IJCAI 2025 Papers — Page 4
International Joint Conference on Artificial Intelligence · 1014 papers
EAVIT: Efficient and Accurate Human Value Identification From Text Data via LLMs
Wenhao Zhu (Peking University), Xin Zhang (Peking University)
ClassificationExplainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextChain-of-Thought
🎯 What it does: Propose the EAVIT framework, which combines locally tunable LLMs and online LLMs to efficiently and accurately identify human values in text data.
ECC-SNN: Cost-Effective Edge-Cloud Collaboration for Spiking Neural Networks
Di Yu (Zhejiang University), Shuiguang Deng (Zhejiang University)
ClassificationComputational EfficiencyKnowledge DistillationSpiking Neural NetworkImage
🎯 What it does: Proposes the ECC-SNN framework, which integrates SNN with cloud-based ANN to collaborate, leveraging the low energy consumption of edge devices and the high precision of the cloud, supporting incremental learning and dynamic environmental adaptation.
EchoGPT: An Interactive Cardiac Function Assessment Model for Echocardiogram Videos
Bo Xu (Dalian University of Technology), Feng Xia (RMIT University)
SegmentationConvolutional Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelVideoTextBiomedical DataUltrasound
🎯 What it does: Developed and validated an interactive cardiac function assessment system, EchoGPT, which integrates large language models with medical diagnostic models to automatically segment cardiac ultrasound videos, predict ejection fraction, and answer user questions.
EDGE: Efficient Data Selection for LLM Agents via Guideline Effectiveness
Yunxiao Zhang (Peking University), Wen Zhao (Peking University)
Computational EfficiencyData-Centric LearningReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningAgentic AIPrompt EngineeringText
🎯 What it does: Proposed the EDGE method, which automatically selects high-information samples without golden answers for prompt engineering and LLM fine-tuning by using the guidance effectiveness (GE) metric.
EDyGS: Event Enhanced Dynamic 3D Radiance Fields from Blurry Monocular Video
Mengxu Lu (Zhejiang University), Gang Pan (Zhejiang University)
RestorationGenerationNeural Radiance FieldGaussian SplattingVideoMultimodality
🎯 What it does: Construct the EDyGS model to achieve dynamic scene deblurring and novel view synthesis by leveraging event cameras and blurry monocular videos.
EF1 and EFX Orientations
Argyrios Deligkas (Royal Holloway University of London), Viktoriia Korchemna (TU Wien)
Optimization
🎯 What it does: The study addresses the fair division problem under the constraint that each agent can only receive items (orientations) they approve of. It proves the existence of EF1 for monotonic valuations and provides a pseudo-polynomial algorithm; for the EFX direction, it conducts complexity analysis, demonstrating that it remains NP-hard under various constraints, and proposes an FPT algorithm parameterized by slim tree-cut width.
Efficient Algorithms for Electing Successive Committees
Pallavi Jain (Indian Institute of Technology), Andrzej Kaczmarczyk (University of Chicago)
Computational Efficiency
🎯 What it does: Proposed and implemented multiple fixed-parameter tractable (FPT) algorithms for the Successive Committees Election model, addressing the previously proven NP-hard cases where the committee size is three or more.
Efficient and Rigorous Model-Agnostic Explanations
Joao Marques-Silva (ICREA and University of Lleida), Maria Vanina Martinez (Artificial Intelligence Research Institute (IIIA), CSIC)
Explainability and InterpretabilityComputational EfficiencyTabular
🎯 What it does: This paper proposes efficient algorithms for abductive and contrastive explanations in sample-based explainable AI (XAI), along with complexity analysis and experimental validation.
Efficient Constraint-based Window Causal Graph Discovery in Time Series with Multiple Time Lags
Yewei Xia (Fudan University), Shuigeng Zhou (Fudan University)
Computational EfficiencyTime SeriesBiomedical Data
🎯 What it does: Propose a constraint-based multi-delay window causal graph discovery method called CDiT, which reduces the number of CI tests by minimizing trek lag, thereby improving efficiency;
Efficient Counterexample-Guided Fairness Verification and Repair of Neural Networks Using Satisfiability Modulo Convex Programming
Arya Fayyazi (University of Southern California), Massoud Pedram (University of Southern California)
OptimizationExplainability and InterpretabilityComputational EfficiencyTabularFinance Related
🎯 What it does: This work proposes the FaVeR framework, which verifies individual fairness of pre-trained neural networks using Satisfiability Modulo Convex Programming (SMC), and repairs fairness through reverse weight adjustment targeting high-sensitivity neurons.
Efficient Differentiable Approximation of Generalized Low-rank Regularization
Naiqi Li (Tsinghua Shenzhen International Graduate School), Shu-Tao Xia (Tsinghua Shenzhen International Graduate School)
RestorationOptimizationComputational EfficiencyImageVideo
🎯 What it does: Proposes a differentiable general low-rank regularization approximation method that can be directly embedded into deep learning loss functions and optimized via gradient descent.
Efficient Diversity-based Experience Replay for Deep Reinforcement Learning
Kaiyan Zhao (Wuhan University), Xiaoguang Niu (Wuhan University)
Reinforcement LearningImage
🎯 What it does: Propose an efficient diversity-driven experience replay framework called EDER, which uses a deterministic point process to evaluate trajectory diversity and combines Cholesky decomposition and rejection sampling techniques, significantly improving sample efficiency and performance of deep reinforcement learning in high-dimensional state spaces.
Efficient Dynamic Ensembling for Multiple LLM Experts
Jinwu Hu (South China University of Technology), Mingkui Tan (South China University of Technology)
Computational EfficiencyReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningAgentic AIPrompt EngineeringMixture of ExpertsText
🎯 What it does: Proposed an efficient dynamic ensemble reasoning (DER) framework that integrates complementary knowledge from multiple large language model (LLM) experts by training an agent to sequentially invoke them on input questions, generating higher-quality answers.
Efficient Dynamic Graphs Learning with Refined Batch Parallel Training
Zhengzhao Feng (Zhejiang University), Mingli Song (Zhejiang University)
Computational EfficiencyGraph Neural NetworkGraphTime Series
🎯 What it does: Propose the RBT framework to address the memory staleness problem in memory-based temporal graph neural networks (MTGNN) during batch parallel training;
Efficient Hi-Fi Style Transfer via Statistical Attention and Modulation
Zhirui Fang (Dalian University of Technology), Yanqing Guo (Dalian University of Technology)
Image TranslationComputational EfficiencyConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: Propose an efficient and high-fidelity style transfer framework SRCA-SM, integrating statistical row-column attention, statistical modulation, and contrastive learning to achieve fast inference and content detail preservation.
Efficient Inter-Operator Scheduling for Concurrent Recommendation Model Inference on GPU
Shuxi Guo (Beijing University of Posts and Telecommunications), Jingyu Wang (Beijing University of Posts and Telecommunications)
Recommendation SystemComputational Efficiency
🎯 What it does: Propose RecOS, a scheduling system for concurrent inference of recommendation models (RM) on GPUs, addressing issues caused by traditional single-stream or topology-based multi-stream scheduling, such as operator queuing, cache conflicts, and low GPU resource utilization.
Efficient Multi-view Clustering via Reinforcement Contrastive Learning
Qianqian Wang (Xidian University), Quanxue Gao (Xidian University)
Representation LearningReinforcement LearningAuto EncoderContrastive LearningMultimodality
🎯 What it does: Propose an efficient multi-view clustering framework EMVC-RCL, which utilizes reinforced contrastive learning to dynamically optimize the clustering process, combined with a memory-enhanced mechanism to achieve feature enhancement and adaptive correction of uncertain samples.
Efficient Quantum Approximate kNN Algorithm via Granular-Ball Computing
Shuyin Xia (Chongqing University of Posts and Telecommunications), Jeremiah D. Deng (University of Otago)
OptimizationComputational Efficiency
🎯 What it does: Propose a GB-Q k-NN algorithm that combines Granular-Ball theory, HNSW graph, and quantum computing for efficient nearest neighbor search on large-scale datasets.
Efficient Visual Representation Learning with Heat Conduction Equation
Zhemin Zhang (Southwest Jiaotong University), Xun Gong (Southwest Jiaotong University)
ClassificationObject DetectionSegmentationRepresentation LearningConvolutional Neural NetworkImagePhysics RelatedOrdinary Differential Equation
🎯 What it does: Proposed a visual representation learning framework HcNet based on the heat conduction equation, designed the Heat Conduction Layer and Refinement Approximation Layer, and constructed an interpretable visual backbone network.
EfficientPIE: Real-Time Prediction on Pedestrian Crossing Intention with Sole Observation
Fang Qu (Chongqing University), Songtao Guo (Chongqing University)
Autonomous DrivingComputational EfficiencyConvolutional Neural NetworkImage
🎯 What it does: Propose the EfficientPIE framework, which uses single-frame images to predict in real-time whether pedestrians have the intention to cross the street;
EFormer: An Effective Edge-based Transformer for Vehicle Routing Problems
Dian Meng (Dalian University of Technology), Qiang Zhang (Dalian University of Technology)
OptimizationGraph Neural NetworkTransformerReinforcement LearningGraph
🎯 What it does: Designed and trained an edge-information-based Transformer model (EFormer) for autoregressive solving of vehicle routing problems.
EFX Feasible Scheduling for Time-dependent Resources
Jiazhu Fang (Ocean University of China), Wenjing Liu (Ocean University of China)
Optimization
🎯 What it does: Studied the fair interval scheduling problem (FISP) for time-dependent resources, extending fairness metrics to the stronger EFX and incorporating efficiency metrics (MaxNSW, PO, WIO).
Egocentric Object-Interaction Anticipation with Retentive and Predictive Learning
Guo Chen (Nanjing University), Tong Lu (Nanjing University)
RecognitionObject DetectionKnowledge DistillationTransformerPrompt EngineeringVision-Language-Action ModelVideoText
🎯 What it does: Propose the EgoAnticipator framework, combining Retentive Pre-Training and Predictive Pre-Training, and introduce Mirror Distillation and Long-Term Memory Prompting to enhance short-term object interaction prediction performance in first-person perspective scenarios.
ElaD-Net: An Elastic Semantic Decoupling Network for Lesion Segmentation in Breast Ultrasound Images
Lijuan Xu (Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Shandong Computer Science Center (National Supercomputer Center in Jinan), Qilu University of Technology (Shandong Academy of Sciences)), Dawei Zhao (Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Shandong Computer Science Center (National Supercomputer Center in Jinan), Qilu University of Technology (Shandong Academy of Sciences))
SegmentationConvolutional Neural NetworkBiomedical DataUltrasound
🎯 What it does: This study proposes an elastic semantic decoupling network called ElaD-Net for automatically segmenting lesion regions in breast ultrasound images, addressing challenges such as similar intensity between foreground and background pixels and inconsistent multi-scale detail transmission.
Electron Density-enhanced Molecular Geometry Learning
Hongxin Xiang (Hunan University), Xiangxiang Zeng (Hunan University)
Knowledge DistillationDrug DiscoveryGraph Neural NetworkTransformerAuto EncoderImageMultimodalityGraphPhysics Related
🎯 What it does: Proposed the EDG framework, which enhances molecular geometry learning by utilizing multi-view RGB-D electron density images.
Empowering Multimodal Road Traffic Profiling with Vision Language Models and Frequency Spectrum Fusion
Haolong Xiang, Wei Fan (University of Auckland)
Autonomous DrivingTransformerPrompt EngineeringVision Language ModelContrastive LearningMultimodality
🎯 What it does: Proposes the TraffiCFUS framework, which utilizes image and text bimodalities for traffic road analysis, combining VLM-generated text, spectral transformation, and contrastive learning to achieve multi-modal fusion.
Empowering Vision Transformers with Multi-Scale Causal Intervention for Long-Tailed Image Classification
Xiaoshuo Yan (Shandong University), Xiangxu Meng (Shandong University)
ClassificationTransformerImage
🎯 What it does: Proposed a two-stage causal debiasing method named TSCNet, which first enhances ViT's fine-grained causal representations for tail classes through hierarchical causal interventions at the patch-level and feature-level, and then reduces semantic and distributional biases in long-tailed image classification by generating counterfactual distributions via adversarial Fourier perturbations and adaptively calibrating logit biases.
Endogenous Recovery via Within-modality Prototypes for Incomplete Multimodal Hashing
Sa Zhu (Chinese Academy of Sciences), Bo Li (Chinese Academy of Sciences)
RetrievalMultimodality
🎯 What it does: Proposes a modality completion hashing method PMCH based on intra-modal prototypes, effectively restoring missing modalities and enhancing multi-modal retrieval performance.
Endowing Interpretability for Neural Cognitive Diagnosis by Efficient Kolmogorov-Arnold Networks
Shangshang Yang (Anhui University), Ye Tian (Anhui University)
Explainability and InterpretabilityTabularSequential
🎯 What it does: Proposed KAN2CD, which utilizes Kolmogorov-Arnold networks to enhance the explainability of neural cognitive diagnostic models while maintaining or even improving prediction accuracy.
EnergyCompress: A General Case Base Learning Strategy
Fadi Badra (Université Sorbonne Paris Nord), David Leake (Indiana University)
ClassificationCompressionTabular
🎯 What it does: Proposed a general case-based compression learning strategy called EnergyCompress, which evaluates the case base using an energy model and iteratively removes the weakest cases to obtain a smaller and more optimal case subset.
Enhanced Graph Similarity Learning via Adaptive Multi-scale Feature Fusion
Cuifang Zou (Guangxi Normal University), Shichao Zhang (Guangxi Normal University)
Representation LearningGraph Neural NetworkGraph
🎯 What it does: Proposed the AMFF framework for graph similarity learning based on adaptive multi-scale feature fusion.
Enhanced Unsupervised Discriminant Dimensionality Reduction for Nonlinear Data
Qianqian Wang (Xidian University), Zhengming Ding (Tulane University)
OptimizationRepresentation LearningImage
🎯 What it does: Propose an unsupervised dimensionality reduction and clustering method that integrates centerless K-means with linear discriminant analysis (LDA), eliminating the dependency on centroids in traditional K-means while simultaneously capturing local neighborhood and discriminative structures;
Enhancing Chemical Reaction and Retrosynthesis Prediction with Large Language Model and Dual-task Learning
Xuan Lin (Xiangtan University), Xiangxiang Zeng (Hunan University)
Drug DiscoveryTransformerLarge Language ModelTextBiomedical DataBenchmark
🎯 What it does: For the chemical reaction and retrosynthesis prediction tasks in drug discovery, this paper proposes ChemDual, an enhanced LLaMA framework, which achieves precise modeling of the chemical synthesis process by leveraging a large-scale instruction dataset and dual-task learning.
Enhancing Counterfactual Estimation: A Focus on Temporal Treatments
Xin Wang (University of Science and Technology of China), Chunyan Miao (Nanyang Technological University)
Computational EfficiencyRepresentation LearningRecurrent Neural NetworkAuto EncoderTabularTime SeriesBiomedical DataElectronic Health Records
🎯 What it does: Constructed the CTD-NKO model based on the Koopman operator and RNN for multi-timepoint causal counterfactual estimation, focusing on capturing interactions in time series treatments.
Enhancing Long-Tail Bundle Recommendations Utilizing Composition Pattern Modeling
Tianhui Ma (University of Science and Technology of China), Hui Xiong (Hong Kong University of Science and Technology)
Recommendation SystemGraph Neural NetworkContrastive LearningSequential
🎯 What it does: This paper addresses the long-tail bundle recommendation problem by proposing the CALBRec framework, which enhances the recommendation performance of long-tail bundles through a combination pattern-aware adapter and dual-view prototype learning.
Enhancing Mixture of Experts with Independent and Collaborative Learning for Long-Tail Visual Recognition
Yanhao Chen (Xiamen University), Qingqiang Wu (Xiamen University)
ClassificationRecognitionMixture of ExpertsContrastive LearningImage
🎯 What it does: A framework for independent and collaborative learning (ICL) is proposed to address long-tailed visual recognition tasks, achieving independent optimization of each expert while maintaining their advantages within a Mixture of Experts (MoE) model to enhance overall performance.
Enhancing Multimodal Model Robustness Under Missing Modalities via Memory-Driven Prompt Learning
Yihan Zhao (Xi'an Jiaotong University), Jizhong Zhao (Xi'an Jiaotong University)
Representation LearningTransformerSupervised Fine-TuningPrompt EngineeringMultimodality
🎯 What it does: Propose the Memory-Driven Prompt Learning framework, using generative prompts and shared prompts to compensate for missing modalities, enabling the unified Transformer to maintain robustness under missing modalities.
Enhancing Multimodal Protein Function Prediction Through Dual-Branch Dynamic Selection with Reconstructive Pre-Training
Xiaoling Luo (Shenzhen University), Junsong Wang (Shenzhen Technology University)
Drug DiscoveryTransformerLarge Language ModelMixture of ExpertsMultimodalityBiomedical Data
🎯 What it does: This paper proposes a dual-branch, reconstructive pre-training based multi-modal protein function prediction framework called DSRPGO.
Enhancing Nighttime Semantic Segmentation with Visual-Linguistic Priors and Wavelet Transform
Jianhou Zhou (Hangzhou Dianzi University), Xiaoqin Zhang (Zhejiang University Of Technology)
SegmentationTransformerVision Language ModelImageText
🎯 What it does: This paper proposes an end-to-end nighttime semantic segmentation framework called Text-WaveletFormer, which combines text prompts and wavelet transforms to improve segmentation performance under low-light conditions.
Enhancing Sampling Protocol for Point Cloud Classification Against Corruptions
Chongshou Li (Southwest Jiaotong University), Xinke Li (City University of Hong Kong)
ClassificationPoint Cloud
🎯 What it does: Propose an improved point cloud sampling protocol called PointSP to enhance robustness against noise and missing data;
Enhancing Semantic Clarity: Discriminative and Fine-grained Information Mining for Remote Sensing Image-Text Retrieval
Yu Liu (Jilin University), Yingda Lyu (Jilin University)
RetrievalTransformerVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Propose the DFIM model for remote sensing image-text retrieval, addressing semantic confusion caused by visual redundancy and inter-class similarity.
Enhancing Table Recognition with Vision LLMs: A Benchmark and Neighbor-Guided Toolchain Reasoner
Yitong Zhou (University of Science and Technology of China), Xin Li (iFLYTEK Co., Ltd)
RecognitionTransformerVision Language ModelImageBenchmarkRetrieval-Augmented Generation
🎯 What it does: Proposed a Vision LLM-based table recognition evaluation benchmark and designed the Neighbor-Guided Toolchain Reasoner (NGTR) framework to enhance table recognition performance in no-training scenarios.
Enhancing the Performance of Global Model by Improving the Adaptability of Local Models in Federated Learning
Wujun Zhou (Nanjing University), Wei Wang (Nanjing University)
ClassificationFederated LearningImage
🎯 What it does: Propose to improve the performance of the global model in federated learning by enhancing the adaptability of local models across different client distributions, designing an adaptive constraint-based local training objective and aggregation strategy.
Enhancing Transferability of Audio Adversarial Example for Both Frequency- and Time-domain
Zilin Tian (Harbin Engineering University), Jiahong Zhao (University of Southampton)
ClassificationAdversarial AttackAudio
🎯 What it does: Propose the Adaptive Inter-domain Ensemble (AIE) attack method, which generates transferable audio adversarial examples by leveraging complementary information between time-domain and frequency-domain models.
Enhancing User-Oriented Proactivity in Open-Domain Dialogues with Critic Guidance
Yufeng Wang (South China University of Technology), Mingkui Tan (South China University of Technology)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes a user-oriented proactive chatbot (UPC) that enhances the proactiveness of conversations by learning user backgrounds and guiding dialogues toward user interests.
Equitable Mechanism Design for Facility Location
Toby Walsh
Optimization
🎯 What it does: In the facility location problem without using money, a series of strategy-oblivious mechanisms are designed and analyzed to achieve limited approximation in fairness (complementary Gini index).
ESBN: Estimation Shift of Batch Normalization for Source-free Universal Domain Adaptation
Jiao Li (University of Electronic Science and Technology of China), Jingcai Guo (Hong Kong Polytechnic University)
Domain AdaptationConvolutional Neural NetworkImage
🎯 What it does: Propose the ESBN framework by analyzing the estimation bias of batch normalization (BN), selectively replacing BN layers with batch-free normalization (BFN), and combining instance statistics and Gaussian Mixture Model (GMM) to achieve source unsupervised domain adaptation.
Escaping Saddle Point Efficiently in Minimax and Bilevel Optimizations
Wenhan Xian (University of Maryland), Heng Huang (University of Maryland)
OptimizationComputational Efficiency
🎯 What it does: Proposes an algorithm named PRGDA, capable of finding second-order optimal points in stochastic non-convex-strongly concave min-max and non-convex-strongly convex bilevel optimization problems;
Evaluating and Mitigating Linguistic Discrimination in Large Language Models: Perspectives on Safety Equity and Knowledge Equity
Guoliang Dong (Singapore Management University), Xinyu Wang (Zhejiang University)
Safty and PrivacyAdversarial AttackLarge Language ModelTextBenchmark
🎯 What it does: Evaluate the safety and quality bias of large language models in multilingual scenarios, and propose a lightweight consistency voting method called LDFighter to mitigate this bias.
EVICheck: Evidence-Driven Independent Reasoning and Combined Verification Method for Fact-Checking
Lingxiao Wang (Communication University of China), Zeyu Li (Communication University of China)
ClassificationTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: Propose the EVICheck method, which independently reasons about each piece of evidence and combines fine-grained authenticity criteria to achieve automated fact-checking.
Evolutionary Algorithms Are Significantly More Robust to Noise When They Ignore It
Denis Antipov (Sorbonne Universit' e), Benjamin Doerr (Institut Polytechnique de Paris)
OptimizationBenchmark
🎯 What it does: Analyzes how the (1+1) evolutionary algorithm optimizes the LeadingOnes problem without re-evaluation in a priori noisy environment.
Evolvable Conditional Diffusion
Zhao Wei (Agency for Science, Technology and Research), Yew-Soon Ong (Agency for Science, Technology and Research)
OptimizationDiffusion modelScore-based ModelPhysics Related
🎯 What it does: This paper proposes an evolutionary conditional diffusion method that effectively guides the diffusion process under non-differentiable multi-physics black-box models (e.g., CFD and electromagnetic simulations), enabling the generation of design samples that meet specific optimization objectives.
Exact Algorithms with New Upper Bounds for the Maximum k-plex Problem
Jiongzhi Zheng (Huazhong University of Science and Technology), Kun He (Huazhong University of Science and Technology)
OptimizationGraph
🎯 What it does: In the branch-and-bound framework for the maximum k-plex problem (MKP), the authors propose two new upper bounds: RelaxGCB (a relaxed upper bound based on graph coloring) and RelaxPUB (a combination of RelaxGCB and the latest partitioning upper bound), using them to improve eight existing exact solvers.
Expanding the Category of Classifiers with LLM Supervision
Derui Lyu (University of Science and Technology of China), Huanhuan Chen (University of Science and Technology of China)
ClassificationRepresentation LearningLarge Language ModelVision Language ModelAuto EncoderContrastive LearningImageText
🎯 What it does: Propose a fully image-free and human-intervention-free classifier expansion framework called CEMIL, which is entirely based on large language models and can automatically generate multi-perspective descriptions and map them to existing classifier weights.
ExpertDiff: Head-less Model Reprogramming with Diffusion Classifiers for Out-of-Distribution Generalization
Jee Seok Yoon (Korea University), Heung-Il Suk (Korea University)
Domain AdaptationPrompt EngineeringDiffusion modelImageBiomedical DataUltrasound
🎯 What it does: This paper proposes an output-layer-free model reprogramming method called ExpertDiff, which leverages a diffusion classifier and learnable prompts to rapidly adapt the Stable Diffusion model in cross-domain OOV tasks.
Explainable Graph Neural Networks via Structural Externalities
Lijun Wu (University of Electronic Science and Technology of China), Zhiyi Fan (University of Electronic Science and Technology of China)
Explainability and InterpretabilityGraph Neural NetworkTextGraphBiomedical Data
🎯 What it does: Propose a GNN explainability framework called GraphEXT based on collaborative game theory and externality concepts, which measures node importance by calculating their marginal contributions in different subgraphs, directly explaining tasks such as graph classification, node classification, and link prediction;
Explainable Graph Representation Learning via Graph Pattern Analysis
Xudong Wang (Chinese University of Hong Kong, Shenzhen), Jicong Fan (Chinese University of Hong Kong, Shenzhen)
Explainability and InterpretabilityRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: This paper proposes a method for learning interpretable graph representations through graph pattern analysis, constructing two pattern-based interpretable learning frameworks, PXGL-EGK and PXGL-GNN.
Explaining Black-box Model Predictions via Two-level Nested Feature Attributions with Consistency Property
Yuya Yoshikawa (Chiba Institute of Technology), Yuki Saito (ZOZO Research)
ClassificationExplainability and InterpretabilityImageText
🎯 What it does: Propose a model-agnostic local explanation method called C2FA, which can simultaneously estimate the high-level feature importance (HiFA) and low-level feature importance (LoFA) of nested inputs
Exploiting Label Skewness for Spiking Neural Networks in Federated Learning
Di Yu (Zhejiang University), Shuiguang Deng (Zhejiang University)
Federated LearningSafty and PrivacyKnowledge DistillationSpiking Neural NetworkImageTime Series
🎯 What it does: This paper proposes the FedLEC framework for federated learning on label-skewed edge devices to train deep spiking neural networks (SNN) while preserving data privacy.
Exploiting Position Information in Convolutional Kernels for Structural Re-parameterization
Tianxiang Hao (Tsinghua University), Guiguang Ding (Tsinghua University)
ClassificationRestorationObject DetectionSegmentationConvolutional Neural NetworkImage
🎯 What it does: Proposed a position-enhanced convolutional PBConv, which introduces position embeddings into the convolution kernel and uses parallel small convolution branches to reinforce important positions. Finally, structural reparameterization restores it to a single convolution during inference without additional overhead.
Exploiting Self-Refining Normal Graph Structures for Robust Defense against Unsupervised Adversarial Attacks
Bingdao Feng (Tianjin University), Zhen Wang (Northwestern Polytechnical University)
Anomaly DetectionRepresentation LearningAdversarial AttackGraph Neural NetworkContrastive LearningGraph
🎯 What it does: This paper proposes an unsupervised graph representation learning framework based on self-correction, which generates a rough graph through anomaly detection and then gradually refines the graph structure using information theory to measure the Graph Pollution Degree (GPD), thereby obtaining robust node representations.
Exploiting Text Semantics for Few and Zero Shot Node Classification on Text-attributed Graph
Yuxiang Wang (Wuhan University), Jiawei Jiang (Wuhan University)
ClassificationGraph Neural NetworkTransformerContrastive LearningTextGraph
🎯 What it does: Propose a text semantic-enhanced TSA framework for few-shot and zero-shot node classification on text-attribute graphs.
Exploring Efficient and Effective Sequence Learning for Visual Object Tracking
Dongdong Li (National University of Defense Technology), Rui Chen (National University of Defense Technology)
Object TrackingTransformerVideo
🎯 What it does: Proposes the FastSeqTrack framework, achieving efficient visual tracking by generating target bounding box coordinates in a single parallel step
Exploring Semantic Masked Autoencoder for Self-supervised Point Cloud Understanding
Yixin Zha (University of Science and Technology of China), Tianzhu Zhang (University of Science and Technology of China)
Representation LearningTransformerPrompt EngineeringAuto EncoderContrastive LearningPoint Cloud
🎯 What it does: Propose a self-supervised point cloud pre-training framework called Semantic Masked Autoencoder, combining prototype-driven component semantic modeling and semantic-enhanced masking strategies to enhance the semantic expressiveness of point cloud features;
Exploring the Frontiers of Animation Video Generation in the Sora Era: Method, Dataset and Benchmark
Yudong Jiang (Bilibili Inc), Huyang Sun (Bilibili Inc)
GenerationData SynthesisTransformerReinforcement LearningDiffusion modelVideoTextBenchmark
🎯 What it does: Propose the AniSora system, which realizes functions such as animation video generation, frame interpolation, and local guidance, and provides a large-scale dataset and evaluation benchmark.
Exploring the Over-smoothing Problem of Graph Neural Networks for Graph Classification: An Entropy-based Viewpoint
Feifei Qian (Beijing Normal University), Edwin Hancock (University of York)
ClassificationGraph Neural NetworkGraphStochastic Differential Equation
🎯 What it does: Investigated the over-smoothing problem in graph neural networks for graph classification tasks, and proposed an SDE method based on high-entropy node sampling and discretization.
Exploring the Trade-Offs: Quantization Methods, Task Difficulty, and Model Size in Large Language Models From Edge to Giant
Jemin Lee (Electronics and Telecommunications Research Institute), Yongin Kwon (Electronics and Telecommunications Research Institute)
OptimizationComputational EfficiencyTransformerLarge Language ModelTextBenchmark
🎯 What it does: This paper evaluates systematically quantized instruction-tuned LLMs with parameters ranging from 1B to 405B on 13 benchmark datasets;
Exploring Transferable Homogenous Groups for Compositional Zero-Shot Learning
Zhijie Rao (Hong Kong Polytechnic University), Mengzhu Wang (Hong Kong Polytechnic University)
ClassificationRepresentation LearningTransformerPrompt EngineeringMixture of ExpertsVision Language ModelContrastive LearningMultimodality
🎯 What it does: Proposed Homogeneous Group Representation Learning (HGRL), which balances transferability and discriminability by learning representations of states and objects within homogeneous groups (clusters of the same attribute category), and introduced multi-expert networks, distributed group prompts, and intra-group enhancement in visual, prompt, and pair branches;
ExpTalk: Diverse Emotional Expression via Adaptive Disentanglement and Refined Alignment for Speech-Driven 3D Facial Animation
Zhan Qu (Zhejiang University), Fei Wu (Zhejiang University)
GenerationData SynthesisTransformerDiffusion modelAuto EncoderContrastive LearningMultimodalityMeshAudio
🎯 What it does: The paper proposes a system named ExpTalk for generating 3D facial animations synchronized with emotion based on speech.
External Memory Matters: Generalizable Object-Action Memory for Retrieval-Augmented Long-Term Video Understanding
Jisheng Dang (Lanzhou University), Tat Seng Chua (National University of Singapore)
RecognitionRetrievalTransformerLarge Language ModelVision-Language-Action ModelContrastive LearningVideoTextRetrieval-Augmented Generation
🎯 What it does: Propose a retrieval-enhanced long video understanding framework REVU, which fills the knowledge gaps of visual models through external text-object-action memory.
EyeSeg: An Uncertainty-Aware Eye Segmentation Framework for AR/VR
Zhengyuan Peng (Shanghai Jiao Tong University), Lizhuang Ma (Shanghai Jiao Tong University)
SegmentationConvolutional Neural NetworkImage
🎯 What it does: Proposed an eye segmentation framework called EyeSeg for AR/VR environments, which can achieve accurate and efficient eye segmentation and gaze estimation under conditions such as motion blur, eyelid occlusion, and training-testing domain differences.
Facets in Argumentation: A Formal Approach to Argument Significance
Johannes K. Fichte (Linkoping University), Jonathan Persson (Linkoping University)
GraphBenchmark
🎯 What it does: This paper introduces the concept of 'facet (panel)' to measure the extent to which an argument appears across different extensions in an abstract argumentation framework, providing a fine-grained reasoning approach between decision-making and enumeration. It also classifies the computational complexity of various decision and counting problems related to facets, while implementing an ASP-based tool and conducting experiments on ICCMA competition instances, demonstrating that facet counting is more efficient than full extension enumeration.
FADE: Towards Fairness-aware Data Generation for Domain Generalization via Classifier-Guided Score-based Diffusion Models
Yujie Lin (Tianjin University), Chen Zhao (Baylor University)
GenerationData SynthesisDomain AdaptationMeta LearningDiffusion modelScore-based ModelTabular
🎯 What it does: Built a three-stage FADE framework, leveraging pre-trained score-based diffusion models and classifier guidance to generate bias-removed, transferable fair data, which is then used to train downstream classifiers.
Fair Incomplete Multi-View Clustering via Distribution Alignment
Qianqian Wang (Xidian University), Xiangdong Zhang (Xidian University)
Domain AdaptationRepresentation LearningData-Centric LearningAuto EncoderContrastive LearningTabular
🎯 What it does: Proposed a fair missing multi-view clustering method called FIMVC-DA, which achieves unbiased clustering regarding sensitive attributes by leveraging distribution alignment and fairness constraints.
Fair Submodular Maximization over a Knapsack Constraint
Lijun Li (City University of Hong Kong), Ruilong Zhang (Technical University of Munich)
Optimization
🎯 What it does: Study the problem of maximizing fair submodular functions, proposing a new algorithm framework under knapsack constraints;
fairGNN-WOD: Fair Graph Learning Without Complete Demographics
Zichong Wang (Florida International University), Wenbin Zhang (Florida International University)
Representation LearningGraph Neural NetworkAuto EncoderGraphFinance Related
🎯 What it does: Proposes a framework called fairGNN-WOD to achieve fairness in graph learning when demographic information is missing.
FairSMOE: Mitigating Multi-Attribute Fairness Problem with Sparse Mixture-of-Experts
Changdi Yang (Northeastern University), Yanzhi Wang (Northeastern University)
ClassificationTransformerMixture of ExpertsImage
🎯 What it does: Propose a framework named FairSMoE, which models multi-attribute fairness issues as multi-task learning and achieves dynamic expert allocation through reusable sparse Mixture of Experts (SMoE) to simultaneously optimize multi-attribute fairness and predictive performance.
Fast and Stronger Lower Bounds for Planar Euclidean Shortest Paths
Stefan Funke (University of Stuttgart), Felix Weitbrecht (University of Stuttgart)
OptimizationComputational Efficiency
🎯 What it does: This paper proposes two key technologies: an algorithm that generates a strong lower bound by trimming convex corners of obstacle polygons, and a Hub Labeling acceleration method based on a visibility graph;
Fast Explanations via Policy Gradient-Optimized Explainer
Deng Pan (University of Notre Dame), Nitesh V. Chawla (University of Notre Dame)
Explainability and InterpretabilityReinforcement LearningImageText
🎯 What it does: Designed and implemented the Fast Explanation (FEX) framework, which leverages reinforcement learning and policy gradients to directly learn a Bernoulli distribution-based explainer, thereby rapidly generating high-quality feature attribution explanations for any black-box model.
Fast Guaranteed Tensor Recovery with Adaptive Tensor Nuclear Norm
Jiangjun Peng (Northwestern Polytechnical University), Shuang Xu (Northwestern Polytechnical University)
RestorationComputational EfficiencyImageVideo
🎯 What it does: Proposed an Adaptive Tensor Nuclear Norm (ATNN) framework based on adaptive column orthogonal transform, achieving fast and theoretically provable tensor recovery from sparse noise and missing values.
Fast Second-Order Online Kernel Learning Through Incremental Matrix Sketching and Decomposition
Dongxie Wen (Renmin University of China), Weinan Zhang (Shanghai Jiao Tong University)
ClassificationRecommendation SystemOptimizationTabular
🎯 What it does: Designed a second-order online kernel learning framework named FORKS, which employs incremental matrix sketching and incremental SVD to achieve dynamic feature mapping and fast model updates.
FAST: A Lightweight Mechanism Unleashing Arbitrary Client Participation in Federated Learning
Zhe Li (Rochester Institute of Technology), Haibo Yang (Rochester Institute of Technology)
Federated LearningImageText
🎯 What it does: Propose the FAST mechanism, which allows most communication rounds in federated learning to have arbitrary client participation, while only enforcing random global participation during intermittent snapshot rounds, thereby being compatible with arbitrary client participation (ACP) and maintaining convergence performance.
Fault Diagnosis in REDNet Model Space
Xiren Zhou (University of Science and Technology of China), Huanhuan Chen (University of Science and Technology of China)
Anomaly DetectionRecurrent Neural NetworkTransformerTime Series
🎯 What it does: Propose a model space fault diagnosis framework based on multi-path reservoir REDNet, fitting each time-series multivariate data into the REDNet model and performing classification in the model space.
FBQuant: FeedBack Quantization for Large Language Models
Yijiang Liu (Nanjing University), Li Du (Nanjing University)
Computational EfficiencyKnowledge DistillationTransformerLarge Language ModelText
🎯 What it does: Introduce a feedback sub-branch (FBQuant) for large language models based on weight-only quantization to improve accuracy and support deployment on edge devices.
FCKT: Fine-Grained Cross-Task Knowledge Transfer with Semantic Contrastive Learning for Targeted Sentiment Analysis
Wei Chen (Beihang University), Fuzhen Zhuang (Beihang University)
ClassificationTransformerContrastive LearningText
🎯 What it does: Propose the FCKT fine-grained cross-task knowledge transfer framework for the aspect-sentiment analysis task.
Featured Argumentation Framework: Semantics and Complexity
Gianvincenzo Alfano (University of Calabria), Irina Trubitsyna (University of Calabria)
🎯 What it does: Proposes the Featured Argumentation Framework (FAF) and its extension EFAF, enabling the addition of features to arguments within abstract argumentation frameworks and selecting sub-frameworks through First-Order Logic (FOL) constraints
FedAPA: Server-side Gradient-Based Adaptive Personalized Aggregation for Federated Learning on Heterogeneous Data
Yuxia Sun (Jinan University), Jingcai Guo (Hong Kong Polytechnic University)
ClassificationFederated LearningConvolutional Neural NetworkImage
🎯 What it does: Propose FedAPA under the federated learning framework, which generates personalized models through server-side gradient adaptive aggregation and reduces communication by partially sharing models.
FedBG: Proactively Mitigating Bias in Cross-Domain Graph Federated Learning Using Background Data
Sheng Huang (Sun Yat-sen University), Chuan Chen (Sun Yat-sen University)
Domain AdaptationFederated LearningGraph Neural NetworkGraph
🎯 What it does: Proposed the FedBG framework, which actively corrects cross-domain bias in federated graph learning by leveraging background graph data;
FedCCH: Automatic Personalized Graph Federated Learning for Inter-Client and Intra-Client Heterogeneity
Pengfei Jiao (Hangzhou Dianzi University), HuaMing Wu (Tianjin University)
Federated LearningRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposes FedCCH, a graph federated learning framework that simultaneously considers intra-client and inter-client heterogeneity, combining a learnable personalization factor (PF) with hashing clustering to achieve automatic personalization and privacy protection for graph data.
FedCM: Client Clustering and Migration in Federated Learning via Gradient Path Similarity and Update Direction Deviation
Peng Wang (Sichuan University), Cheng Dai (Sichuan University)
Federated LearningImage
🎯 What it does: Propose the Fed-CM framework, which clusters clients using the similarity of gradient paths in personalized layers and dynamically migrates clients during training to address data drift.
FedCPD:Personalized Federated Learning with Prototype-Enhanced Representation and Memory Distillation
Kaili Jin (Fujian Normal University), Limei Lin (Fujian Normal University)
Federated LearningKnowledge DistillationRepresentation LearningContrastive LearningImage
🎯 What it does: Studies personalized federated learning, proposing the FedCPD framework that employs hierarchical feature distillation and prototype alignment to alleviate historical information forgetting and enhance model generalization.
FedDLAD: A Federated Learning Dual-Layer Anomaly Detection Framework for Enhancing Resilience Against Backdoor Attacks
Binbin Ding (Nanjing University of Aeronautics and Astronautics), Sheng-Jun Huang (Nanjing University of Aeronautics and Astronautics)
Anomaly DetectionFederated LearningConvolutional Neural NetworkImage
🎯 What it does: Propose the FedDLAD two-layer anomaly detection framework, combining COF, IQR, and Pardon modules to enhance robustness against backdoor attacks in federated learning.
Federated Deconfounding and Debiasing Learning for Out-of-Distribution Generalization
Zhuang Qi (Shandong University), Xiangxu Meng (Beijing Institute of Technology)
Federated LearningConvolutional Neural NetworkVision Language ModelContrastive LearningImage
🎯 What it does: To address attribute bias caused by data sources and background in federated learning, the FedDDL method is proposed, achieving disentanglement and debiasing by decomposing the model's reasoning path.
Federated Domain Generalization with Decision Insight Matrix
Tianchi Liao (Sun Yat-sen University), Zibin Zheng (Sun Yat-sen University)
Domain AdaptationFederated LearningExplainability and InterpretabilityComputational EfficiencyRepresentation LearningImage
🎯 What it does: Propose the FedDIM framework, achieving federated domain generalization through a decision insight matrix, balancing fine-grained invariance of features and classifiers.
Federated Multi-view Graph Clustering with Incomplete Attribute Imputation
Wei Feng (Northwest A F University), Bo Dong (Xi'an Jiaotong University)
Federated LearningGraph Neural NetworkAuto EncoderContrastive LearningGraph
🎯 What it does: Proposed a federated multi-view graph clustering method FMVC-IAI, which can perform clustering in distributed scenarios with missing views and without sharing features.
Federated Stochastic Bilevel Optimization with Fully First-Order Gradients
Yihan Zhang (Temple University), Hongchang Gao (Temple University)
OptimizationFederated LearningRepresentation LearningHyperparameter SearchTabular
🎯 What it does: Proposed a global first-order gradient federated stochastic bilevel optimization algorithm FedSVRBGD-FO, which achieves efficient updates using a single-scale constant learning rate and variance reduction, avoiding the computation of second-order Hessian and Jacobian matrices;
FedHAN: A Cache-Based Semi-Asynchronous Federated Learning Framework Defending Against Poisoning Attacks in Heterogeneous Clients
Xiaoding Wang (Fujian Normal University), Limei Lin (Fujian Normal University)
Federated LearningSafty and PrivacyImage
🎯 What it does: This paper proposes FedHAN, a cache-based semi-asynchronous federated learning framework designed to defend against model poisoning attacks in heterogeneous client environments.
FedSaaS: Class-Consistency Federated Semantic Segmentation via Global Prototype Supervision and Local Adversarial Harmonization
Xiaoyang Yu (Qilu University of Technology (Shandong Academy of Sciences)), Peng Cheng (Zhejiang University)
SegmentationAutonomous DrivingFederated LearningGenerative Adversarial NetworkContrastive LearningImage
🎯 What it does: Propose the FedSaaS framework to address the problem of inconsistent class representations in federated semantic segmentation, achieving the unification of local and global class representations through global prototype supervision and local adversarial harmonization.
Feint and Attack: Jailbreaking and Protecting LLMs via Attention Distribution Modeling
Rui Pu (Beijing University of Posts and Telecommunications), Zaisheng Ye (Fujian Cancer Hospital)
Safty and PrivacyAdversarial AttackTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: To address jailbreaking and defense for large language models (LLM), this paper proposes an Attack (ABA) and Defense (ABD) framework based on attention distribution, which reconstructs prompts using attention strength and entropy information to shift the model's focus;
Few-Shot Incremental Multi-modal Learning via Touch Guidance and Imaginary Vision Synthesis
Lina Wei (Hangzhou City University), Dapeng Chen (Nanjing University of Information Science and Technology)
Representation LearningTransformerContrastive LearningImageVideoMultimodality
🎯 What it does: This paper proposes a few-shot incremental multi-modal learning framework based on Tactile-Guided and Imaginary Visual Synthesis (TIFS), addressing the problems of catastrophic forgetting and modality imbalance in multi-modal incremental learning.