IJCAI 2024 Papers — Page 4
International Joint Conference on Artificial Intelligence · 790 papers
FedFa: A Fully Asynchronous Training Paradigm for Federated Learning
Haotian Xu (Hong Kong Polytechnic University), Jiannong Cao (Hong Kong Polytechnic University)
Federated LearningConvolutional Neural NetworkTransformerImageTextAudio
🎯 What it does: Propose FedFa, a fully asynchronous federated learning parameter update strategy that eliminates synchronization waiting and ensures convergence.
FedGCS: A Generative Framework for Efficient Client Selection in Federated Learning via Gradient-based Optimization
Zhiyuan Ning (Chinese Academy of Sciences), Yuanchun Zhou (Chinese Academy of Sciences)
OptimizationFederated LearningRecurrent Neural NetworkLarge Language ModelImageText
🎯 What it does: Propose the FedGCS framework, reformulating the client selection task in federated learning as a generation task, generating the optimal subset of clients through gradient optimization in a continuous representation space.
FedPFT: Federated Proxy Fine-Tuning of Foundation Models
Zhaopeng Peng (Xiamen University), Cheng Wang (Xiamen University)
Federated LearningKnowledge DistillationTransformerSupervised Fine-TuningImageText
🎯 What it does: Propose the FedPFT method, which performs proxy fine-tuning on a base model through federated learning without sharing server models or client data.
FedSSA: Semantic Similarity-based Aggregation for Efficient Model-Heterogeneous Personalized Federated Learning
Liping Yi (Nankai University), Xiaoxiao Li (University Of British Columbia)
ClassificationFederated LearningImage
🎯 What it does: Propose a FedSSA framework for model heterogeneity personalized federated learning, which can achieve local and global knowledge transfer through semantic similarity aggregation and adaptive parameter stabilization without sharing a public dataset.
FedTAD: Topology-aware Data-free Knowledge Distillation for Subgraph Federated Learning
Yinlin Zhu (Sun Yat-sen University), Rong-Hua Li (Beijing Institute of Technology)
Federated LearningKnowledge DistillationGraph Neural NetworkGraph
🎯 What it does: In subgraph federated learning, first decouple node and topological heterogeneity, revealing their impact on class-level knowledge reliability, and based on this, propose FedTAD: a topology-aware data-unsupervised knowledge distillation method, helping the global model more reliably integrate knowledge from local models.
Feedback-Based Adaptive Crossover-Rate in Evolutionary Computation
Xiaoyuan Guan (Sun Yat-sen University), Yuren Zhou (Sun Yat-sen University)
OptimizationReinforcement LearningBenchmark
🎯 What it does: Proposed a feedback-based adjustable crossover rate multi-objective evolutionary algorithm, improving crossover point selection and introducing virtual points to prevent probability drift;
Finding Increasingly Large Extremal Graphs with AlphaZero and Tabu Search
Abbas Mehrabian (Google DeepMind), Adam Zsolt Wagner
OptimizationGraph Neural NetworkReinforcement LearningGraphBenchmark
🎯 What it does: This paper proposes a new benchmark based on learning search, leveraging AlphaZero and incremental Tabu search to discover better graphs in extremal graph theory problems (finding the maximum number of edges in graphs without 3-cycles and 4-cycles).
Fine-grained Analysis of Stability and Generalization for Stochastic Bilevel Optimization
Xuelin Zhang (Huazhong Agricultural University), Feng Zheng (Southern University of Science and Technology)
OptimizationImage
🎯 What it does: This paper proposes a generalization analysis framework based on 'on-average argument stability' for Stochastic Bilevel Optimization (SBO), providing precise bounds on stability and generalization error for two mainstream first-order algorithms—Single-Scale Stochastic Gradient Descent (SSGD) and Two-Scale Stochastic Gradient Descent (TSGD). Subsequent experiments validate the theoretical predictions.
Fine-tuning Pre-trained Models for Robustness under Noisy Labels
Sumyeong Ahn (Michigan State University), Se-Young Yun (Korea Advanced Institute of Science and Technology)
ClassificationData-Centric LearningSupervised Fine-TuningImage
🎯 What it does: A two-step fine-tuning method is proposed on datasets with noisy labels: first, a robust classifier is obtained using linear probing (LP), then a Gaussian Mixture Model (GMM) selects clean samples, followed by full fine-tuning (FFT) on these samples, effectively enhancing the generalization performance of pre-trained models under noisy labels.
FineFMPL: Fine-grained Feature Mining Prompt Learning for Few-Shot Class Incremental Learning
Hongbo Sun (Peking University), Yuxin Peng (Peking University)
ClassificationRepresentation LearningMeta LearningPrompt EngineeringVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Propose the Fine-grained Feature Mining Prompt Learning (FineFMPL) method, which employs visual probe prompts and text context prompts on vision-language pre-trained models to construct visual and text prototypes, achieving high performance in few-shot class incremental learning.
Finite Groundings for ASP with Functions: A Journey through Consistency
Lukas Gerlach (TU Dresden), Markus Hecher (Massachusetts Institute of Technology)
🎯 What it does: This paper studies the extreme undecidability of consistency determination in answer set programming (ASP) with function symbols, and proposes two types of programs, 'frugal' and 'non-proliferous,' to achieve semi-decision; meanwhile, it presents an improved solver based on forbidden atoms and an algorithm capable of generating finite effective grounding.
Finite-Time Convergence Rates of Decentralized Local Markovian Stochastic Approximation
Pengfei Wang (Zhejiang University), Nenggan Zheng (Zhejiang University)
CompressionOptimizationFederated LearningBenchmark
🎯 What it does: Proposed distributed Markov stochastic approximation algorithm DLMSA and its compressed variant C-DLMSA for achieving near-optimal convergence rates in multi-agent systems.
First-Order Progression beyond Local-Effect and Normal Actions
Daxin Liu (University of Edinburgh), Jens Claßen (Roskilde University)
🎯 What it does: This paper studies the progression problem in action theory, proposing a new class of 'acyclic actions' and proving that progression can be defined and effectively computed under this class using first-order logic.
FlagVNE: A Flexible and Generalizable Reinforcement Learning Framework for Network Resource Allocation
Tianfu Wang (University of Science and Technology of China), Hui Xiong (Hong Kong University of Science and Technology)
OptimizationMeta LearningGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: Propose FlagVNE, a flexible and generalizable Virtual Network Embedding (VNE) framework that leverages reinforcement learning, graph neural networks, and meta-learning.
FLDM-VTON: Faithful Latent Diffusion Model for Virtual Try-on
Chenhui Wang (Fudan University), Hongming Shan (Fudan University)
Image TranslationGenerationTransformerDiffusion modelOptical FlowImage
🎯 What it does: Propose FLDM-VTON, a virtual try-on system based on latent diffusion models, utilizing deformed clothing as a prior and incorporating a clothing flattening network for detail supervision, improving the sampling strategy to enhance clothing detail fidelity.
Focus on the Whole Character: Discriminative Character Modeling for Scene Text Recognition
Bangbang Zhou (University of Science and Technology of China), Hongtao Xie (University of Science and Technology of China)
RecognitionTransformerContrastive LearningImage
🎯 What it does: Designed and implemented a scene text recognition model called Character Features Enriched (CFE), specifically addressing Large Intra-Class Variance (LICV) and Small Inter-Class Variance (SICV) issues to enhance the discriminability of character features.
Formal Verification of Parameterised Neural-symbolic Multi-agent Systems
Panagiotis Kouvaros (University of Limassol), Cosmo De Bonis-Campbell (University of Kent)
OptimizationExplainability and InterpretabilityContrastive LearningOptical Flow
🎯 What it does: Proposed a formal verification method for parameterized neural symbolic multi-agent systems, along with the corresponding abstract model and solving program.
Formalisation and Evaluation of Properties for Consequentialist Machine Ethics
Raynaldio Limarga (University of New South Wales), Maurice Pagnucco (University of New South Wales)
Review/Survey Paper
🎯 What it does: Formalize the consequentialist principle in machine ethics and define ethical permissible plans and a set of judgment attributes within the situation calculus framework.
Fostering Collective Action in Complex Societies Using Community-Based Agents
Jonathan Skaggs (Brigham Young University), Jacob W. Crandall (Brigham Young University)
Graph
🎯 What it does: This paper proposes Junior High Game (JHG) as a testbed for simulating complex social collective actions, and designs a community-based agent CAB that learns to form and maintain communities to achieve collective goals.
Fraud Risk Mitigation in Real-Time Payments: A Strategic Agent-Based Analysis
Katherine Mayo (University of Michigan), Michael P. Wellman (University of Michigan)
Anomaly DetectionAgentic AIGraphFinance Related
🎯 What it does: This paper analyzes the strategies employed by banks in real-time payments (RTP) to combat fraud through threshold setting and investment in fraud detection using a surrogate model.
FreqFormer: Frequency-aware Transformer for Lightweight Image Super-resolution
Tao Dai (Shenzhen University), Zexuan Zhu (Shenzhen University)
Super ResolutionTransformerImage
🎯 What it does: Introduces FreqFormer, a lightweight Transformer that integrates spatial, frequency, and channel information to enhance image super-resolution performance.
From Optimization to Generalization: Fair Federated Learning against Quality Shift via Inter-Client Sharpness Matching
Nannan Wu (Huazhong University of Science and Technology), Li Yu (Huazhong University of Science and Technology)
ClassificationFederated LearningImageBiomedical DataComputed Tomography
🎯 What it does: This paper addresses image quality imbalance caused by equipment differences in medical images, proposing the FedISM method under a federated learning framework to achieve unified sharpness across clients, thereby improving the model's generalization performance on low-quality images while maintaining performance on high-quality images.
Full Bayesian Significance Testing for Neural Networks in Traffic Forecasting
Zehua Liu (Beihang University), Yue He (Tsinghua University)
Autonomous DrivingOptimizationExplainability and InterpretabilityRecurrent Neural NetworkGraph Neural NetworkGraphTime Series
🎯 What it does: This paper proposes a full Bayesian significance testing framework, STn FBST, which models traffic prediction using Bayesian neural networks. It can simultaneously quantify and utilize both aleatoric and epistemic uncertainties, and employs gradient evidence for significance testing to reveal traffic spatiotemporal evolution patterns;
Fusion from a Distributional Perspective: A Unified Symbiotic Diffusion Framework for Any Multisource Remote Sensing Data Classification
Teng Yang (Xidian University), Yueguang Yang (Xidian University)
ClassificationTransformerVision Language ModelDiffusion modelMultimodality
🎯 What it does: Propose a unified self-supervised SymDiffuser framework to achieve joint classification of remote sensing images from any two sources.
G2LTraj: A Global-to-Local Generation Approach for Trajectory Prediction
Zhanwei Zhang (Zhejiang University), Wenxiao Wang (Zhejiang University)
GenerationAutonomous DrivingRecurrent Neural NetworkTime SeriesSequential
🎯 What it does: Propose a global-local generation framework called G2LTraj for predicting the future trajectories of traffic agents;
Game Transformations That Preserve Nash Equilibria or Best-Response Sets
Emanuel Tewolde (Carnegie Mellon University), Vincent Conitzer (Carnegie Mellon University)
🎯 What it does: Studied the computational complexity of determining best responses and game equivalence in multi-player games, systematically classified separable game transformations, and proved that only positive projection transformations (PAT) are uniquely effective in universally preserving Nash equilibria and best response sets.
General Epistemic Abstract Argumentation Framework: Semantics and Complexity
Gianvincenzo Alfano (University of Calabria), Irina Trubitsyna (University of Calabria)
🎯 What it does: Proposes an intuitive and descriptive semantics for general (including cyclic) Epistemic Abstract Argumentation Framework (EAAF), extending the semantics of traditional AAF and acyclic EAAF;
Generalized Taxonomy-Guided Graph Neural Networks
Yu Zhou (Tianjin University), Weixiong Zhang (Hong Kong Polytechnic University)
ClassificationRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposes a generic hierarchical classification guided graph neural network (TG-GNN), which constructs position-aware sibling networks and position-enhanced graph attention networks to learn the semantics and structure of hierarchical classification. It further integrates classification knowledge into network representation learning through a Markov random field (MRF) guided mechanism, thereby improving the quality of node embeddings.
Generate Synthetic Text Approximating the Private Distribution with Differential Privacy
Wenhao Zhao (Renmin University of China), Chunlai Zhou (Renmin University of China)
GenerationData SynthesisSafty and PrivacyText
🎯 What it does: This paper proposes a synthetic text generation framework based on differential privacy, which uses the idea of genetic algorithms to iteratively optimize the synthetic text distribution to maintain high similarity with the private text distribution. It also incorporates semantic readability detection and a limited domain selection mechanism to ensure dual guarantees of privacy and quality.
Generating More Audios for End-to-End Spoken Language Understanding
Xuxin Cheng (Peking University), Yuexian Zou (Peking University)
Data SynthesisTransformerLarge Language ModelSupervised Fine-TuningAudio
🎯 What it does: This paper proposes the GMA-SLU framework, which leverages a two-stage language model to generate more speech data through labels, thereby improving the performance of end-to-end speech understanding without transcriptions.
GenSeg: On Generating Unified Adversary for Segmentation
Yuxuan Zhang (University of Science and Technology of China), Yinxing Xue (University of Science and Technology of China)
SegmentationAdversarial AttackGenerative Adversarial NetworkImage
🎯 What it does: Propose a generative unified adversarial attack framework called GenSeg, designed to attack semantic, instance, and panoptic segmentation models.
Geometry-Guided Conditional Adaptation for Surrogate Models of Large-Scale 3D PDEs on Arbitrary Geometries
Jingyang Deng (Peking University), Jinwen Ma (Peking University)
Computational EfficiencyTransformerPoint CloudMeshPhysics Related
🎯 What it does: Propose a general framework 3DGeoCA, which extracts boundary geometric features through a point cloud geometry encoder and performs conditional adaptation in the hidden layer to improve the prediction accuracy of large-scale 3D PDEs (steady-state Navier-Stokes) on arbitrary geometries.
Getting More by Knowing Less: Bayesian Incentive Compatible Mechanisms for Fair Division
Vasilis Gkatzelis (Drexel University), Paritosh Verma (Purdue University)
Optimization
🎯 What it does: This paper studies the use of Bayesian Incentive Compatibility (BIC) mechanisms to address the conflict between fairness and efficiency in traditional Dominant Strategy Incentive Compatibility (DSIC) mechanisms for fair resource allocation.
Global Optimality of Single-Timescale Actor-Critic under Continuous State-Action Space: A Study on Linear Quadratic Regulator
Xuyang Chen (National University Of Singapore), Lin Zhao (National University Of Singapore)
OptimizationReinforcement Learning
🎯 What it does: This paper proves that the single-sample single-time-scale actor-critic (Actor-Critic) algorithm globally converges to the optimal policy for the linear quadratic regulator (LQR) problem in continuous state-action spaces, with a sample complexity of O(ε⁻²).
Gradformer: Graph Transformer with Exponential Decay
Chuang Liu (Wuhan University), Wenbin Hu (Wuhan University)
Representation LearningDrug DiscoveryProtein Structure PredictionGraph Neural NetworkTransformerGraphBiomedical Data
🎯 What it does: Designed a novel Graph Transformer called Gradformer, which incorporates graph structural priors naturally during information aggregation by multiplying an exponential decay mask on the self-attention matrix, using graph structure distances (short paths) to decay attention weights.
Graph Attention Network with High-Order Neighbor Information Propagation for Social Recommendation
Fei Xiong (Beijing Jiaotong University), Liang Wang (Northwestern Polytechnical University)
Recommendation SystemRecurrent Neural NetworkGraph Neural NetworkGraph
🎯 What it does: Propose GAIPSRec: a social recommendation model based on graph attention networks, which effectively propagates high-order neighbor information through random walk sampling, path encoding, and path aggregation.
Graph Collaborative Expert Finding with Contrastive Learning
Qiyao Peng (Tianjin University), Minglai Shao (Tianjin University)
Recommendation SystemGraph Neural NetworkTransformerContrastive LearningTextGraph
🎯 What it does: This paper proposes a contrastive learning-based expert discovery model called CGEF, which utilizes an expert-question interaction graph and a graph attention network to capture higher-order connections, and enhances expert representations under sparse interactions through behavior-level and interest-level contrastive tasks.
Graph Contrastive Learning with Reinforcement Augmentation
Ziyang Liu (Tsinghua University), Cheng Wu (Tsinghua University)
Representation LearningGraph Neural NetworkReinforcement LearningContrastive LearningGraph
🎯 What it does: This paper proposes a graph contrastive learning framework based on reinforcement learning called Graph Reinforcement Augmentation (GRA), and implements the GA2C model, achieving continuous evolution of views and structural information retention during graph data augmentation by incorporating the Actor-Critic mechanism.
GRASP: A Novel Benchmark for Evaluating Language GRounding and Situated Physics Understanding in Multimodal Language Models
Serwan Jassim (Osnabrück University), Elia Bruni (Osnabrück University)
TransformerLarge Language ModelPrompt EngineeringVision Language ModelVideoMultimodalityBenchmarkPhysics RelatedChain-of-Thought
🎯 What it does: Propose and release the GRASP benchmark, utilizing Unity-generated videos to evaluate the ability of multimodal large language models (LLMs) in language localization and intuitive physics understanding.
Group-Aware Coordination Graph for Multi-Agent Reinforcement Learning
Wei Duan (University of Technology Sydney), Junyu Xuan (University of Technology Sydney)
Graph Neural NetworkReinforcement LearningBenchmark
🎯 What it does: Proposed Group-Aware Coordination Graph (GACG), a dynamic coordination graph that simultaneously considers individual pairwise relationships and group-level dependencies, for information exchange and decision-making in multi-agent reinforcement learning;
Guidance Graph Optimization for Lifelong Multi-Agent Path Finding
Yulun Zhang (Carnegie Mellon University), Jiaoyang Li (Carnegie Mellon University)
OptimizationConvolutional Neural NetworkGraph
🎯 What it does: Proposes the concept of 'guidance graph' for lifelong multi-agent pathfinding (MAPF), and studies the task of improving system throughput by optimizing the edge weights of the graph;
GUIDE: A Guideline-Guided Dataset for Instructional Video Comprehension
Jiafeng Liang (Harbin Institute of Technology), Bing Qin (Harbin Institute of Technology)
TransformerLarge Language ModelVision Language ModelVideoTextMultimodalityBenchmark
🎯 What it does: Constructed the GUIDE dataset, containing 560 tasks, 3.5K videos, 15K specific steps and guiding principles, and proposed three benchmark tasks.
Guiding GBFS through Learned Pairwise Rankings
Mingyu Hao (Australian National University), Jörg Hoffmann
OptimizationComputational EfficiencyGraph Neural NetworkGraphBenchmark
🎯 What it does: Propose a learning framework based on ranking, replacing traditional heuristics with pairwise ranking to guide GBFS, and integrating it with DirectRanker;
Hacking Task Confounder in Meta-Learning
Jingyao Wang (Institute of Software Chinese Academy of Sciences), Wenwen Qiang (Institute of Software Chinese Academy of Sciences)
Explainability and InterpretabilityRepresentation LearningMeta LearningDrug DiscoveryImageTextTabularBiomedical Data
🎯 What it does: This paper reveals the 'task confounder' in meta-learning, which causes negative knowledge transfer problems, through a causal model, and proposes a pluggable meta-learning causal representation learner (MetaCRL) to eliminate this confounding factor and enhance model generalization.
Hard-Thresholding Meets Evolution Strategies in Reinforcement Learning
Chengqian Gao (Mohamed bin Zayed University of Artificial Intelligence), Zhiqiang Xu (Mohamed bin Zayed University of Artificial Intelligence)
Reinforcement LearningImage
🎯 What it does: Proposed an algorithm called NESHT that combines hard thresholding (Hard-Thresholding) with natural evolution strategies (NES) to remove task-irrelevant features and achieve sparse policies in reinforcement learning.
HeterGCL: Graph Contrastive Learning Framework on Heterophilic Graph
Chenhao Wang (Heilongjiang University), Wei Li (Harbin Engineering University)
Representation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Propose HeterGCL, an unsupervised graph contrastive learning framework for heterogeneous graphs, integrating structural and semantic information through adaptive neighbor aggregation and multi-layer contrastive loss for representation learning.
Heterogeneous Causal Metapath Graph Neural Network for Gene-Microbe-Disease Association Prediction
Kexin Zhang (Huazhong Agricultural University), Wen Zhang (Huazhong Agricultural University)
Graph Neural NetworkGraphBiomedical Data
🎯 What it does: Construct a gene-microbe-disease heterogeneous graph, extract subgraphs using six causal meta-paths, and design a causal semantic shared message passing network combined with subgraph attention fusion to predict triplet associations.
Heterogeneous Graph Transformer with Poly-Tokenization
Zhiyuan Lu (Beijing University of Posts and Telecommunications), Chuan Shi (Beijing University of Posts and Telecommunications)
Representation LearningGraph Neural NetworkTransformerGraph
🎯 What it does: Propose Poly-tokenized Heterogeneous Graph Transformer (PHGT) for node-level representation learning on heterogeneous graphs.
Hierarchical Reinforcement Learning for Point of Interest Recommendation
Yanan Xiao (Northeast Normal University), Minghao Yin (Northeast Normal University)
Recommendation SystemReinforcement LearningSequential
🎯 What it does: Researchers proposed the HRL-PRP framework, which uses hierarchical reinforcement learning to preprocess users' historical POI data, removing noise points to improve the accuracy of POI recommendations.
Hierarchical Reinforcement Learning on Multi-Channel Hypergraph Neural Network for Course Recommendation
Lu Jiang (Northeast Normal University), Minghao Yin (Northeast Normal University)
Recommendation SystemGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: This paper proposes a hierarchical reinforcement learning framework (HHCoR) based on multi-channel hypergraph neural networks to address the problem of personalized course recommendation in MOOCs;
Higher-Order Argumentation Frameworks: Principles and Gradual Semantics
Leila Amgoud (CNRS - IRIT), Marie-Christine Lagasquie-Schiex (Universite Toulouse 3 - IRIT)
🎯 What it does: Proposed progressive semantics for the high-order attack framework (HO-AF), assigning values to arguments and attacks simultaneously for the first time;
History Repeats Itself: A Baseline for Temporal Knowledge Graph Forecasting
Julia Gastinger (NEC Laboratories Europe), Heiner Stuckenschmidt (University of Mannheim)
Representation LearningGraphTime Series
🎯 What it does: Propose an intuitive baseline model based on fact recursion for temporal knowledge graph prediction.
How Hard Is It to Impact the Impact of Your Paper?
Yongjie Yang (Saarland University)
Graph
🎯 What it does: Studied and proved that the manipulation problem of the CD index (measuring the destructiveness/fusiveness of scientific papers) is generally intractable under paper merging, adding, and deletion operations in parameterized complexity.
How to Learn Domain-Invariant Representations for Visual Reinforcement Learning: An Information-Theoretical Perspective
Shuo Wang (Beijing Jiaotong University), Kai Lv (Beijing Jiaotong University)
Domain AdaptationAutonomous DrivingRepresentation LearningRobotic IntelligenceReinforcement LearningImageVideo
🎯 What it does: This paper proposes a visual reinforcement learning framework called MIIR based on information theory, which improves the generalization performance of visual RL by learning domain-invariant representations.
Hundred-Kilobyte Lookup Tables for Efficient Single-Image Super-Resolution
Binxiao Huang (University of Hong Kong), Ngai Wong (University of Hong Kong)
Super ResolutionConvolutional Neural NetworkImage
🎯 What it does: Propose an extremely compact lookup table (HKLUT) to achieve single-image super-resolution.
Hundredfold Accelerating for Pathological Images Diagnosis and Prognosis through Self-reform Critical Region Focusing
Xiaotian Yu (Zhejiang University), Zunlei Feng (Zhejiang University)
ClassificationComputational EfficiencyTransformerBiomedical Data
🎯 What it does: Propose a self-reconfigurable multi-layer Transformer (SMT), accelerating pathological image diagnosis and prognosis through hierarchical focusing from macro to micro and backward reconsideration.
HVOFusion: Incremental Mesh Reconstruction Using Hybrid Voxel Octree
Shaofan Liu (Zhejiang University), Jianke Zhu (Zhejiang University)
OptimizationSimultaneous Localization and MappingPoint CloudMesh
🎯 What it does: Proposes HVOFusion, an incremental mesh reconstruction method utilizing a hybrid voxel-octree structure, capable of online generating and directly optimizing explicit triangular meshes, combining sparse storage with high surface quality.
Hybrid Frequency Modulation Network for Image Restoration
Yuning Cui (Technical University of Munich), Alois Knoll (Technical University of Munich)
RestorationConvolutional Neural NetworkImage
🎯 What it does: Proposes the CSNet network based on channel and spatial dual-frequency modulation for image restoration tasks such as dehazing, defocusing blur removal, and snow removal.
HyDiscGAN: A Hybrid Distributed cGAN for Audio-Visual Privacy Preservation in Multimodal Sentiment Analysis
Zhuojia Wu (Tongji University), Liang Hu (Tongji University)
Federated LearningSafty and PrivacyTransformerGenerative Adversarial NetworkContrastive LearningMultimodality
🎯 What it does: Designed the HyDiscGAN framework, leveraging hybrid distributed learning and cross-modal conditional GANs to generate audio and visual pseudo-features under text conditions, achieving privacy-preserving multimodal sentiment analysis.
HypBO: Accelerating Black-Box Scientific Experiments Using Experts’ Hypotheses
Abdoulatif Cissé (University of Liverpool), Andrew I. Cooper (University of Liverpool)
OptimizationHyperparameter Search
🎯 What it does: Proposed a dual-layer Bayesian optimization method called HypBO based on expert hypotheses, which utilizes parameter interval constraints provided by experts to guide the search and accelerate convergence in black-box scientific experiments.
Hypergraph Self-supervised Learning with Sampling-efficient Signals
Fan Li (University of New South Wales), Xuemin Lin (Shanghai Jiaotong University)
Computational EfficiencyRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposed the SE-HSSL framework, leveraging unsampled CCA objectives and hierarchical member contrastive learning to achieve self-supervised representation learning on hypergraphs.
Hyperparameter Optimization Can Even Be Harmful in Off-Policy Learning and How to Deal with It
Yuta Saito (Cornell University), Masahiro Nomura (CyberAgent Inc)
OptimizationHyperparameter SearchReinforcement LearningTabular
🎯 What it does: This paper investigates the optimistic bias and unsafe behaviors that occur when using typical unbiased estimators (e.g., IPS) as proxy targets for automatic hyperparameter optimization (HPO) in offline (off-policy) learning, and proposes the CIR-HPO algorithm to address these two issues.
HyQ: Hardware-Friendly Post-Training Quantization for CNN-Transformer Hybrid Networks
Nam Joon Kim (Seoul National University of Science and Technology), Hyun Kim (Seoul National University of Science and Technology)
ClassificationComputational EfficiencyConvolutional Neural NetworkTransformerImage
🎯 What it does: Propose a post-training quantization (PTQ) method for hybrid models combining CNN and Vision Transformer, achieving 8-bit integer inference while maintaining high accuracy.
Image Retrieval with Self-Supervised Divergence Minimization and Cross-Attention Classification
Vivek Trivedy (Temple University), Longin Jan Latecki (Temple University)
ClassificationRetrievalTransformerContrastive LearningImage
🎯 What it does: Propose the DMCAC framework, which jointly learns query and database embeddings through self-supervised distribution similarity minimization and cross-attention classification;
IMM: An Imitative Reinforcement Learning Approach with Predictive Representation Learning for Automatic Market Making
Hui Niu (Tsinghua University), Jian Guo (International Digital Economy Academy)
Representation LearningConvolutional Neural NetworkReinforcement LearningTime SeriesFinance Related
🎯 What it does: Propose the IMM framework in high-frequency trading to achieve automated market making across multiple price levels
Imperfect-Recall Games: Equilibrium Concepts and Their Complexity
Emanuel Tewolde (Carnegie Mellon University), Vincent Conitzer (Carnegie Mellon University)
Optimization
🎯 What it does: This paper systematically studies the computational complexity of solving Nash, EDT, and CDT equilibria in imperfect-recall games, and presents multiple new results in complexity classes such as NP, ΣP₂, ∃R, ∃∀R, PPAD, PLS, and CLS;
Imperio: Language-Guided Backdoor Attacks for Arbitrary Model Control
Ka-Ho Chow (University of Hong Kong), Lei Yu (Rensselaer Polytechnic Institute)
Adversarial AttackConvolutional Neural NetworkLarge Language ModelImageText
🎯 What it does: This paper proposes a language-guided backdoor attack method named Imperio, enabling attackers to achieve arbitrary output control over image classification models through natural language instructions.
Implicit Prompt Learning for Image Denoising
Yao Lu (Harbin Institute of Technology Shenzhen), Bob Zhang (University of Macau)
RestorationConvolutional Neural NetworkImage
🎯 What it does: Propose an image denoising framework IPLID based on implicit prompt learning, which utilizes frozen weights of pre-trained visual models, generates adaptive prompts through linear prompt blocks (LP), fuses multi-scale prompt features via compact feature fusion blocks (CFF), and enhances optimization effects through gradient accumulation learning (GA).
Imprecise Probabilities Meet Partial Observability: Game Semantics for Robust POMDPs
Eline M. Bovy (Radboud University), Nils Jansen (Radboud University)
OptimizationReinforcement Learning
🎯 What it does: Under the theoretical framework of robust POMDP, a POSG model based on game semantics is proposed, and it is proven that different uncertainty assumptions (stickiness, order of actions) lead to different optimal values;
Improved Approximation Algorithms for Capacitated Location Routing
Jingyang Zhao (University of Electronic Science and Technology of China), Shunwang Wang (University of Electronic Science and Technology of China)
OptimizationBenchmark
🎯 What it does: Proposed two improved approximation algorithms (Tree-Alg and Path-Alg), achieving approximation ratios of 4.169 and 4.092, respectively, for solving the capacity-constrained location routing problem (CLR).
Improved Approximation of Weighted MMS Fairness for Indivisible Chores
Fangxiao Wang (Hong Kong Polytechnic University), Pinyan Lu (Shanghai University of Finance and Economics)
Optimization
🎯 What it does: Propose a WMMS approximation algorithm for fair division problems with different weight participants, providing an O(log n) approximation scheme for any n, achieving the optimal 1.366 approximation ratio in the two-person case; also present an online O(√n) competitive algorithm.
Improved Encodings of Acyclicity for Translating Answer Set Programming into Integer Programming
Masood Feyzbakhsh Rankooh (Tampere University), Tomi Janhunen (Tampere University)
OptimizationBenchmark
🎯 What it does: This paper proposes two novel integer programming (IP) translation methods to improve acyclicity constraints in answer set programming (ASP), thereby more effectively solving ASP problems.
Improved Evolutionary Algorithms for Submodular Maximization with Cost Constraints
Yanhui Zhu (Iowa State University), A. Pavan (Iowa State University)
OptimizationGraphTabularBenchmark
🎯 What it does: Proposed two evolutionary algorithms (EVO-SMC and ST-EVO-SMC) for solving monotonic submodular maximization with cost constraints.
Improved Parallel Algorithm for Non-Monotone Submodular Maximization under Knapsack Constraint
Tan D. Tran (Phenikaa University), Phuong N. H. Pham
OptimizationImageGraph
🎯 What it does: Proposed a new parallel algorithm AST for maximizing non-monotonic submodular functions under knapsack constraints, improving the approximation ratio while maintaining low adaptivity.
Improving Adversarial Robustness via Feature Pattern Consistency Constraint
Jiacong Hu (Zhejiang University), Mingli Song (Zhejiang University)
ClassificationAdversarial AttackConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: This paper proposes the Feature Pattern Consistency Constraint (FPCC) mechanism by constraining the consistency of hidden layer feature patterns in CNNs, enhancing the model's robustness against unknown adversarial samples.
Improving Multi-agent Reinforcement Learning with Stable Prefix Policy
Yue Deng (Zhejiang University), Yin Zhang (Zhejiang University)
Computational EfficiencyReinforcement LearningBenchmark
🎯 What it does: Construct a Monte-Carlo trajectory tree and introduce Stable Prefix Policy (SPP) to control the exploration-exploitation balance in multi-agent reinforcement learning.
Improving Paratope and Epitope Prediction by Multi-Modal Contrastive Learning and Interaction Informativeness Estimation
Zhiwei Wang (Huazhong Agricultural University), Wen Zhang (Huazhong Agricultural University)
Drug DiscoveryProtein Structure PredictionGraph Neural NetworkTransformerLarge Language ModelContrastive LearningMultimodalityGraphBiomedical Data
🎯 What it does: Propose the MIPE method, which utilizes multimodal contrastive learning of sequence and structural information, as well as interaction information estimation, to predict antibody and antigen binding sites (peptide epitopes).
Improving Pseudo Labels with Global-Local Denoising Framework for Cross-lingual Named Entity Recognition
Zhuojun Ding (Huazhong University of Science and Technology), Dangyang Chen (Ping An Property & Casualty Insurance Company of China)
RecognitionTransformerContrastive LearningText
🎯 What it does: Propose the GLoDe global-local denoising framework to enhance the generalization capability of cross-lingual named entity recognition models.
Improving Zero-Shot Cross-Lingual Transfer via Progressive Code-Switching
Zhuoran Li (Beihang University), Richong Zhang (Beihang University)
Domain AdaptationRepresentation LearningTransformerSupervised Fine-TuningText
🎯 What it does: Propose a Progressive Code-Switching (PCS) technique that enhances multilingual representations and improves zero-shot cross-lingual transfer performance by progressively generating code-mixed samples with increasing difficulty.
Incorporating Schema-Aware Description into Document-Level Event Extraction
Zijie Xu (Southeast University), Chenxiao Wu (Southeast University)
TransformerContrastive LearningTextFinance Related
🎯 What it does: This paper proposes the SEELE framework, which jointly learns document-level event extraction (trigger words and arguments) by leveraging natural language descriptions generated from event schemas.
Individual Causal Structure Learning from Population Data
Wei Chen (Guangdong University of Technology), Zhifeng Hao (Shantou University)
TabularBiomedical DataFinance Related
🎯 What it does: This paper proposes a new individual linear acyclic model (ILAM) and develops an individual causal structure learning (ICSL) method based on shared independent component analysis (ShICA), which can utilize aggregate data from multiple individuals to recover their specific causal structures even when each individual's sample size is limited.
Individual Fairness under Group Fairness Constraints in Bipartite Matching - One Framework to Approximate Them All
Atasi Panda (Indian Institute of Science), Prajakta Nimbhorkar (Chennai Mathematical Institute)
OptimizationTabular
🎯 What it does: Designed multiple multi-objective approximation algorithms to achieve individual fairness probability distribution matching while satisfying group fairness constraints
Individual Rationality in Topological Distance Games Is Surprisingly Hard
Argyrios Deligkas (Royal Holloway, University of London), Šimon Schierreich (Czech Technical University in Prague)
Computational Efficiency
🎯 What it does: The study investigates the computational complexity of finding individually rational outcomes in topological distance games, systematically demonstrating NP-hardness and parameterized complexity under various input constraints, and providing feasible XP algorithms and FPT conditions.
InfoMatch: Entropy Neural Estimation for Semi-Supervised Image Classification
Qi Han (Lanzhou University), Kun Zhan (Lanzhou University)
ClassificationRepresentation LearningConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: Design and implement the InfoMatch framework, combining information entropy neural estimation with pseudo labels and strong contrastive learning to enhance semi-supervised image classification performance.
Information-Theoretic Opacity-Enforcement in Markov Decision Processes
Chongyang Shi (University of Florida), Jie Fu (University of Florida)
OptimizationSafty and PrivacyReinforcement LearningTabular
🎯 What it does: Explored the planning problem of transparency (opacity) in Markov decision processes from an information-theoretic perspective, and proposed observability enhancement strategies for both terminal and initial states;
Innovative Directional Encoding in Speech Processing: Leveraging Spherical Harmonics Injection for Multi-Channel Speech Enhancement
Jiahui Pan (Inner Mongolia University), Xueliang Zhang (Inner Mongolia University)
RestorationRepresentation LearningRecurrent Neural NetworkAudio
🎯 What it does: This paper proposes using spherical harmonic transform (SHT) coefficients as auxiliary input, designing parallel and serial dual-encoder networks for multi-channel speech enhancement, achieving unified processing for different microphone array configurations.
Instance-Level Metalearning for Outlier Detection
Long Vu (IBM Research), Horst Samulowitz (IBM Research)
Anomaly DetectionMeta LearningTabular
🎯 What it does: This work proposes a novel anomaly detection method called T-AutoOD based on instance-level meta-learning. It learns the structural relationships between abnormal and normal instances from an imbalanced labeled classification dataset, and builds a meta-model to combine scores from multiple unsupervised anomaly detection pipelines, enabling anomaly detection on new unlabeled datasets.
Instantiations and Computational Aspects of Non-Flat Assumption-based Argumentation
Tuomo Lehtonen (University of Helsinki), Johannes P. Wallner (Graz University of Technology)
Computational EfficiencyBenchmark
🎯 What it does: This paper studies the instantiation and computation of non-flat assumption-based argumentation (non-flat ABA), proposing an instantiation method based on translation to the bipolar argumentation framework (BAF), and providing a direct ASP algorithm without instantiation;
InstructEdit: Instruction-Based Knowledge Editing for Large Language Models
Ningyu Zhang (Zhejiang University), Huajun Chen (Zhejiang University)
Representation LearningMeta LearningTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: Proposed an instruction-based knowledge editing method called InstructEdit, which can perform multi-task knowledge modification using a single unified editor.
InstructME: An Instruction Guided Music Edit Framework with Latent Diffusion Models
Bing Han (Shanghai Jiao Tong University), Xuchen Song (ByteDance)
GenerationData SynthesisTransformerLarge Language ModelDiffusion modelAuto EncoderTextAudio
🎯 What it does: An instruction-driven music editing framework called InstructME based on latent diffusion models, supporting operations such as adding, deleting, replacing, and mixing audio tracks.
Integrating Intent Understanding and Optimal Behavior Planning for Behavior Tree Generation from Human Instructions
Xinglin Chen (National University of Defense Technology), Ji Wang (National University of Defense Technology)
Robotic IntelligenceLarge Language ModelPrompt EngineeringText
🎯 What it does: This paper proposes a two-stage framework that first uses a large language model to interpret human natural language instructions into first-order logic goal conditions, and then generates an optimal and theoretically successful behavior tree through a new Optimal Behavior Tree Expansion Algorithm (OBTEA), enabling robots to automatically execute instructions.
Integrating Vision-Language Semantic Graphs in Multi-View Clustering
JunLong Ke (University of Electronic Science and Technology of China), Lifang He (Lehigh University)
Representation LearningData-Centric LearningGraph Neural NetworkVision Language ModelAuto EncoderContrastive LearningImageTextMultimodality
🎯 What it does: Propose a method called IVSGMV that constructs a semantic graph using the large-scale vision-language pre-trained model CLIP, and achieves unsupervised multi-view clustering through a graph joint process and adaptive hybrid graph filtering.
IntensPure: Attack Intensity-aware Secondary Domain Adaptive Diffusion for Adversarial Purification
Eun-Gi Lee (Chonnam National University), Seok Bong Yoo (Chonnam National University)
Domain AdaptationComputational EfficiencyAdversarial AttackConvolutional Neural NetworkDiffusion modelImage
🎯 What it does: Proposes IntensPure, an adversarial attack purification method for pedestrian re-identification, based on attack intensity awareness and quadratic domain adaptive diffusion, purifying only low/mid-frequency coefficients.
Intention Progression with Temporally Extended Goals
Yuan Yao (University of Nottingham Ningbo China), Brian Logan (University of Aberdeen)
Robotic IntelligenceReinforcement LearningAgentic AIBenchmark
🎯 What it does: Proposed a new BDI agent intent advancement method that can simultaneously and in parallel process multiple temporal extended goals within the same agent (including both reachability goals and invariant constraints), and allows seamless integration of manually written plans with reinforcement learning strategies.
Interpretable Network Visualizations: A Human-in-the-Loop Approach for Post-hoc Explainability of CNN-based Image Classification
Matteo Bianchi (Politecnico di Milano), Marco Brambilla (Politecnico di Milano)
ClassificationExplainability and InterpretabilityConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: Designed and implemented a post-hoc interpretable framework called Interpretable Network Visualizations (INV), which hierarchically displays CNN-extracted features and generates visual saliency maps by clustering feature maps. It also enhances interpretability through gamified crowdsourcing to collect textual labels.
Interpretable Tensor Fusion
Saurabh Varshneya (RPTU Kaiserslautern-Landau), Marius Kloft (RPTU Kaiserslautern-Landau)
Explainability and InterpretabilityMultimodality
🎯 What it does: Propose an interpretable tensor fusion method named InTense for multi-modal learning, which can simultaneously learn multi-modal representations and their interpretable fusion;
Invertible Residual Rescaling Models
Jinmin Li (Tsinghua University), Jingyun Zhang (Tencent)
Super ResolutionFlow-based ModelImage
🎯 What it does: Propose a reversible residual rescaling model (IRRM) that can efficiently learn bijective relationships between high-resolution and low-resolution images, and enhance high-frequency information reconstruction through residual modules.
It Ain’t That Bad: Understanding the Mysterious Performance Drop in OOD Generalization for Generative Transformer Models
Xingcheng Xu (Shanghai Artificial Intelligence Laboratory), Yanqing Yang (Shanghai Artificial Intelligence Laboratory)
Explainability and InterpretabilityRepresentation LearningTransformerLarge Language ModelTextSequential
🎯 What it does: This work trains small generative Transformers such as NanoGPT/MinGPT on n-bit addition/multiplication tasks, systematically investigates ID and OOD generalization behaviors, and reveals the model's equivalence generalization mechanism during OOD scenarios through equivalence class analysis.
Joint Domain Adaptive Graph Convolutional Network
Niya Yang (Tianjin University), Di Jin (Tianjin University)
Domain AdaptationGraph Neural NetworkGraph
🎯 What it does: Propose a Joint Adversarial Domain Adaptation Graph Convolutional Network (JDA-GCN) for cross-network node classification tasks.
Joint Input and Output Coordination for Class-Incremental Learning
Shuai Wang (Wuhan University), Dacheng Tao (Nanyang Technological University)
ClassificationKnowledge DistillationImage
🎯 What it does: Proposed a joint input-output coordination mechanism (JIOC) to alleviate class imbalance and task interference in class-incremental learning.