoye Miao (Zhejiang University), Xiaohua Pan (Zhejiang University)
RestorationConvolutional Neural NetworkTransformerTime SeriesBiomedical Data
🎯 What it does: This paper proposes a multi-scale missing-aware memory-enhanced network called MMNet for imputing missing values in multivariate time series;
Modality-Fair Preference Optimization for Trustworthy MLLM Alignment
Songtao Jiang (Zhejiang University), Zuozhu Liu (Zhejiang University)
OptimizationExplainability and InterpretabilityData-Centric LearningReinforcement Learning from Human FeedbackReinforcement LearningVision Language ModelDiffusion modelImageTextMultimodalityBenchmark
🎯 What it does: By implementing Modal Fair Preference Optimization (MFPO) in multimodal large language models, significantly reduces visual errors (hallucinations) and improves model credibility.
Modality-Guided Dynamic Graph Fusion and Temporal Diffusion for Self-Supervised RGB-T Tracking
Shenglan Li (China University of Mining and Technology), Jiaqi Zhao (China University of Mining and Technology)
Object TrackingGraph Neural NetworkTransformerDiffusion modelVideoMultimodality
🎯 What it does: Propose GDSTrack, a self-supervised RGB-T tracker that combines dynamic graph fusion and temporal graph diffusion to reduce pseudo-label noise and enhance tracking performance.
Model Rake: A Defense Against Stealing Attacks in Split Learning
Qinbo Zhang (Wuhan University), Jiawei Jiang (Wuhan University)
Federated LearningSafty and PrivacyContrastive Learning
🎯 What it does: This paper proposes a dual-base model defense mechanism called Model Rake to prevent model and data theft in split learning.
Model-Based Closed-Loop Control Algorithm for Stochastic Partial Differential Equation Control
Peiyan Hu (Academy of Mathematics and Systems Science Chinese Academy of Sciences), Zhiming Ma (Academy of Mathematics and Systems Science Chinese Academy of Sciences)
Convolutional Neural NetworkPhysics RelatedStochastic Differential Equation
🎯 What it does: Proposed a model-driven closed-loop control framework called MB-CC for stochastic partial differential equations (SPDEs), which can simultaneously perform state prediction and real-time control.
Modular Deep Reinforcement Learning for Multi-Workload Offloading in Edge Networks
Hongchang Ke (Changchun Institute of Technology of China), Jia Zhao (Changchun Institute of Technology of China)
OptimizationReinforcement LearningTabular
🎯 What it does: Designed and implemented a flexible modular deep reinforcement learning framework (DRL-MWF) to address dynamic task offloading in multi-workload, multi-edge server environments. The framework includes key technologies such as state representation and normalization, modular policy networks, weighted policy correction, and prioritized experience replay.
MonoMixer: Marrying Convolution and Vision Transformer for Efficient Self-Supervised Monocular Depth Estimation
Zhiyong Chang, Yan Wang (Zhejiang University)
Depth EstimationConvolutional Neural NetworkTransformerImage
🎯 What it does: Propose MonoMixer, a lightweight CNN-Transformer hybrid architecture for self-supervised monocular depth estimation.
More Efforts Towards Fixed-Parameter Approximability of Multiwinner Rules
Sushmita Gupta, Anannya Upasana (Institute of Mathematical Sciences HBNI)
OptimizationComputational Efficiency
🎯 What it does: This paper studies the multi-choice committee problem based on the Thiele rule, proposing a fixed-parameter tractable approximation scheme (FPT-AS), lossy kernel, and a one-time additive approximation algorithm on K_{d,d}-free graphs, and proves the FPT solvability of the PAV rule under the total score parameter.
Most General Explanations of Tree Ensembles
Yacine Izza (National University of Singapore), Peter J. Stuckey (Monash University)
Explainability and InterpretabilityTabularBenchmark
🎯 What it does: This paper studies how to compute the maximum widening inductive explanation (Max-iAXp) in tree ensemble models, which is the most general explanation that covers as much of the input space as possible while still maintaining the correctness of the model's predictions.
Most Probable Explanation in Probabilistic Answer Set Programming
Damiano Azzolini (University of Ferrara), Fabrizio Riguzzi (University of Ferrara)
Graph
🎯 What it does: This paper proposes three new methods to address the most probable explanation (MPE) problem in probabilistic answer set programming (PASP).
MPPQ: Enhancing Post-Training Quantization for LLMs via Mixed Supervision, Proxy Rounding, and Pre-Searching
Mingrun Wei (Beijing Jiaotong University), Dong Wang (Beijing Jiaotong University)
Computational EfficiencyTransformerLarge Language ModelText
🎯 What it does: This paper proposes a novel post-training quantization framework called MPPQ for efficiently quantizing large language models at extremely low bitwidths (e.g., W4A4, W2A16).
MS-DPPs: Multi-Source Determinantal Point Processes for Contextual Diversity Refinement of Composite Attributes in Text to Image Retrieval
Naoya Sogi (NEC Corporation), Takayuki Okatani (Tohoku University)
RetrievalVision Language ModelImageText
🎯 What it does: Propose a post-processing method called MS-DPP in text-to-image retrieval, which utilizes multi-source Determinantal Point Process (DPP) to refine the diversity of multiple attributes in retrieval results, thereby controlling the diversity of attributes such as image appearance, capture time, and location while maintaining relevance.
MSCI: Addressing CLIP's Inherent Limitations for Compositional Zero-Shot Learning
Yue Wang (Nanjing University of Aeronautics and Astronautics), Yicong Li (Nanjing University of Aeronautics and Astronautics)
TransformerSupervised Fine-TuningPrompt EngineeringVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Designed the MSCI model by adaptively aggregating low-level details and high-level global features, and gradually injecting these information into text prompts through multi-stage cross-modal interactions, thereby enhancing CLIP's performance in synthetic zero-shot recognition.
MSMAR-RL: Multi-Step Masked-Attention Recovery Reinforcement Learning for Safe Maneuver Decision in High-Speed Pursuit-Evasion Game
Yang Zhao (Northwestern Polytechnical University), Xuelong Li (China Telecom)
Robotic IntelligenceTransformerReinforcement Learning
🎯 What it does: This paper studies safe decision-making in high-speed drone pursuit-evasion games and proposes a zero-violation recovery reinforcement learning framework.
MsRAG: Knowledge Augumented Image Captioning with Object-level Multi-source RAG
Yuming Qiao (OPPO Research Institute), Xudong Zhang (OPPO AI Center)
Object DetectionGenerationRetrievalTransformerLarge Language ModelPrompt EngineeringVision Language ModelImageTextMultimodalityRetrieval-Augmented Generation
🎯 What it does: Proposes the MsRAG framework, which enhances image descriptions through multi-source retrieval and visual-RAG alignment without requiring user queries.
MSVIT: Improving Spiking Vision Transformer Using Multi-scale Attention Fusion
Wei Hua (China Nanhu Academy of Electronics and Information Technology), Yangyang Shu (University of New South Wales)
ClassificationRecognitionSpiking Neural NetworkTransformerImage
🎯 What it does: Propose a multi-scale spiking attention fusion Spike-Transformer (MSVIT), enhancing the representation capability of SNN in visual tasks through multi-scale feature fusion.
MTGIB-UNet: A Multi-Task Graph Information Bottleneck and Uncertainty Weighted Network for ADMET Prediction
Xuqiang Li (University of Science and Technology of China), Yang Wang (University of Science and Technology of China)
Drug DiscoveryGraph Neural NetworkGraph
🎯 What it does: Propose MTGIB-UNet, a multi-task graph neural network that leverages the Graph Information Bottleneck (GIB) to extract task-related subgraphs and dynamically adjusts task weights through uncertainty weighting, aiming to enhance ADMET prediction performance.
MTPNet: Multi-Grained Target Perception for Unified Activity Cliff Prediction
Zishan Shu (Peking University), Jie Chen (Peking University)
Drug DiscoveryGraph Neural NetworkTransformerLarge Language ModelBiomedical Data
🎯 What it does: Proposed MTPNet, which utilizes macro and micro protein semantic-guided multi-granularity target-aware modules to achieve unified representation of molecular and receptor interaction information, thereby enabling regression and classification prediction of activity cliffs.
Multi Objective Quantile Based Reinforcement Learning for Modern Urban Planning
Lukasz Pelcner (Lancaster University), Peter M. Atkinson (Lancaster University)
OptimizationReinforcement LearningTabularTime Series
🎯 What it does: This paper proposes a dual-agent multi-objective reinforcement learning framework to simultaneously balance government policies and residents' needs, achieving sustainable urban land use planning.
Multi-Agent Communication with Information Preserving Graph Contrastive Learning
Wei Du (Shandong University), Lizhen Cui (Shandong University)
Representation LearningRecurrent Neural NetworkGraph Neural NetworkReinforcement LearningContrastive LearningGraph
🎯 What it does: This paper proposes a multi-agent communication mechanism based on information-preserving graph contrastive learning (MAIL) to enhance message representation quality in collaborative reinforcement learning.
Multi-Agent Corridor Generating Algorithm
Arseni Pertzovskiy (Ben Gurion University Of Negev), Ariel Felner (Ben Gurion University Of Negev)
OptimizationGraphBenchmark
🎯 What it does: Proposed a corridor-based multi-agent path planning algorithm MACGA and its integrated version with PIBT, MACGA+PIBT, for efficiently solving the multi-agent path finding (MAPF) problem in dense graphs.
Multi-granularity Knowledge Transfer for Continual Reinforcement Learning
Chaofan Pan (Southwestern University of Finance and Economics), Xin Yang (Southwestern University of Finance and Economics)
TransformerLarge Language ModelReinforcement LearningImageText
🎯 What it does: Propose a multi-grained knowledge transfer framework named MT-Core, which uses LLM to generate coarse-grained target sequences, employs fine-grained RL learning to realize specific actions, and achieves cross-task transfer through a policy library.
Multi-Label Text Classification with Label Attention Aware and Correlation Aware Contrastive Learning
Zhengzhong Zhu (Sichuan University), Jiangping Zhu (Sichuan University)
ClassificationTransformerContrastive LearningText
🎯 What it does: This paper proposes a unified context and label-aware framework called UCLAF to address the complex associations and partial overlaps between labels in multi-label text classification.
Multi-modal Anchor Gated Transformer with Knowledge Distillation for Emotion Recognition in Conversation
Jie Li (China University of Mining and Technology), Xuan Li (China University of Mining and Technology)
ClassificationRecognitionKnowledge DistillationTransformerPrompt EngineeringVideoTextMultimodalityAudio
🎯 What it does: Propose the MAGTKD framework, which integrates prompt learning, knowledge distillation, and a multi-modal anchor gating Transformer to efficiently fuse features from text, audio, and video modalities for emotion recognition.
Multi-Modal Point Cloud Completion with Interleaved Attention Enhanced Transformer
Chenghao Fang (Shanxi University), Feilong Cao
RestorationAutonomous DrivingTransformerMultimodalityPoint Cloud
🎯 What it does: Proposed a multimodal point cloud completion framework IAET, which jointly utilizes complete images and partial point clouds for 3D reconstruction.
Multi-Objective Neural Bandits with Random Scalarization
Ji Cheng (City University of Hong Kong), Qingfu Zhang (City University of Hong Kong)
OptimizationReinforcement LearningImage
🎯 What it does: This paper proposes a neural network-based multi-objective multi-armed bandit (MONB) framework, utilizing stochastic normalization to address the multi-objective weight allocation problem;
Multi-Omics Analysis for Cancer Subtype Inference via Unrolling Graph Smoothness Priors
Jielong Lu (Zhejiang University), Haishuai Wang (Zhejiang University)
ClassificationGraph Neural NetworkContrastive LearningBiomedical Data
🎯 What it does: This paper proposes the GTMancer framework, which achieves the fusion of multi-omics data from a global perspective for cancer subtyping by aligning and employing multi-scale graph attention mechanisms.
Multi-Organizational Scheduling: Individual Rationality, Optimality, and Complexity
Jiehua Chen (TU Wien), Christian Hatschka (TU Wien)
Optimization
🎯 What it does: This paper introduces individual rationality constraints under a multi-organization scheduling framework, studying the scheduling problem of minimizing global makespan (Cmax) and total completion time (CΣ), while ensuring that no organization's performance decreases due to cooperation.
Multi-player Multi-armed Bandits with Delayed Feedback
Jingqi Fan (Northeastern University), Linghe Kong (Shanghai Jiao Tong University)
OptimizationReinforcement Learning
🎯 What it does: In the multi-player multi-armed bandit problem, this paper proposes a decentralized learning algorithm called DDSE that considers random delayed feedback.
Multi-Scale Temporal Neural Network for Stock Trend Prediction Enhanced by Temporal Hyepredge Learning
Lingyun Song (Northwestern Polytechnical University), Xuequn Shang (Northwestern Polytechnical University)
Convolutional Neural NetworkTransformerGraphTime SeriesFinance Related
🎯 What it does: Propose the MSTNN framework based on multi-scale 3D convolution and temporal hypergraph attention to capture periodic fluctuations of individual stocks and higher-order associations at the industry level for predicting stock trends
Multi-Sourced Compositional Generalization in Visual Question Answering
Chuanhao Li (Beijing Institute Of Technology), Yunde Jia (Shenzhen Msu Bit University)
Representation LearningTransformerSupervised Fine-TuningMultimodalityBenchmarkRetrieval-Augmented Generation
🎯 What it does: Investigate the performance of multi-source compositional generalization in visual question answering, and propose a retrieval-enhanced training framework to improve the model's unified representation capability for cross-modal primitives.
Multi-Task Curriculum Graph Contrastive Learning with Clustering Entropy Guidance
Chusheng Zeng (Northwestern Polytechnical University), Xuelong Li (China Telecom)
Representation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposes a clustering entropy guided multi-task curriculum graph contrastive learning framework, CurGL, for unsupervised graph clustering.
Multi-view Clustering via Multi-granularity Ensemble
Jie Yang (University of Technology Sydney), Bingbing Jiang (Hangzhou Normal University)
🎯 What it does: Proposed a multi-view clustering framework called MGE based on multi-granularity label sets, which utilizes constrained hierarchical clustering to generate labels from fine-grained to coarse-grained levels. A co-association matrix is constructed through cross-view and cross-granularity fusion, followed by secondary clustering to obtain the final clustering results.
Multi-View Learning with Context-Guided Receptance for Image Denoising
Binghong Chen (Harbin Institute of Technology), Xiangqian Wu (Harbin Institute of Technology)
RestorationRecurrent Neural NetworkImage
🎯 What it does: Proposed a context-guided receptive field weighted key-value network CRWKV for real-world image denoising.
MultiDreamer3D: Multi-concept 3D Customization with Concept-Aware Diffusion Guidance
Wooseok Song (Ulsan National Institute of Science and Technology), Jaejun Yoo (Ulsan National Institute of Science and Technology)
GenerationData SynthesisPrompt EngineeringVision Language ModelDiffusion modelScore-based ModelGaussian SplattingTextPoint Cloud
🎯 What it does: Proposes MultiDreamer3D to address the issues of object missing and concept mixing in 3D multi-concept customization by adopting phased layout generation and concept-aware diffusion guidance;
Multimodal Cancer Survival Analysis via Hypergraph Learning with Cross-Modality Rebalance
Mingcheng Qu (Harbin Institute of Technology), Lei Fan (UNSW Sydney)
Representation LearningGraph Neural NetworkMultimodalityBiomedical Data
🎯 What it does: This paper proposes the MRePath framework, which utilizes sheaf-based hypergraph learning to extract contextual and hierarchical information from pathological images, and rebalances the contributions of pathological and genomic modalities through dynamic weighting and interactive alignment fusion techniques, ultimately achieving multi-modal cancer survival prediction.
Multimodal Fake News Detection: MFND Dataset and Shallow-Deep Multitask Learning
Ye Zhu (Hebei University of Technology), Zitong Yu (Great Bay University)
ClassificationObject DetectionTransformerVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: This paper proposes a multimodal fake news detection dataset named MFND, which includes 11 types of deepfake methods, and designs a shallow-deep multi-task learning (SDML) framework to detect news authenticity and locate image/text forgeries.
Multimodal Image Matching Based on Cross-Modality Completion Pre-training
Meng Yang (Wuhan University), Jiayi Ma (Wuhan University)
Pose EstimationRetrievalConvolutional Neural NetworkTransformerSupervised Fine-TuningContrastive LearningImageMultimodality
🎯 What it does: For the matching task of multi-modal images such as visible light and infrared, the XCP-Match scheme is proposed, adopting a two-stage training approach: first, self-supervised cross-modal pretraining is conducted on real multi-modal data, and then supervised fine-tuning is performed on the enhanced MegaDepth dataset, constructing a complete multi-modal multi-scale feature extraction, coarse-to-fine matching, and sub-pixel refinement network.
Multimodal Inference with Incremental Tabular Attributes
Xinda Chen (Fudan University), Bo Yan (Fudan University)
Auto EncoderContrastive LearningImageTabular
🎯 What it does: Proposed and implemented the MIITA framework to unsupervisedly integrate incremental tabular attributes into pre-trained multi-modal models during the inference phase, addressing the issue of dynamic changes in table columns.
Multimodal Inverse Attention Network with Intrinsic Discriminant Feature Exploitation for Fake News Detection
Tianlin Zhang (Shandong Normal University), Jiande Sun (Shandong Normal University)
ClassificationAnomaly DetectionTransformerLarge Language ModelImageTextMultimodality
🎯 What it does: Propose a Multimodal Inverse Attention Network (MIAN) for fake news detection, combining text and image features while explicitly extracting consistency and inconsistency information.
Multimodal Knowledge Retrieval-Augmented Iterative Alignment for Satellite Commonsense Conversation
Qian Li (Beijing University of Posts and Telecommunications), Shangguang Wang (Beihang University)
RetrievalTransformerLarge Language ModelVision Language ModelMultimodalityRetrieval-Augmented Generation
🎯 What it does: Proposed the Sat-RIA framework, combining multi-modal knowledge retrieval with iterative alignment to achieve satellite common-sense dialogue.
Multimodal Prior Learning with Double Constraint Alignment for Snapshot Spectral Compressive Imaging
Mingjin Zhang (Xidian University), Jie Guo (Xidian University)
RestorationVision Language ModelMultimodality
🎯 What it does: Proposed the CAMM framework, integrating textual description information into the snapshot spectral compressed imaging reconstruction network.
MVP-CBM: Multi-layer Visual Preference-enhanced Concept Bottleneck Model for Explainable Medical Image Classification
Chunjiang Wang (University Of Science And Technology Of China), S. Kevin Zhou (University Of Science And Technology Of China)
ClassificationExplainability and InterpretabilityTransformerLarge Language ModelVision Language ModelContrastive LearningBiomedical DataComputed TomographyUltrasound
🎯 What it does: This paper proposes a multi-layer visual preference enhanced concept bottleneck model (MVP-CBM), achieving explainable and high-precision medical image classification by capturing visual layer preferences for medical concepts and sparsely fusing multi-layer concept activations.
NAAST-GNN: Neighborhood Adaptive Aggregation and Spectral Tuning for Graph Anomaly Detection
Ronghui Guo (Tianjin University), Zhiyong Feng (Tianjin University)
Anomaly DetectionGraph Neural NetworkGraph
🎯 What it does: This paper proposes NAAST-GNN, a GNN model specifically designed for graph anomaly detection, which addresses the challenges of overlapping and discrete neighborhood distributions through two modules: neighborhood adaptive aggregation and spectral regulation.
Navigating Social Dilemmas with LLM-based Agents via Consideration of Future Consequences
Dung Nguyen (Deakin University), Truyen Tran (Deakin University)
TransformerLarge Language ModelAgentic AIPrompt EngineeringText
🎯 What it does: Utilize large language models (LLMs) to construct agents, investigating how considering future consequences (CFC) can enhance agents' sustainable cooperative behavior in text-based common resource games (Common Harvest) and more complex government simulations (GovSim).
Negative Metric Learning for Graphs
Yiyang Zhao (Sichuan University), Ning Yang (Sichuan University)
Representation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Propose a negative sample measurement learning (NML) framework called NML-GCL, which utilizes a learnable negative sample measurement network to distinguish real negative samples from fake negative samples in graph contrastive learning, significantly enhancing the discriminativeness of node representations and the performance of downstream tasks.
NeSyA: Neurosymbolic Automata
Nikolaos Manginas, Luc De Raedt (KU Leuven)
ClassificationRecognitionConvolutional Neural NetworkRecurrent Neural NetworkTransformerImageVideo
🎯 What it does: Developed a neural symbolic automaton called NESYA for sequence classification and annotation, integrating neural network perception with symbolic automata and providing probabilistic semantics.
NeuBM: Mitigating Model Bias in Graph Neural Networks Through Neutral Input Calibration
Jiawei Gu (Great Bay University), Xiao Luo (University of California)
ClassificationData-Centric LearningGraph Neural NetworkGraph
🎯 What it does: Propose the NeuBM method, which constructs a dynamic neutral graph and performs difference calibration on logits to mitigate model bias in GNNs under class imbalance scenarios.
Neuromorphic Sequential Arena: A Benchmark for Neuromorphic Temporal Processing
Xinyi Chen (Hong Kong Polytechnic University), Jibin Wu (Hong Kong Polytechnic University)
Spiking Neural NetworkTransformerTime SeriesSequentialBenchmark
🎯 What it does: Proposed the Neuromorphic Sequential Arena (NSA) as a multi-task, application-oriented neuromorphic temporal processing benchmark, and developed the Segregated Temporal Probe (STP) tool to verify the temporal dependency of tasks, followed by systematic evaluations of various synapse neuron models and network architectures on NSA.
Neuron Similarity-Based Neural Network Verification via Abstraction and Refinement
Yuehao Liu (Xidian University), Cong Tian (Xidian University)
Computational EfficiencyAdversarial AttackImageBenchmark
🎯 What it does: This paper proposes an abstraction-refinement method based on neuron similarity to reduce the scale of deep neural networks (DNNs), thereby improving the efficiency of formal verification;
New Algorithms for #2-SAT and #3-SAT
Junqiang Peng (University of Electronic Science and Technology of China), Mingyu Xiao (University of Electronic Science and Technology of China)
OptimizationComputational Efficiency
🎯 What it does: Proposed new weighted #2-SAT and #3-SAT counting algorithms, and provided an upper bound on the worst-case runtime with the number of clauses m as the parameter.
New Sequence-Independent Lifting Techniques for Cover Inequalities and When They Induce Facets
Siddharth Prasad (Carnegie Mellon University), Tuomas Sandholm (Carnegie Mellon University)
OptimizationTabular
🎯 What it does: This paper studies sequence-independent lifting methods for covering inequalities in integer programming and proposes a new piecewise constant (PC) lifting technique, proving that it can generate face-defining cuts under specific conditions;
No Regret Reinforcement Learning Algorithms for Online Scheduling with Multi-Stage Tasks
Yongxin Xu (ShanghaiTech University), Xin Liu (ShanghaiTech University)
OptimizationReinforcement Learning
🎯 What it does: Study the online scheduling problem for multi-stage tasks, proposing the RM VI 2 algorithm to achieve sublinear regret without degradation.
Noise Optimized Conditional Diffusion for Domain Adaptation
Lingkun Luo (Shanghai Jiao Tong University), Liming Chen (Ecole Centrale de Lyon)
Domain AdaptationDiffusion modelImageTime Series
🎯 What it does: Propose a framework named NOCDDA that integrates conditional diffusion models with unsupervised domain adaptation (UDA) decision-making to generate high-confidence pseudo-labels in the target domain and achieve cross-domain feature alignment.
Noise-Resistant Label Reconstruction Feature Selection for Partial Multi-Label Learning
Wanfu Gao (Jilin University), Kunpeng Liu (Portland State University)
ClassificationImageBenchmark
🎯 What it does: To address the challenges of noisy labels and sparse positive samples in partial multi-label learning, the authors propose a three-stage feature selection framework: ① Reconstruct the label set using a label mutual information matrix to remove noise and enhance label reliability; ② Apply an improved low-rank assumption and graph Laplacian regularization on the reconstructed labels to learn a weight matrix while preserving the high-dimensional structure; ③ Further refine the weight matrix by leveraging label mutual information reconstruction to emphasize important labels, thereby enhancing the identification of positive labels.
Non-collective Calibrating Strategy for Time Series Forecasting
Bin Wang (Ocean University of China), Yanwei Yu (Ocean University of China)
OptimizationTransformerTime SeriesBenchmarkPhysics Related
🎯 What it does: Proposed a universal, model-agnostic calibration strategy called Socket+Plug (SoP), achieving non-ensemble calibration of trained deep time series models by training a Plug for each prediction target individually.
Non-expansive Fuzzy ALC
Stefan Gebhart (Friedrich-Alexander-Universität Erlangen-Nürnberg), Paul Wild (Friedrich-Alexander-Universität Erlangen-Nürnberg)
🎯 What it does: Proposed a new fuzzy description logic called non-expansive fuzzy ALC, and provided an unlabelled tableau computation method for general TBox, achieving EXPTIME complexity for concept satisfiability checking under this logic.
Non-Obvious Manipulability in Additively Separable and Fractional Hedonic Games
Diodato Ferraioli (University of Salerno), Giovanna Varricchio (University of Calabria)
Optimization
🎯 What it does: This paper studies the design of non-manipulable (NOM) mechanisms for agents lacking full rationality (i.e., without conditional reasoning) in additive decomposable and fractional hedonic games, aiming to achieve optimal or near-optimal social welfare while ensuring non-manipulability;
Not All Layers of LLMs Are Necessary During Inference
Siqi Fan (University Of Electronic Science And Technology Of China), Yequan Wang (Beijing Academy Of Artificial Intelligence)
ClassificationComputational EfficiencyTransformerLarge Language ModelText
🎯 What it does: This paper proposes AdaInfer, an input-instance-based adaptive inference algorithm that can dynamically decide when to terminate inference early without modifying LLM parameters;
Not in My Backyard! Temporal Voting Over Public Chores
Edith Elkind (Northwestern University), Nicholas Teh (University of Oxford)
Optimization
🎯 What it does: Studied a temporal voting model in public works voting, exploring how voters dynamically trade off preferences through disapproval expressions, and conducted theoretical analysis on two welfare objectives (minimizing total dissatisfaction and minimizing the maximum individual voter dissatisfaction);
NS4S: Neighborhood Search for Scheduling Problems Via Large Language Models
Junjie Zhang (Huazhong University of Science and Technology), Yan Jin (Huazhong University of Science and Technology)
OptimizationTransformerLarge Language ModelTabular
🎯 What it does: Proposed a neighborhood search framework NS4S based on large language models for solving JSP, FJSP, and FJSP-SDST.
NuMDS: An Efficient Local Search Algorithm for Minimum Dominating Set Problem
Rui Sun (Northeast Normal University), Jiejiang Chen (Hangzhou Normal University)
OptimizationGraphBenchmark
🎯 What it does: Propose an efficient local search algorithm called NuMDS for solving the minimum dominating set (MDS) problem.
Object-Level Backdoor Attacks in RGB-T Semantic Segmentation with Cross-Modality Trigger Optimization
Xianghao Jiao, Xiaochun Cao (Sun Yat-Sen University)
SegmentationAdversarial AttackMultimodality
🎯 What it does: This paper proposes an object-level backdoor attack framework called OBA for RGB-T semantic segmentation, achieving precise manipulation of individual target objects.
Odyssey : Empowering Minecraft Agents with Open-World Skills
Shunyu Liu (Zhejiang University), Mingli Song (Zhejiang University)
Robotic IntelligenceTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningAgentic AITextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Developed the ODYSSEY framework, which enables multi-task, long-term planning, and autonomous exploration in Minecraft through an interactive agent based on large language models and an open-world skill library containing 40 primitive skills and 183 composite skills;
Omni-Dimensional State Space Model-driven SAM for Pixel-level Anomaly Detection
Chao Huang (University of Macau), Bob Zhang (University of Macau)
Anomaly DetectionConvolutional Neural NetworkTransformerImageBiomedical Data
🎯 What it does: Propose a pixel-level anomaly detection method ODS-SAM based on the Segment Anything Model (SAM), which automatically generates multi-scale global prompts.
OMS: One More Step Noise Searching to Enhance Membership Inference Attacks for Diffusion Models
Xiaomeng Fu (Chinese Academy of Sciences), Xingyu Gao (Chinese Academy of Sciences)
GenerationSafty and PrivacyConvolutional Neural NetworkTransformerDiffusion modelImage
🎯 What it does: This paper studies member inference attacks on diffusion models, proposing to convert member inference into an optimization problem for finding training noise, solving the problem via fixed-point iteration, and introducing an improved method called 'One More Step (OMS)' to enhance discriminative performance.
On Definite Iterated Belief Revision with Belief Algebras
Hua Meng (Southwest Jiaotong University), Zhengchun Zhou (Southwest Jiaotong University)
🎯 What it does: Proposes an iterative belief revision framework based on Belief Algebra, which can unify the current belief state, new evidence, and revision results as belief algebra.
On Independence and SCC-Recursiveness in Assumption-Based Argumentation
Lydia Blümel, Francesca Toni (Imperial College London)
Review/Survey Paper
🎯 What it does: Studied the concept of conditional independence in assumption-based argumentation (ABA), proving that it satisfies the semi-Johnson-Groves property and establishing an equivalence relationship with independence in abstract argumentation frameworks (AF).
On Integrating Logical Analysis of Data into Random Forests
David Ing (CRIL, Université d'Artois), Lakhdar Saïs (CRIL, Université d'Artois)
Explainability and InterpretabilityComputational EfficiencyTabularBenchmark
🎯 What it does: This paper introduces the Minimum Support Sets (MSSes) from the Logical Analysis of Data (LAD) method into Random Forest (RF), replacing traditional random feature subsets with MSSes to construct more discriminative decision trees, reduce feature redundancy, and enhance model interpretability.
On Middle Grounds for Preference Statements
Anne-Marie George (University of Oslo), Ana Ozaki (University of Oslo)
🎯 What it does: This paper proposes and formalizes the concept of finding a middle ground in preference statements (e.g., 'I prefer a over b'), providing general definitions, existence conditions, and construction algorithms.
On Temporal ASP with Eager Unfoldable Operators
Thomas Eiter (Vienna University of Technology), Davide Soldà (Vienna University of Technology)
🎯 What it does: Proposed the 'eager unfolding' temporal operator and defined non-splitting temporal programs (NDTP) and compact temporal programs (TTP), enabling polynomial translation of these programs into LTL, thereby reducing the complexity of original TEL and allowing the use of existing LTL model checkers.
On the Discrimination and Consistency for Exemplar-Free Class Incremental Learning
Tianqi Wang (Hong Kong Polytechnic University), Zhi Chen (University of Southern Queensland)
ClassificationRepresentation LearningContrastive LearningImage
🎯 What it does: In sample-free class-incremental learning (EF-CIL), this paper emphasizes the importance of discriminativeness and consistency in the feature space through theoretical analysis, and proposes the DCNet framework, which achieves stable and discriminative feature spaces via orthogonal embedding and dynamic aggregation compensation.
On the Generalization of Feature Incremental Learning
Chao Xu (National University of Defense Technology), Chenping Hou (National University of Defense Technology)
ClassificationTabular
🎯 What it does: This paper systematically summarizes four mainstream strategies for feature incremental learning (feature trimming, data adaptation, model reuse, and data reconstruction), and derives upper bounds on generalization error for each strategy, providing theoretical guidance and experimental validation.
On the Learning with Augmented Class via Forests
Fan Xu (Nanjing University), Wei Gao (Nanjing University)
ClassificationImageTabular
🎯 What it does: This paper proposes two methods, LACForest and Deep Neural LACForest, which proactively mine unseen classes (augmented classes) in decision trees/forests using the 'enhanced Gini impurity' criterion, and further improve classification performance through pseudo labels.