IJCAI 2025 Papers — Page 5
International Joint Conference on Artificial Intelligence · 1014 papers
Few-shot Novel Category Discovery
Chunming Li (Nanjing University of Science and Technology), Haofeng Zhang (Nanjing University of Science and Technology)
ClassificationRecognitionMeta LearningTransformerContrastive LearningImage
🎯 What it does: Designed the Few-Shot New Class Discovery (FSNCD) task, proposed two baseline methods (Semi-supervised Hierarchical Clustering SHC and Uncertainty-aware K-means Clustering UKC), and trained feature representations using supervised contrastive learning, achieving both recognition of known classes and discovery of unknown classes on the query set.
FGeo-HyperGNet: Geometric Problem Solving Integrating FormalGeo Symbolic System and Hypergraph Neural Network
Xiaokai Zhang (Shanghai University), Tuo Leng (Shanghai University)
Graph Neural NetworkTransformerImageTextMultimodalityChain-of-Thought
🎯 What it does: Built a neuro-symbolic system, FGeo-HyperGNet, to automatically generate readable and verifiable solution processes from geometric images and text.
Filling the Missings: Spatiotemporal Data Imputation by Conditional Diffusion
Wenying He (Hebei University of Technology), Yude Bai (Tiangong University)
RestorationData SynthesisConvolutional Neural NetworkGraph Neural NetworkDiffusion modelTime Series
🎯 What it does: Proposed a conditional diffusion model called CoFILL for high-quality spatiotemporal data imputation.
Find and Perceive: Tell Visual Change with Fine-Grained Comparison
Feixiao Lv (Chinese Academy of Sciences), Lijun Liu (Chinese Academy of Sciences)
RecognitionGenerationTransformerVision Language ModelContrastive LearningImageMultimodalityBenchmark
🎯 What it does: This paper proposes a new image difference description learning framework called Find and Perceive (F&P), which achieves precise capture of subtle changes between similar images and language generation through two steps: fine-grained feature learning and weakly supervised region discrimination.
Finding Possible Winners in Spatial Voting with Incomplete Information
Hadas Shachnai (Technion Israel Institute of Technology), Andreas Wiese (Technical University of Munich)
OptimizationComputational Efficiency
🎯 What it does: This paper studies, under the spatial voting model, how to determine whether a candidate can possibly win when voters' ideal points are only given as interval information, and presents various complexity results.
Fine-Grained and Efficient Self-Unlearning with Layered Iteration
Hongyi Lyu (Macquarie University), Lianyong Qi (China University of Petroleum (East China))
ClassificationSafty and PrivacyComputational EfficiencyKnowledge DistillationImage
🎯 What it does: Designed and proposed the Self-Unlearning with Layered Iteration (SULI) method, achieving efficient fine-grained forgetting in machine learning models through layered iteration and soft label selective probability adjustment.
Fine-grained Prompt Screening: Defending Against Backdoor Attack on Text-to-Image Diffusion Models
Yiran Xu (Fudan University), Xinpeng Zhang (Fudan University)
Safty and PrivacyAdversarial AttackTransformerLarge Language ModelPrompt EngineeringDiffusion modelImageText
🎯 What it does: Propose the Fine-grained Prompt Screening (GrainPS) method, which detects and locates backdoor triggers in text-to-image diffusion models during the inference phase by splitting prompts at a fine-grained level and calculating semantic alignment scores.
Finite-Time Analysis of Heterogeneous Federated Temporal Difference Learning
Ye Zhu (Auburn University), Shiwen Mao (Auburn University)
Federated LearningReinforcement LearningOrdinary Differential Equation
🎯 What it does: Propose a federated temporal difference learning algorithm named HFTD tailored for environmental and computational heterogeneity, aimed at collaboratively evaluating the value function of a given policy;
First-Order Coalition Logic
Davide Catta (Université Sorbonne Paris Nord), Aniello Murano (University of Naples Federico II)
🎯 What it does: This paper proposes a new logic called First-Order Collaborative Logic (FOCL), which integrates the core ideas of Collaborative Logic (CL) and Strategy Logic (SL), allowing arbitrary quantification over agent actions and explicitly referencing action labels;
FissionVAE: Federated Non-IID Image Generation with Latent Space and Decoder Decomposition
Chen Hu (Swansea University), Xiaoke Ma (Xi'Dian University)
GenerationFederated LearningAuto EncoderImage
🎯 What it does: Propose the FissionVAE model, achieving higher quality image synthesis in federated non-IID image generation by decoupling the latent space and constructing dedicated decoder branches for each client group.
Flexible Generalized Low-Rank Regularizer for Tensor RPCA
Zhiyang Gong (Huazhong Agricultural University), Yulong Wang (Huazhong Agricultural University)
RestorationOptimizationImageVideo
🎯 What it does: This paper proposes a flexible and generalizable low-rank regularization framework called FGTNN, based on which two TRPCA methods, FGTRPCA and SFGTRPCA, are constructed. Experimental results verify that their denoising and recovery performance on color images, gray-scale videos, hyperspectral images, and multispectral images outperforms existing methods.
Flow Matching Based Sequential Recommender Model
Feng Liu (Wuhan University), Chenliang Li (Wuhan University)
Recommendation SystemTransformerFlow-based ModelSequentialOrdinary Differential Equation
🎯 What it does: This paper proposes a sequence recommendation model called FMREC based on flow matching, achieving precise prediction of users' next interaction by simplifying the forward noise path and backward denoising process of diffusion models.
Flow-based Time-aware Causal Structure Learning for Sequential Recommendation
Hangtong Xu (Jilin University), En Wang (Jilin University)
Recommendation SystemTransformerFlow-based ModelSequential
🎯 What it does: Developed a new recommendation framework FCSRec that explicitly models unobserved confounding factors and their time-varying effects, jointly learning them with sequential dependencies.
ForgDiffuser: General Image Forgery Localization with Diffusion Models
Mengxi Wang (Beijing University of Posts and Telecommunications), Jiwei Zhang (Beijing University of Posts and Telecommunications)
SegmentationConvolutional Neural NetworkTransformerDiffusion modelImage
🎯 What it does: Propose an image forgery localization framework called ForgDiffuser based on conditional diffusion models, which generates masks using forged images as conditions, and combines attention guidance and edge-driven modules to improve localization accuracy.
Formal Synthesis of Safe Kolmogorov-Arnold Network Controllers with Barrier Certificates
Xiongqi Zhang (Zhejiang Sci-Tech University), Zuohua Ding (Zhejiang Sci-Tech University)
BenchmarkPhysics Related
🎯 What it does: Propose a formal synthesis method for control boundary certificates and safe controllers based on Kolmogorov–Arnold networks (KAN).
Free Lunch of Image-mask Alignment for Anomaly Image Generation and Segmentation
Xiangyue Li (Soochow University), Mingjie Sun (Soochow University)
SegmentationGenerationAnomaly DetectionDiffusion modelImageBiomedical Data
🎯 What it does: This paper proposes a dual-branch training strategy, enabling the generative model to simultaneously generate anomalous images and their masks, while improving the correspondence between images and masks through alignment regularization. Subsequently, the pre-trained generative model provides high-quality data and generates feedback loss to further enhance segmentation performance.
FreEformer: Frequency Enhanced Transformer for Multivariate Time Series Forecasting
Wenzhen Yue (Peking University), Ji Shi (Peking University)
TransformerTime SeriesFinance Related
🎯 What it does: Propose a multivariate time series forecasting model called FreEformer based on a frequency-domain Transformer.
FreqLLM: Frequency-Aware Large Language Models for Time Series Forecasting
Shunnan Wang (Chongqing University), Guansong Pang (Singapore Management University)
TransformerLarge Language ModelPrompt EngineeringTime Series
🎯 What it does: Propose the FreqLLM framework, which embeds dual-scale frequency domain signals into LLMs and aligns them with the semantic space through soft prompts to enhance time series prediction performance.
FreqMoE: Dynamic Frequency Enhancement for Neural PDE Solvers
Tianyu Chen (Beihang University), Jianxin Li (Beihang University)
Mixture of ExpertsTabularBenchmarkPhysics Related
🎯 What it does: Propose the FreqMoE framework, leveraging sparse Mixture-of-Experts to dynamically enhance the capability of Fourier Neural Operators in the high-frequency domain, achieving high-resolution PDE solving;
Frequency-Aware Deep Depth from Focus
Tao Yan (Shanxi University), Feijiang Li (Shanxi University)
Depth EstimationConvolutional Neural NetworkImage
🎯 What it does: Proposed a frequency-aware deep focusing (FAD) network that achieves precise depth prediction for multi-focus image sequences by utilizing a time-frequency joint module combining multi-scale spatial features and frequency domain global features.
From End-to-end to Step-by-step: Learning to Abstract via Abductive Reinforcement Learning
Zilong Wang (Nanjing University), Wang-Zhou Dai (Nanjing University)
Representation LearningReinforcement LearningImage
🎯 What it does: Proposed the Abductive Abstract Reinforcement Learning (A2RL) framework, which integrates neural networks with symbolic reasoning to directly learn abstract high-level steps from raw perceptual inputs and represent them in the form of an abstract state machine (ASM);
From General Relation Patterns to Task-Specific Decision-Making in Continual Multi-Agent Coordination
Chang Yao (Beijing Jiaotong University), Kai Lv (Beijing Jiaotong University)
Recurrent Neural NetworkReinforcement LearningSequentialBenchmark
🎯 What it does: Propose a relation-based persistent multi-agent coordination method called RPG, which can extract generic relational patterns from continuously emerging new tasks and map them to specific action spaces;
From Individual to Universal: Regularized Multi-view Joint Representation for Multi-view Subspace-Preserving Recovery
Libin Wang (Huazhong Agricultural University), Yuan Yan Tang (University of Macau)
Representation Learning
🎯 What it does: Proposed and theoretically analyzed a regularized multi-view joint sparse representation (RMJSR) model for multi-view subspace preservation recovery, and applied it to multi-view subspace classification (MSCla) and clustering (MSClu)
From Sparse to Complete: Semantic Understanding Based on Stroke Evolution in On-the-fly Sketch-based Image Retrieval
Yingge Liu (Chongqing University of Posts and Telecommunications), Guoyin Wang (Chongqing University of Posts and Telecommunications)
RetrievalMixture of ExpertsContrastive LearningImage
🎯 What it does: This paper proposes a framework based on stroke consistency detection and adaptive gating Mixture of Experts for noise stroke filtering and feature extraction in instant sketch retrieval.
FS-KEN: Few-shot Knowledge Graph Reasoning by Adversarial Negative Enhancing
Lingyuan Meng (National University of Defense Technology), Wenpeng Lu (Shandong Computer Science Center(National Supercomputer Center in Jinan))
Representation LearningMeta LearningReinforcement LearningGenerative Adversarial NetworkGraphBenchmark
🎯 What it does: Proposed FS-KEN, a few-shot knowledge graph reasoning framework based on adversarial learning, which enhances few-shot learning by generating high-quality negative samples using GAN.
Fully Test-Time Adaptation for Feature Decrement in Tabular Data
Zi-Jian Cheng (Nanjing University), Lan-Zhe Guo (Nanjing University)
Domain AdaptationTransformerLarge Language ModelTabularChain-of-Thought
🎯 What it does: This paper studies the fully test-time adaptation (FTTA) problem under feature decrement scenarios, proposing a missing feature imputation method based on large language models, LLM-IMPUTE, and an incremental training method called ATLLM that introduces simulated feature missingness during training.
Fusion of Granular-Ball Visual Spatial Representations for Enhanced Facial Expression Recognition
Shuaiyu Liu (University of Electronic Science and Technology of China), Guoyin Wang (Chongqing University of Posts and Telecommunications)
RecognitionRepresentation LearningGraph Neural NetworkTransformerImage
🎯 What it does: Propose a CS-GBSBF method that converts facial expression images into a graph structure via granular spherical representation, separately extracting visual and spatial features. A component separation network is used to obtain visual/spatial representations of key facial regions, and in the fusion network, spatial representations adaptively guide the fusion of visual features, ultimately completing emotion recognition.
G3PT: Unleash the Power of Autoregressive Modeling in 3D Generation via Cross-Scale Querying Transformer
Jinzhi Zhang (AMAP), Mu Xu (AMAP)
GenerationData SynthesisTransformerContrastive LearningImageTextMultimodalityPoint Cloud
🎯 What it does: Propose a scalable cross-scale autoregressive model, G3PT, which directly maps unordered 3D point clouds to multi-scale discrete tokens, and generates them through cross-scale query Transformer (CQT) and cross-scale autoregressive (CAR) mechanisms.
GarmentDiffusion: 3D Garment Sewing Pattern Generation with Multimodal Diffusion Transformers
Xinyu Li (Zhejiang University), Yuanda Wang (Shenfu Research)
GenerationTransformerVision Language ModelDiffusion modelImageTextMultimodalityMesh
🎯 What it does: Designed and implemented a multi-modal diffusion model called GarmentDiffusion, which generates centimeter-level precision vectorized 3D sewing patterns from text, images, or incomplete pattern inputs, and supports completion of full or partial patterns.
GATES: Cost-aware Dynamic Workflow Scheduling via Graph Attention Networks and Evolution Strategy
Ya Shen (Victoria University of Wellington), Mengjie Zhang (Victoria University of Wellington)
OptimizationGraph Neural NetworkTransformerReinforcement LearningGraph
🎯 What it does: A deep reinforcement learning scheduling framework named GATES, which integrates graph attention networks (GAT) with evolutionary strategies (ES), is proposed to address the cost-aware dynamic workflow scheduling (CADWS) problem in cloud computing environments.
Gaussian Mixture Model for Graph Domain Adaptation
Mengzhu Wang (Hebei University of Technology), Nan Yin (Hong Kong University of Science and Technology)
Domain AdaptationGraph Neural NetworkGraph
🎯 What it does: Proposes a graph domain adaptation framework integrating Gaussian Mixture Model (GMM), achieving more precise alignment and clustering between source and target domains by modeling multimodal distributions within graph structures.
GBGC: Efficient and Adaptive Graph Coarsening via Granular-ball Computing
Shuyin Xia (Chongqing University of Posts and Telecommunications), Guoyin Wang (Chongqing Normal University)
ClassificationComputational EfficiencyRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: Propose a multi-granularity adaptive graph coarsening method (GBGC) based on granular ball computing, which constructs super nodes by hierarchically generating and refining granular balls at the global-local level, significantly compressing the graph scale while preserving key structural information.
GCNT: Graph-Based Transformer Policies for Morphology-Agnostic Reinforcement Learning
Yingbo Luo (Jilin University), Xueming Xiao (Changchun University of Science and Technology)
Robotic IntelligenceGraph Neural NetworkTransformerReinforcement LearningGraphBenchmark
🎯 What it does: Propose GCNT, a structured network that integrates GCN and Transformer to train a universal control policy compatible with various robot morphologies;
GCTAM: Global and Contextual Truncated Affinity Combined Maximization Model For Unsupervised Graph Anomaly Detection
Xiong Zhang (Yunnan University), Hua Jiang (Yunnan University)
Anomaly DetectionGraph Neural NetworkGraph
🎯 What it does: Proposed the GCTAM model for unsupervised graph anomaly detection, combining context and global truncation affinity maximization;
General Incomplete Time Series Analysis via Patch Dropping Without Imputation
Yangyang Wu (Zhejiang University), Meng Xi (Zhejiang University)
ClassificationAnomaly DetectionTransformerLarge Language ModelSupervised Fine-TuningTime Series
🎯 What it does: Propose an end-to-end framework called INTER that directly analyzes incomplete multivariate time series, completely bypassing traditional missing value imputation steps.
Generalized Safe Conditional Syntax Splitting of Belief Bases
Lars-Phillip Spiegel (FernUniversitat in Hagen), Christoph Beierle (FernUniversitat in Hagen)
🎯 What it does: This paper studies how to generalize safe conditional syntax splitting into a broader form that allows sub-bases to share non-self-consistent conditions, and proposes the concept of genuine splitting, followed by new postconditions for inductive reasoning operators.
Generate or Re-Weight? A Mutual-Guidance Method for Class-Imbalanced Graphs
Zhongying Zhao (Shandong University of Science and Technology), Qingtian Zeng (Shandong University of Science and Technology)
ClassificationGraph Neural NetworkGraph
🎯 What it does: Propose a GraphMuGu method that integrates generation and reweighting to alleviate class imbalance issues in graph data.
Generic Adversarial Attack Framework Against Vertical Federated Learning
Yimin Liu (Beijing Institute of Technology), Peng Jiang (Beijing Institute of Technology)
Federated LearningAdversarial AttackImage
🎯 What it does: Designed PGAC, an attack framework for vertical federated learning (VFL), which can generate adversarial samples capable of dominating joint inference without querying the server model, without real opponent test samples, and using only non-training domain labeled auxiliary data.
GLDiTalker: Speech-Driven 3D Facial Animation with Graph Latent Diffusion Transformer
Yihong Lin (South China University of Technology), Huang Xu (Huawei Cloud)
GenerationGraph Neural NetworkTransformerDiffusion modelAuto EncoderMeshAudio
🎯 What it does: Proposed GLDiTalker, a two-stage speech-driven 3D facial animation model based on graph-quantized diffusion Transformer, addressing the audio-mesh mismatch problem and improving lip synchronization and action diversity.
Global Information Compensation Network for Image Denoising
Shifei Ding (China University of Mining and Technology), Lili Guo (China University of Mining and Technology)
RestorationConvolutional Neural NetworkImage
🎯 What it does: Propose a Global Information Compensation Network (GICN) for image denoising, capable of simultaneously capturing global features in both spatial and frequency domains;
Going Beyond Consistency: Target-oriented Multi-view Graph Neural Network
Sujia Huang (Nanjing University of Science and Technology), Tong Zhang (Nanjing University of Science and Technology)
ClassificationRepresentation LearningGraph Neural NetworkMultimodalityGraph
🎯 What it does: Proposed a Target-oriented Graph Neural Network (TGNN), which enhances the representation capability of multi-view graph data by separating the determinative and incidental features of each view and using a class-level dual-objective loss to achieve task-oriented semantic learning.
Good Advisor for Source Localization: Using Large Language Model to Guide the Source Inference Process
Dongpeng Hou (Northwestern Polytechnical University), Zhen Wang (Northwestern Polytechnical University)
Anomaly DetectionGraph Neural NetworkTransformerLarge Language ModelContrastive LearningText
🎯 What it does: Proposed the CRSLL framework, which uses LLM as an 'advisor' to generate comment source analysis, combining contrastive learning, differentiable feature masking, and cross-modal attention to achieve the localization of rumor sources
GPI-Net: Gestalt-Guided Parallel Interaction Network via Orthogonal Geometric Consistency for Robust Point Cloud Registration
Weikang Gu (Fujian Agriculture and Forestry University), Lifang Wei (Fujian Agriculture and Forestry University)
Pose EstimationTransformerPoint Cloud
🎯 What it does: This paper proposes a point cloud registration network called GPI-Net, which effectively removes outliers and accurately estimates rigid transformations in the initial correspondence point set.
GPL4SRec: Graph Multi-Level Aware Prompt Learning for Streaming Recommendation
Hao Cang (Soochow University), Pengpeng Zhao (Soochow University)
Recommendation SystemGraph Neural NetworkPrompt EngineeringContrastive LearningGraph
🎯 What it does: Proposed a stream recommendation framework named GPL4SRec based on graph pre-training and multi-level prompt learning, which simultaneously captures user long-term and short-term preferences, and models incremental updates and cascading changes of graphs through three levels of prompts (node-aware, structure-aware, and hierarchical-aware).
Gradient-based Causal Feature Selection
Zhaolong Ling (Anhui University), Zhangling Duan (Institute of Artificial Intelligence Hefei Comprehensive National Science Center)
OptimizationAuto EncoderGraphBiomedical Data
🎯 What it does: GCFS employs gradient optimization combined with autoencoders, acyclic constraints, and mask methods to select the Markov Blanket (causal features) for the target variable.
GRAML: Goal Recognition As Metric Learning
Matan Shamir (Bar-Ilan University), Reuth Mirsky (Tufts University)
RecognitionRecurrent Neural NetworkReinforcement LearningSequential
🎯 What it does: Proposes the GRAML (Goal Recognition As Metric Learning) framework, which uses deep metric learning (Siamese LSTM) to distinguish trajectories of different goals in the embedding space, achieving online dynamic goal recognition (ODGR) and supporting offline self-supervised training, one-time adaptation to new goals, and unified processing of continuous and discrete environments.
Granular-Ball-Induced Multiple Kernel K-Means
Shuyin Xia (Chongqing University of Posts and Telecommunications), Guoyin Wang (Chongqing Normal University)
OptimizationComputational EfficiencyImageTabular
🎯 What it does: This paper proposes the GB-MKKM framework, a multi-kernel K-means approach based on granular spheres, to improve clustering efficiency and robustness.
GRAPE: Heterogeneous Graph Representation Learning for Genetic Perturbation with Coding and Non-Coding Biotype
Changxi Chi (Zhejiang University), Stan Z. Li (Westlake University)
Representation LearningGraph Neural NetworkLarge Language ModelContrastive LearningMultimodalityGraphBiomedical Data
🎯 What it does: This paper proposes the GRAPE model, which initializes gene representations using multimodal features from gene descriptions and DNA sequences, constructs a heterogeneous graph to learn gene regulatory networks, and achieves accurate predictions in single-cell gene perturbation tasks.
Graph Embedded Contrastive Learning for Multi-View Clustering
Hongqing He (Guangxi Normal University), Xiaofeng Zhu (University of Electronic Science and Technology of China)
Representation LearningGraph Neural NetworkAuto EncoderContrastive LearningGraph
🎯 What it does: Proposes the GMVC framework, unifying multi-view clustering (MVC) and multi-graph clustering (MVGC) under a single model, and achieving representation learning and clustering through contrastive learning on graph embeddings;
Graph OOD Detection via Plug-and-Play Energy-based Evaluation and Propagation
Yunxia Zhang (Jilin University), Ying Wang (Jilin University)
Anomaly DetectionRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: This paper proposes an energy propagation-based graph neural network, EPGNN, for detecting nodes with unknown distributions in graphs.
Graph Prompts: Adapting Video Graph for Video Question Answering
Yiming Li (Nanjing University of Posts and Telecommunications), Changsheng Xu (Institute of Automation Chinese Academy of Sciences)
Graph Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringVideoGraph
🎯 What it does: Proposed the GP-VQA model, which adopts a two-stage training approach for video graphs and question graphs: first performing task pre-training using randomly masked graphs, then conducting prompt tuning for VideoQA through graph prompts and cross-modal message passing.
Graph Random Walk with Feature-Label Space Alignment: A Multi-Label Feature Selection Method
Wanfu Gao (Jilin University), Kunpeng Liu (Portland State University)
ClassificationGraph Neural NetworkTabular
🎯 What it does: This paper proposes a multi-label feature selection method based on random walks on feature-label composite graphs and feature-label space alignment.
GraphAD: Interaction Scene Graph for End-to-end Autonomous Driving
Yunpeng Zhang (PhiGent Robotics), Dalong Du (PhiGent Robotics)
Autonomous DrivingGraph Neural NetworkTransformerGaussian SplattingMultimodality
🎯 What it does: Proposed the GraphAD end-to-end autonomous driving architecture, which models the geometric priors of dynamic vehicles and static map elements through an Interactive Scene Graph (ISG), achieving efficient interaction aggregation and trajectory prediction;
GraphProt: Certified Black-Box Shielding Against Backdoored Graph Models
Xiao Yang (Shanghai Jiao Tong University), Hang Zhang (Cornell University)
Adversarial AttackGraph Neural NetworkGraph
🎯 What it does: Proposes a black-box backdoor defense method called GRAPHPROT, which suppresses backdoor attacks in graph neural networks by leveraging graph anomaly filtering, subgraph sampling, and majority voting integration.
Grounding Methods for Neural-Symbolic AI
Rodrigo Castellano Ontiveros (University of Siena), Michelangelo Diligenti (University of Siena)
Computational EfficiencyRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: Proposed a parameterizable logical reasoning forward-chaining grounding method that controls reasoning depth and width to balance scalability and expressiveness in NeSy models.
GSDet: Gaussian Splatting for Oriented Object Detection
Zeyu Ding (China University of Mining and Technology), Rui Yao (China University of Mining and Technology)
Object DetectionTransformerGaussian SplattingImage
🎯 What it does: Propose the GSDet framework, which performs oriented object detection using 3D Gaussian distributions in the 3D feature space;
GSDNet: Revisiting Incomplete Multimodality-Diffusion Emotion Recognition from the Perspective of Graph Spectrum
Yuntao Shou (Anhui Normal University), Keqin Li (State University of New York, New Paltz)
RecognitionGraph Neural NetworkDiffusion modelScore-based ModelMultimodalityGraphStochastic Differential Equation
🎯 What it does: Propose a graph spectral diffusion network (GSDNet), which precisely completes missing modalities in dialogue emotion recognition tasks by injecting Gaussian noise into the graph spectral domain and using fractional estimation to recover missing modalities;
Guiding Large Language Models in Modeling Optimization Problems via Question Partitioning
Xiaotian Pan (University of Science and Technology of China), Xiang-Yang Li (University of Science and Technology of China)
OptimizationTransformerLarge Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: Proposed the PaMOP framework, which leverages large language models (LLMs) to automate modeling of optimization problems, and achieves large-scale problem modeling through techniques such as tree structure decomposition, incremental prompting, iterative error correction, and reverse translation.
Guiding LLM-based Smart Contract Generation with Finite State Machine
Hao Luo (Wuhan University), Jiawei Jiang (Wuhan University)
GenerationAI Code AssistantTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextGraph
🎯 What it does: Propose an LLM-assisted intelligent contract auto-generation framework FSM-SCG based on finite state machines (FSM). First, user requirements are converted into SmartFSM, then the LLM generates the contract, and continuously improves it through a feedback loop of compilation and security checks.
HA-SCN: Learning Hierarchical Aligned Subtree Convolutional Networks for Graph Classification
Xinya Qin (Beijing Normal University), Edwin Hancock (University of York)
ClassificationConvolutional Neural NetworkGraphBiomedical Data
🎯 What it does: Proposes the Hierarchical Aligned Subtree Convolutional Network (HA-SCN) for graph classification;
Handling Infinite Domain Parameters in Planning Through Best-First Search with Delayed Partial Expansions
Ángel Aso-Mollar (Valencian Research Institute for Artificial Intelligence, Universitat Politècnica de València), Eva Onaindia (Valencian Research Institute for Artificial Intelligence, Universitat Politècnica de València)
OptimizationTabularBenchmark
🎯 What it does: This paper proposes a method for automated planning problems involving infinite-domain control parameters, treating control parameters as decision points rather than constraints. It implements systematic search using a Sampling Best-First Search (SBFS) algorithm based on best-first search with delayed partial expansion, and proves its probabilistic completeness under certain conditions;
HeTa: Relation-wise Heterogeneous Graph Foundation Attack Model
Yuling Wang (Hangzhou Dianzi University), Xiao Wang (Beihang University)
Adversarial AttackGraph Neural NetworkSupervised Fine-TuningGraph
🎯 What it does: Designed and implemented a fundamental attack model HeTa based on relational units, which can migrate across different heterogeneous graph neural networks (HGNN) and quickly adapt to new graphs, achieving low-budget node injection attacks.
Heterogeneous Federated Learning with Scalable Server Mixture-of-Experts
Jingang Jiang (South China Normal University), Chenyou Fan (South China Normal University)
Federated LearningMixture of ExpertsImageText
🎯 What it does: Propose Fed-MoE: a heterogeneous federated learning framework that aggregates lightweight client models into a large Mixture-of-Experts (MoE) on the server side to enhance model performance.
Heterogeneous Temporal Hypergraph Neural Network
Huan Liu (Hangzhou Dianzi University), Di Jin (Tianjin University)
Representation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Propose a time-varying hypergraph representation learning framework named HTHGN that can simultaneously capture heterogeneous nodes and higher-order interactions
Heterophily-Aware Personalized PageRank for Node Classification
Giuseppe Pirrò (University of Calabria)
ClassificationGraph Neural NetworkSupervised Fine-TuningGraphBenchmark
🎯 What it does: Node classification on heterogeneous graphs, proposing the HAPPY framework: combining Heterogeneity-Aware Personalized PageRank (H-PPR) with adaptive subgraph extraction to capture both homogeneous and heterogeneous neighbors.
HGEN: Heterogeneous Graph Ensemble Networks
Jiajun Shen (Florida Atlantic University), Xingquan Zhu (Florida Atlantic University)
ClassificationGraph Neural NetworkMixture of ExpertsGraph
🎯 What it does: Propose HGEN, an ensemble learning framework for heterogeneous graphs, which generates diverse homogeneous subgraphs through meta-paths and trains multiple base learners (GNNs). Subsequently, residual attention fusion and correlation regularization are applied for integration, significantly improving node classification accuracy.
HGMP: Heterogeneous Graph Multi-Task Prompt Learning
Pengfei Jiao (Hangzhou Dianzi University), Yanxian Bi (CETC Academy of Electronics and Information Technology Group China Academy of Electronic and Information Technology)
Representation LearningGraph Neural NetworkPrompt EngineeringContrastive LearningGraph
🎯 What it does: Proposed the HGMP framework, unifying heterogeneous graph tasks into a graph-level task and combining contrastive pre-training with heterogeneous graph prompting.
Hierarchy Knowledge Graph for Parameter-Efficient Entity Embedding
Hepeng Gao (Jilin University), Ying Wang (Jilin University)
Computational EfficiencyRepresentation LearningGraph Neural NetworkGraphBenchmark
🎯 What it does: Proposes a hierarchical parameter-efficient entity embedding model HRL that jointly learns common and specific representations of entities using a meta-encoder and context encoder, enabling knowledge graph completion without storing separate embeddings for each entity.
High-Confident Local Structure Guided Consensus Graph Learning For Incomplete Multi-view Clustering
Shuping Zhao (Guangdong University of Technology), Tingting Chai (Harbin Institute of Technology)
Representation LearningGraph Neural NetworkMultimodality
🎯 What it does: This paper proposes a high-confidence local structure-guided consensus graph learning method, HLSCG IMC, to address the problems of missing views and information imbalance in incomplete multi-view clustering.
High-Fidelity Road Network Generation with Latent Diffusion Models
Jinming Wang (University of Exeter), Man Luo (University of Exeter)
Data SynthesisTransformerDiffusion modelAuto EncoderGraphSequential
🎯 What it does: Propose the GraphWalker framework, which can directly generate high-fidelity road network maps from noisy trajectories, achieving end-to-end road network reconstruction.
Higher-order Logical Knowledge Representation Learning
Suixue Wang (Hainan University), Qingchen Zhang (Hainan University)
Representation LearningGraph Neural NetworkGraph
🎯 What it does: Propose a knowledge graph representation learning framework called LORE, which utilizes network motifs to intuitively capture high-order logical relationships and jointly aggregates attribute features with high-order relationship features, supporting two tasks: entity classification and chain prediction.
HIPP: Protecting Image Privacy via High-Quality Reversible Protected Version
Xi Ye (Wuhan University), Geying Yang (Wuhan University)
Safty and PrivacyFlow-based ModelGenerative Adversarial NetworkImage
🎯 What it does: Propose a reversible thumbnail-based privacy protection scheme HIPP, which utilizes latent space to separate detail and contour information and generates natural protected images
HiTuner: Hierarchical Semantic Fusion Model Fine-Tuning on Text-Attributed Graphs
Zihan Fang (Fuzhou University), Shiping Wang (Fuzhou University)
ClassificationRepresentation LearningGraph Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningTextGraph
🎯 What it does: Propose the HiTuner framework, which enhances the node classification performance of text attribute graphs (TAG) by combining the multi-layer hidden states of LLM with fine-tuned PLM and utilizing a confidence network to adaptively fuse semantic information from different layers.
HLMTrans: A Sim-to-Real Transfer Framework for Spatial Crowdsourcing with Human-Guided Language Models
Qingshun Wu (Zhengzhou University), Mingliang Xu (Zhengzhou University)
Domain AdaptationReinforcement Learning from Human FeedbackGraph Neural NetworkTransformerLarge Language ModelReinforcement LearningChain-of-Thought
🎯 What it does: Propose the HLMTrans framework, combining RL task allocation with human-guided LLM for Sim-to-Real transfer.
How to Mitigate Information Loss in Knowledge Graphs for GraphRAG: Leveraging Triple Context Restoration and Query-Driven Feedback
Manzong Huang (Hefei University of Technology), Xindong Wu (Hefei University of Technology)
RetrievalRepresentation LearningTransformerLarge Language ModelTextGraphRetrieval-Augmented Generation
🎯 What it does: Designed and implemented the Triple Context Restoration and Query-Driven Feedback (TCR-QF) framework, which first restores the original text context of triplets to compensate for information loss, and then dynamically completes missing knowledge graphs through a query-driven feedback loop during inference.
How to Resolve Envy by Adding Goods
Matthias Bentert (University of Bergen), Leon Kellerhals (TU Clausthal)
OptimizationComputational Efficiency
🎯 What it does: The study examines eliminating envy between agents by adding extra items while keeping the initial allocation unchanged, presenting algorithms and complexity results for various scenarios.
HPDM: A Hierarchical Popularity-aware Debiased Modeling Approach for Personalized News Recommender
Xiangfu He (Tianjin University), Hongtao Liu (Du Xiaoman Financial Technology)
Recommendation SystemTransformerText
🎯 What it does: Propose a hierarchical news recommendation model HPDM that considers news popularity, aiming to correct popularity bias in user click data.
HSRMamba: Contextual Spatial-Spectral State Space Model for Single Hyperspectral Image Super-Resolution
Shi Chen (Wuhan University), Liangpei Zhang (Henan Academy of Sciences)
Super ResolutionImage
🎯 What it does: Propose a single-image super-resolution framework for hyperspectral images, HSRMamba, based on a state-space model.
Human Motion Capture from Loose and Sparse Inertial Sensors with Garment-aware Diffusion Models
Andela Ilic (ETH Zürich), Christian Holz (ETH Zürich)
Data SynthesisPose EstimationTransformerDiffusion modelTime Series
🎯 What it does: A new task is proposed, namely full-body human pose estimation using sparse and loosely attached inertial sensors. A transformer-based diffusion model was developed to synthesize loose IMU data and estimate human poses by simulating IMU recordings from existing clothing-aware human motion datasets.
Human-Imperceptible, Machine-Recognizable Images
Fusheng Hao (Shenzhen Institutes of Advanced Technology Chinese Academy of Sciences), Jun Cheng (Shenzhen Institutes of Advanced Technology Chinese Academy of Sciences)
ClassificationObject DetectionSafty and PrivacyRepresentation LearningTransformerImage
🎯 What it does: Propose an image encryption and learnable privacy protection framework, utilizing two strategies: random shuffling (RS) and subblock mixing (MI), to encrypt images into human-unidentifiable but machine-recognizable forms, with minimal modifications to ViT and YOLOS to support classification and detection on encrypted images.
Human-Readable Neuro-Fuzzy Networks from Frequent Yet Discernible Patterns in Reward-Based Environments
John Wesley Hostetter (North Carolina State University), Min Chi (North Carolina State University)
Explainability and InterpretabilityReinforcement LearningAuto EncoderTabular
🎯 What it does: This paper proposes a self-organizing and simplified neural fuzzy network (NFN) method, leveraging fuzzy information granulation and graph theory techniques to retain only frequent but discriminable patterns, thereby generating interpretable strategies.
Hybrid Local Causal Discovery
Zhaolong Ling (Anhui University), Kui Yu (Hefei University of Technology)
TabularBenchmark
🎯 What it does: Proposed a hybrid local causal discovery algorithm named HLCD, which constructs a rough skeleton using constraint-based methods (with OR rules), then refines the skeleton via a scoring function and distinguishes V-structures from equivalent classes through score comparisons, thereby fully determining the direct causes and effects of the target variable.
Hybrid Mesh-Gaussian Representation for Efficient Indoor Scene Reconstruction
Binxiao Huang (University of Hong Kong), Ngai Wong (University of Hong Kong)
GenerationComputational EfficiencyGaussian SplattingMesh
🎯 What it does: This paper proposes a hybrid mesh-Gaussian representation for high-quality reconstruction of indoor scenes. The method uses textured meshes to cover texture-rich planar regions while retaining 3D Gaussian splats to model complex geometry, significantly reducing the number of Gaussians and improving rendering speed.
Hybrid Relational Graphs with Sentiment-laden Semantic Alignment for Multimodal Emotion Recognition in Conversation
Hongru Ji (Northwestern Polytechnical University), Chao Gao (Northwestern Polytechnical University)
ClassificationRecognitionGraph Neural NetworkTransformerContrastive LearningImageTextMultimodalityAudio
🎯 What it does: This paper proposes a multimodal emotion recognition framework based on hybrid relational graphs and emotional semantic alignment (HRG-SSA).
HyperDet: Source Detection in Hypergraphs via Interactive Relationship Construction and Feature-rich Attention Fusion
Le Cheng (Northwestern Polytechnical University), Zhen Wang (Northwestern Polytechnical University)
Anomaly DetectionGraph Neural NetworkAuto EncoderTextPoint CloudGraph
🎯 What it does: Designed a hypergraph-based rumor source detection method called HyperDet, which constructs high-order interaction relationships and automatically learns node representations by utilizing the Interactive Relationship Construction (IRC) module and the Feature-Rich Attention Fusion (FAF) module, thereby achieving source node localization.
Hypernetwork Aggregation for Decentralized Personalized Federated Learning
Weishi Li (National University of Defense Technology), Li Shen (Sun Yat-sen University)
Federated LearningImage
🎯 What it does: Propose a decentralized personalized federated learning framework called DFedHP based on hypernetworks, which generates shared model parameters using hypernetworks to reduce communication overhead and improve convergence speed;
HyperTrans: Efficient Hypergraph-Driven Cross-Domain Pattern Transfer in Image Anomaly Detection
Tengyu Zhang (Xi'an Jiaotong University), Zongze Wu (Shenzhen University)
Domain AdaptationAnomaly DetectionGraph Neural NetworkTransformerImageMultimodalityPoint CloudBenchmark
🎯 What it does: Proposes the HyperTrans method to achieve cross-domain few-shot industrial defect detection, leveraging hypergraphs and a perturbation correction framework to transfer and fuse features from the source domain RGB images with the target domain 3D depth maps.
ID-RemovalNet: Identity Removal Network for EEG Privacy Protection with Enhancing Decoding Tasks
Huabin Wang (Anhui University), Zhao Lv (Anhui University)
Safty and PrivacyConvolutional Neural NetworkGenerative Adversarial NetworkContrastive LearningBiomedical Data
🎯 What it does: This paper proposes ID-RemovalNet, designed to enhance task decoding accuracy while protecting EEG identity privacy.
Identifying and Reusing Learnwares Across Different Label Spaces
Jian-Dong Liu (Nanjing University), Zhi-Hua Zhou (Nanjing University)
Computational EfficiencyKnowledge DistillationImageTabular
🎯 What it does: Proposes a method to identify and reuse combinations of learned models (learnware) across different label spaces to address the problem of missing label spaces in user tasks.
Identifying Causal Mechanism Shifts Under Additive Models with Arbitrary Noise
Yewei Xia (Fudan University), Shuigeng Zhou (Fudan University)
Domain AdaptationExplainability and InterpretabilityScore-based ModelGraphBiomedical Data
🎯 What it does: Propose an algorithm called CMSI that uses a hybrid distribution score function to identify changes in causal mechanisms under different environments, applicable to additive noise models with arbitrary noise distributions.
Identifying Drivers of Predictive Aleatoric Uncertainty
Pascal Iversen (University of Potsdam), Bernhard Y. Renard (University of Potsdam)
Explainability and InterpretabilityImageTabular
🎯 What it does: This paper proposes a simple method that modifies neural network outputs to follow a Gaussian distribution and directly interprets the variance output using existing interpreters to explain the model's predictive uncertainty.
IE-PMMA:Point Cloud Completion Through Inverse Edge-aware Upsampling and Precise Multi-Modal Feature Alignment
Ran Jia (Sichuan University), Kelei Wang (Sichuan University)
RestorationMultimodalityPoint Cloud
🎯 What it does: Proposed a point cloud completion framework named IE-PMMA, combining inverse edge-aware upsampling and precise multi-modal feature alignment;
ILIF: Temporal Inhibitory Leaky Integrate-and-Fire Neuron for Overactivation in Spiking Neural Networks
Kai Sun (Monash University), Bin Zhang (Northeastern University)
ClassificationRecognitionSpiking Neural NetworkImageTime Series
🎯 What it does: Propose a temporal inhibition-type spiking neuron (ILIF) based on biological inhibition mechanisms, addressing the two major issues of over-activation and gradient vanishing in traditional LIF neurons caused by the proxy gradient support width γ, through two inhibition units (MPIU and CIU).
Image-Enhanced Hybrid Encoding with Reinforced Contrastive Learning for Spatial Domain Identification in Spatial Transcriptomics
Daoyuan Wang (Central South University), Fei Guo (Central South University)
Graph Neural NetworkTransformerReinforcement LearningAuto EncoderContrastive LearningImageMultimodalityBiomedical Data
🎯 What it does: Propose the IE-HERCL framework, which integrates gene expression, spatial coordinates, and tissue images for spatial domain identification in spatial transcriptomics through hybrid encoding, cross-attention, and enhanced contrastive learning.
Imagination-Limited Q-Learning for Offline Reinforcement Learning
Wenhui Liu (East China Normal University), Shuigeng Zhou (Fudan University)
Reinforcement LearningDiffusion modelWorld Model
🎯 What it does: Proposed a new offline reinforcement learning method called Imagination-Limited Q-learning (ILQ), which corrects value estimation by generating out-of-distribution (OOD) action values through a model and capping them with the maximum value of the behavior policy.
Imitation Learning via Focused Satisficing
Rushit N. Shah (University of Illinois Chicago), Brian Ziebart (University of Illinois Chicago)
Reinforcement Learning
🎯 What it does: Propose a imitation learning method called MinSubFI based on satisfaction theory, which directly optimizes the policy through subdominance minimization to generate acceptable behaviors within an unknown demonstrator's acceptable set.
Implicitly Aligning Humans and Autonomous Agents through Shared Task Abstractions
Stéphane Aroca-Ouellette (University of Colorado Boulder), Alessandro Roncone
Explainability and InterpretabilityReinforcement Learning from Human FeedbackReinforcement LearningAgentic AI
🎯 What it does: Propose a hierarchical agent architecture named HA2, leveraging hierarchical reinforcement learning to achieve zero-shot collaboration, and conduct cooperative experiments with humans and unknown agents in the Overcooked game.
Improved Approximation Ratio for Strategyproof Facility Location on a Cycle
Krzysztof Rogowski (University of Warsaw), Marcin Dziubiński (University of Warsaw)
OptimizationGraph
🎯 What it does: Studied the facility location problem on cyclic graphs where strategy-proofness without money is challenging, proposed a hybrid random mechanism combining the random representative mechanism with the proportional circular distance mechanism, and proved that its approximation ratio does not exceed 7/4; subsequently, numerical experiments further observed that this ratio may decrease to 3/2.
Improved MMS Approximations for Few Agent Types
Parnian Shahkar (University of California, Irvine), Jugal Garg (University of Illinois at Urbana-Champaign)
Optimization
🎯 What it does: Studied fair allocation of indivisible items with few types, proposing an improved approximation guarantee for the maximin share (MMS).
Improved Rank Aggregation Under Fairness Constraint
Diptarka Chakraborty (National University of Singapore), Alvin Hong Yao Yan (National University of Singapore)
OptimizationTabular
🎯 What it does: This paper proposes a new fair ranking aggregation algorithm that provides a (2+ε) approximation for the Kendall-tau count.