IJCAI 2025 Papers — Page 3
International Joint Conference on Artificial Intelligence · 1014 papers
Contamination Budget: Trade-offs Between Breadth, Depth and Difficulty
Behzad Mehrbakhsh (Universitat Politecnica De Valencia), José Hernández-Orallo (Universitat Politecnica De Valencia)
Computational EfficiencyData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: This paper studies how to improve the performance of large language models (LLMs) under limited fine-tuning budgets by implementing targeted pollution intervention (i.e., memorizing specific error samples), with a focus on analyzing the trade-offs between the breadth, depth of intervention, and sample difficulty.
Continuous Diffusive Prediction Network for Multi-Station Weather Prediction
Chujie Xu (Beihang University), Xianglong Liu (Beihang University)
Convolutional Neural NetworkRecurrent Neural NetworkDiffusion modelOptical FlowTime SeriesPhysics Related
🎯 What it does: For multi-site weather forecasting, this paper proposes the Continuous Diffusion Prediction Network (CDPNet), which models spatial and temporal continuous weather changes through Continuous Calibration Initialization (CCI) and Diffusion Difference Estimation (DDE).
Continuous-Time Reward Machines
Amin Falah (University of Colorado Boulder), Ashutosh Trivedi
Autonomous DrivingReinforcement LearningBenchmark
🎯 What it does: This paper proposes the Continuous-Time Reward Machine (CTRM) framework, combining Continuous-Time Markov Decision Processes (CTMDPs) with reward machines to achieve reinforcement learning in temporally complex environments;
Contrastive Cross-Course Knowledge Tracing via Concept Graph Guided Knowledge Transfer
Wenkang Han (Zhejiang University), Jingyuan Chen (Zhejiang University)
Representation LearningGraph Neural NetworkTransformerLarge Language ModelContrastive LearningTextGraphSequential
🎯 What it does: This paper proposes a cross-course knowledge tracking model called TransKT, which constructs a cross-course concept graph using zero-shot large language models, achieves knowledge transfer through LLM-to-LM semantic enhancement and graph convolutional networks, and enhances knowledge state representation by incorporating cross-course contrastive learning.
Contrastive Unlearning: A Contrastive Approach to Machine Unlearning
Hong kyu Lee (Emory University), Li Xiong (Emory University)
Safty and PrivacyTransformerContrastive LearningImage
🎯 What it does: Propose contrastive learning on the model's representation space, utilizing positive and negative sample pairs to achieve machine model forgetting.
CorrDetail: Visual Detail Enhanced Self-Correction for Face Forgery Detection
Binjia Zhou (Zhejiang University), Yijun Bei (Zhejiang University)
Anomaly DetectionTransformerLarge Language ModelPrompt EngineeringMixture of ExpertsVision Language ModelContrastive LearningImageVideoMultimodality
🎯 What it does: Proposes the CorrDetail framework, which combines visual detail enhancement with self-correcting multimodal deepfake detection methods, capable of providing interpretable text prompts and effectively identifying image details.
Cost-Effective On-Device Sequential Recommendation with Spiking Neural Networks
Di Yu (Zhejiang University), Shuiguang Deng (Zhejiang University)
Recommendation SystemComputational EfficiencySpiking Neural NetworkSequential
🎯 What it does: Proposed the SSR model, enabling sequential recommendation on edge devices by converting dense embeddings into sparse spike representations to reduce memory and energy consumption.
Counterfactual Explanations for Continuous Action Reinforcement Learning
Shuyang Dong (University of Virginia), Lu Feng (University of Virginia)
Explainability and InterpretabilityReinforcement Learning
🎯 What it does: Proposed a counterfactual explanation generation method for continuous action reinforcement learning, extending the TD3 algorithm to enhance cumulative rewards while maintaining minimal action deviation.
Counterfactual Explanations Under Model Multiplicity and Their Use in Computational Argumentation
Gianvincenzo Alfano (University of Calabria), Francesca Toni (Imperial College London)
Explainability and Interpretability
🎯 What it does: Proposes a framework for counting counterfactual explanations (CX) under model diversity (Multiplicity) environments, providing qualitative definitions (∃∀, ∀∀, etc.) and quantitative definitions (k-CX), and instantiating them in an abstract argumentation (AA) environment.
Counterfactual Knowledge Maintenance for Unsupervised Domain Adaptation
Yao Li (China University of Mining and Technology), Bing Liu (China University of Mining and Technology)
Domain AdaptationTransformerVision Language ModelImage
🎯 What it does: Propose an unsupervised domain adaptation framework based on counterfactual causal disentanglement and discriminative knowledge preservation
Counterfactual Strategies for Markov Decision Processes
Paul Kobialka (University of Oslo), Einar Broch Johnsen (University of Oslo)
OptimizationAdversarial AttackReinforcement LearningTabularTime Series
🎯 What it does: This paper proposes a framework for solving adversarial strategies in Markov Decision Processes (MDP), and presents an algorithm based on nonlinear optimization to generate minimal modification strategies that reduce the reachability probability of target states, while supporting the generation of diverse strategies;
Counterfactual Thinking Driven Emotion Regulation for Image Sentiment Recognition
Xinyue Zhang (East China Normal University), Guitao Cao (East China Normal University)
RecognitionConvolutional Neural NetworkImage
🎯 What it does: Propose CTERNet based on emotion regulation theory, which utilizes causal counterfactual thinking to simultaneously learn real and potential emotional regions, and trains the network with total effect to automatically focus on core emotional areas;
Coupling Category Alignment for Graph Domain Adaptation
Nan Yin (Hong Kong University of Science and Technology), Mengzhu Wang (Hebei University of Technology)
Domain AdaptationGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Propose a dual-branch coupled category alignment framework named CoCA for unsupervised graph domain adaptation, transferring source domain label knowledge to an unlabeled target domain.
CRAFT: Time Series Forecasting with Cross-Future Behavior Awareness
Yingwei Zhang (Chinese Academy of Sciences), Detao Lv (Alibaba Group)
Time Series
🎯 What it does: Proposes the CRAFT time series prediction framework, which leverages cross-future behavior (CFB) features to mine and predict future trends of the target sequence by utilizing trend information from CFB.
Credit Assignment and Fine-Tuning Enhanced Reinforcement Learning for Collaborative Spatial Crowdsourcing
Wei Chen (Zhengzhou University), Mingliang Xu (Zhengzhou University)
OptimizationReinforcement LearningAuto EncoderGraphTabular
🎯 What it does: Proposed the CAFE framework, which utilizes causal reward redistribution and implicit gradient meta-learning for credit assignment and parameter fine-tuning in multi-agent reinforcement learning within collaborative crowdsourcing spaces.
Critical Node-aware Augmentation for Hypergraph Contrastive Learning
Zhuo Li (Beijing University Of Technology), Gengyu Lyu (Beijing University Of Technology)
Representation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposed a hypergraph contrastive learning method based on key nodes (CNHCL), which identifies and retains key nodes through hyperedge prediction to improve the quality of hypergraph augmentation and representation learning.
Cross-modal Collaborative Representation Learning for Text-to-Image Person Retrieval
Shuanglin Yan (Nanjing University of Science and Technology), Jinhui Tang (Nanjing University of Science and Technology)
RetrievalRepresentation LearningVision Language ModelContrastive LearningMultimodality
🎯 What it does: Propose a cross-modal collaborative representation learning framework CoRL, which maps images directly to text features via a virtual text embedding synthesizer, and enhances feature representation for text-image retrieval through dual-branch collaborative learning and cross-modal relation consistency loss.
CrossVTON: Mimicking the Logic Reasoning on Cross-Category Virtual Try-On Guided by Tri-Zone Priors
Donghao Luo (Fudan University), Yanwei Fu (Fudan University)
Image TranslationGenerationData SynthesisVision Language ModelDiffusion modelImage
🎯 What it does: Propose the CrossVTON method, leveraging three-region prior (trial area, reconstruction area, imagination area) and iterative cross-category data construction to achieve high-quality synthesis for cross-category virtual try-on.
CSAHFL:Clustered Semi-Asynchronous Hierarchical Federated Learning for Dual-layer Non-IID in Heterogeneous Edge Computing Networks
Aijing Li (Beijing University of Posts and Telecommunications), Guanhua Ye (Beijing University of Posts and Telecommunications)
Federated LearningSafty and PrivacyComputational EfficiencyImageBenchmark
🎯 What it does: Proposes the CSAHFL framework to address the two-layer non-IID and client heterogeneity issues in hierarchical federated learning, enhancing model accuracy and training efficiency.
CSF-GAN: Cross-modal Semantic Fusion-based Generative Adversarial Network for Text-guided Image Inpainting
Shilin Zhang (Tianjin University), Chunjiang Fu (Dalian University of Technology)
RestorationConvolutional Neural NetworkVision Language ModelGenerative Adversarial NetworkMultimodality
🎯 What it does: Propose a one-stage text-guided image inpainting model called CSF-GAN, which completes missing regions by leveraging text semantics.
Curriculum Abductive Learning for Mitigating Reasoning Shortcuts
Wen-Da Wei (Nanjing University), Lan-Zhe Guo (Nanjing University)
Explainability and InterpretabilityComputational EfficiencyImage
🎯 What it does: Proposes the CurABL algorithm, which integrates curriculum learning into Abductive Learning, using simple samples to assist difficult samples, significantly reducing reasoning shortcuts and improving training convergence speed.
Curriculum Hierarchical Knowledge Distillation for Bias-Free Survival Prediction
Chaozhuo Li (Beijing University of Posts and Telecommunications), Xi Zhang (Beijing University of Posts and Telecommunications)
ClassificationKnowledge DistillationTransformerDiffusion modelContrastive LearningBiomedical Data
🎯 What it does: Proposed PathoKD, utilizing hierarchical knowledge distillation and three-stage curriculum learning to address the performance gap in survival prediction caused by discrepancies in the number and resolution of whole slide images (WSI).
Cyclic Vision-Language Manipulator: Towards Reliable and Fine-Grained Image Interpretation for Automated Report Generation
Yingying Fang (Imperial College London), Guang Yang (Imperial College London)
Explainability and InterpretabilityTransformerVision Language ModelDiffusion modelMultimodalityBiomedical Data
🎯 What it does: This paper proposes a cyclic vision-language manipulator (CVLM), which generates adversarial (counterfactual) images for chest X-rays through text-conditioned diffusion models, thereby explaining the visual basis of the report generator for specific medical findings.
D3: Diversity, Difficulty, and Dependability-Aware Data Selection for Sample-Efficient LLM Instruction Tuning
Jia Zhang (Nanjing University), Lan-Zhe Guo (Nanjing University)
OptimizationData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes a data selection method called D3, aiming to achieve sample-efficient LLM instruction tuning by selecting the most valuable subset from large-scale instruction tuning datasets.
DANCE: Resource-Efficient Neural Architecture Search with Data-Aware and Continuous Adaptation
Maolin Wang (City University of Hong Kong), Xiangyu Zhao
Neural Architecture SearchImage
🎯 What it does: The DANCE framework proposes a NAS method based on continuous distribution learning, achieving dynamic adaptation to different computational constraints by learning a continuously sampleable architecture distribution.
DaringFed: A Dynamic Bayesian Persuasion Pricing for Online Federated Learning Under Two-sided Incomplete Information
Yun Xin (Wuhan University of Science and Technology), Guanghui Wen
Federated LearningReinforcement LearningImage
🎯 What it does: Designed DaringFed, a dynamic Bayesian persuasion and pricing mechanism under bidirectional incomplete information, for incentive allocation in online federated learning.
DASS: A Dual-Branch Attention-based Framework for Trajectory Similarity Learning with Spatial and Semantic Fusion
Jiayi Li (Soochow University), Pengpeng Zhao (Soochow University)
RetrievalGraph Neural NetworkTransformerContrastive LearningTextGraphSequential
🎯 What it does: This paper proposes a dual-branch attention framework called DASS for learning trajectory similarity by fusing spatial and semantic information on road networks.
Data Poisoning Attack Defense and Evolutionary Domain Adaptation for Federated Medical Image Segmentation
Min Hyuk Kim (Chonnam National University), Seok Bong Yoo (Chonnam National University)
SegmentationFederated LearningAdversarial AttackBiomedical DataMagnetic Resonance Imaging
🎯 What it does: Propose the AdaShield-FL framework, integrating k-space decoupled reconstruction and segmentation, attack detection, and evolutionary domain adaptation to achieve defense and generalization in federated medical image segmentation
DcDsDiff: Dual-Conditional and Dual-Stream Diffusion Model for Generative Image Tampering Localization
Qixian Hao (Beijing University of Posts and Telecommunications), Kai Wang (Beijing University of Posts and Telecommunications)
SegmentationAnomaly DetectionConvolutional Neural NetworkTransformerDiffusion modelImage
🎯 What it does: Construct a dual-conditional dual-stream network (DcDsDiff) based on diffusion models, simultaneously generating mask maps and detail maps to achieve generative image tampering localization.
DDPA-3DVG: Vision-Language Dual-Decoupling and Progressive Alignment for 3D Visual Grounding
Hongjie Gu (Hangzhou Dianzi University), Yuxiang Yang (Hangzhou Dianzi University)
Object DetectionRepresentation LearningTransformerTextPoint Cloud
🎯 What it does: Designed and implemented the DDPA-3DVG framework, combining dual decoupling and progressive alignment to achieve 3D visual localization.
Decentralized Online Learning by Selfish Agents in Coalition Formation
Saar Cohen (Bar-Ilan University), Noa Agmon (Bar-Ilan University)
OptimizationFederated Learning
🎯 What it does: Proposed a decentralized online learning model and designed a single-sample Frank-Wolfe algorithm, enabling selfish agents to incrementally construct approximately Nash stable coalitions in additive separable exchangeable preference games.
Decision-Aware Preference Modeling for Multi-Behavior Recommendation
Qingfeng Li (Sun Yat-sen University), Jian Yin (Sun Yat-sen University)
Recommendation SystemGraph Neural NetworkContrastive Learning
🎯 What it does: This paper proposes the Decision-Aware Preference Modeling (DAPM) framework for multi-behavior recommendation, which constructs a behavior-agnostic graph to complement behavior-specific representations, and enhances the modeling of multi-behavior user preferences through decision thresholds and improved contrastive learning.
Decomposing Inconsistencies: Marginal Contributions and Pooling Techniques
Christian Straßer (Ruhr University Bochum), Said Jabbour (Université d'Artois)
Explainability and InterpretabilityComputational Efficiency
🎯 What it does: Investigated the relationship between global and local inconsistency measures, and provided a systematic framework for transitioning from global to local and vice versa
Decoupled Imbalanced Label Distribution Learning
Yongbiao Gao (Qilu University of Technology), Guohua Lv (Qilu University of Technology)
ClassificationRepresentation LearningImage
🎯 What it does: Proposes a Decoupled Imbalanced Label Distribution Learning (DILDL) method, which divides labels into dominant and non-dominant parts, learning and aligning them separately.
Decoupling and Reconstructing: A Multimodal Sentiment Analysis Framework Towards Robustness
Mingzheng Yang (University of Science and Technology of China), Min Hou (University of Science and Technology of China)
ClassificationRepresentation LearningConvolutional Neural NetworkTransformerAuto EncoderContrastive LearningMultimodality
🎯 What it does: Propose a multimodal sentiment analysis framework (DAR) based on feature decoupling and adaptive reconstruction, which can maintain robustness under missing or misaligned multimodal sequences.
Deduction with Induction: Combining Knowledge Discovery and Reasoning for Interpretable Deep Reinforcement Learning
Haodi Zhang (Shenzhen University), Fangzhen Lin (Hong Kong University of Science and Technology)
Explainability and InterpretabilityComputational EfficiencyTransformerReinforcement LearningVideo
🎯 What it does: Propose a neural-symbolic framework HRL-ID that integrates automatic rule induction with logical reasoning into hierarchical reinforcement learning, enhancing the explainability and training efficiency of DRL.
Deep Learning-Based Pedestrian Simulation with Limited Real-World Training Data: An Evaluation Framework
Vahid Mahzoon (Temple University), Slobodan Vucetic (Temple University)
GenerationData SynthesisAutonomous DrivingConvolutional Neural NetworkTransformerSupervised Fine-TuningPoint CloudSequentialBenchmark
🎯 What it does: This paper proposes a unified evaluation framework for assessing pedestrian crowd simulators based on deep learning and traditional knowledge-driven models, and systematically experiments on their performance under limited real data.
Deep Opinion-Unaware Blind Image Quality Assessment by Learning and Adapting from Multiple Annotators
Zhihua Wang (Sun Yat-sen University), Chao Huang (Sun Yat-sen University)
Domain AdaptationConvolutional Neural NetworkImage
🎯 What it does: Designed a fully human-score-free deep BIQA model called DUBMA, which first learns through pseudo labels generated by synthetic image pairs from multiple FR-IQA models, and then enhances the model's adaptability to real images using unsupervised domain adaptation.
Denoise-then-Retrieve: Text-Conditioned Video Denoising for Video Moment Retrieval
Weijia Liu (Southeast University), Ajmal Mian (University of Western Australia)
RestorationRetrievalTransformerVision Language ModelContrastive LearningVideoTextMultimodality
🎯 What it does: Propose a denoise-then-retrieve framework that first performs text-conditioned denoising on videos and then retrieves the target moment using clean multimodal representations.
Denoising Diffusion Models are Good General Gaze Feature Learners
Guanzhong Zeng (Hikvision Research Institute), Mingyang Zhou (Shenzhen University)
Pose EstimationRepresentation LearningDiffusion modelContrastive LearningImage
🎯 What it does: Through the self-supervised learning framework GazeDiff, conditional diffusion models are used to generate pre-training on unannotated facial images, thereby learning general features applicable to gaze estimation;
DenseSAM: Semantic Enhance SAM for Efficient Dense Object Segmentation
Linyun Zhou, Zunlei Feng (Zhejiang University)
SegmentationTransformerContrastive LearningImageBiomedical Data
🎯 What it does: Designed and implemented DenseSAM, replacing SAM's location prompts to achieve automatic dense object segmentation;
DepthART: Monocular Depth Estimation as Autoregressive Refinement Task
Bulat Gabdullin (AIRI), Anton Konushin (AIRI)
Depth EstimationTransformerAuto EncoderImagePoint Cloud
🎯 What it does: This paper proposes a novel autoregressive depth estimation framework called DepthART, which migrates the visual autoregressive transformer (VAR) from image generation tasks to monocular depth estimation, and improves model accuracy through a self-correcting objective during training.
DERI: Cross-Modal ECG Representation Learning with Deep ECG-Report Interaction
Jian Chen (Shenzhen MSU-BIT University), Xiping Hu (Shenzhen MSU-BIT University)
Representation LearningTransformerLarge Language ModelAuto EncoderContrastive LearningMultimodalityElectronic Health RecordsElectrocardiogram
🎯 What it does: Learn deep interactive representations across modalities of ECG and clinical reports, generating ECG representations for zero-shot classification and report generation.
Detecting Hallucination in Large Language Models Through Deep Internal Representation Analysis
Luan Zhang (Beijing Institute of Technology), Shuhao Zhang (Huazhong University of Science and Technology)
Anomaly DetectionExplainability and InterpretabilityRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: This paper proposes the MHAD (Model Hallucination Awareness for Hallucination Detection) method, which detects hallucinations by analyzing internal representations of LLMs during the generation process (including attention outputs, feed-forward network outputs, and layer outputs), and constructs the SOQHD (Sustainable Open-Domain QA Hallucination Detection) benchmark dataset, providing internal representations and hallucination labels across multiple LLMs to ensure timeliness consistency.
Device-Cloud Collaborative Correction for On-Device Recommendation
Tianyu Zhan (Zhejiang University), Fei Wu (Zhejiang University)
Recommendation SystemRecurrent Neural NetworkTextSequential
🎯 What it does: Proposed the CoCorrRec framework, deploying self-correcting networks (SCN) on devices and a global correction network (GCN) in the cloud to achieve device-cloud collaborative recommendation.
DFCA: Disentangled Feature Contrastive Learning and Augmentation for Fairer Dermatological Diagnostics
Pengcheng Zhao (Shanghai Jiao Tong University), Xiaowei Ding (Shanghai Jiao Tong University)
ClassificationConvolutional Neural NetworkFlow-based ModelContrastive LearningImageBiomedical Data
🎯 What it does: Propose the Disentangled Feature Contrastive learning and Augmentation (DFCA) framework, leveraging feature disentanglement, vMF distribution contrastive learning, and inverse flow models to enhance fairness and accuracy in dermatological diagnosis models.
DFMU: Distribution-based Framework for Modeling Aleatoric Uncertainty in Multimodal Sentiment Analysis
Chen Tang (South China University of Technology), Tong Zhang (South China University of Technology)
ClassificationTransformerContrastive LearningMultimodality
🎯 What it does: In multi-modal sentiment analysis, distributed modeling of observable uncertainty is proposed, along with the DFMU framework, which combines distributed contrastive learning and sentiment word replacement to achieve robustness.
DGCPL: Dual Graph Distillation for Concept Prerequisite Relation Learning
Miao Zhang (Hubei University), Shihui Wang (Hubei University)
Knowledge DistillationRepresentation LearningGraph Neural NetworkTransformerLarge Language ModelGraphBenchmark
🎯 What it does: Construct and train a dual graph distillation model (DGCPL) to learn concept prerequisite relationships through a concept-resource hypergraph and a learning behavior graph.
DGExplainer: Explaining Dynamic Graph Neural Networks via Relevance Back-propagation
Yezi Liu (University of California Irvine), Yanning Shen (University of California Irvine)
Explainability and InterpretabilityRecurrent Neural NetworkGraph Neural NetworkGraphTime Series
🎯 What it does: Propose DGExplainer, a method utilizing spatiotemporal hierarchical relevance propagation (LRP) to explain predictions of dynamic graph neural networks (Dynamic GNN) in link prediction and node regression tasks.
DGL: Dynamic Global-Local Information Aggregation for Scalable VRP Generalization with Self-Improvement Learning
Yubin Xiao (Jilin University), Yuan Jiang (Nanyang Technological University)
OptimizationTransformerReinforcement LearningGraphBenchmark
🎯 What it does: Propose the DGL model, which solves vehicle routing problems (VRP) by dynamically aggregating global and local information, and enhances robustness through replacement-based self-improvement learning (SIL).
DGraFormer: Dynamic Graph Learning Guided Multi-Scale Transformer for Multivariate Time Series Forecasting
Han Yan (Ocean University of China), Yanwei Yu (Ocean University of China)
Graph Neural NetworkTransformerTime Series
🎯 What it does: Propose a multi-variable time series forecasting model called DGraFormer, which combines dynamic graph learning with multi-scale Transformer.
DHTAGK: Deep Hierarchical Transitive-Aligned Graph Kernels for Graph Classification
Xinya Qin (Beijing Normal University), Edwin Hancock (University of York)
ClassificationGraph Neural NetworkAuto EncoderGraphBiomedical DataBenchmark
🎯 What it does: This paper proposes a hierarchical alignment-based deep graph kernel called DHTAGK, which generates transferable embeddings across graphs using an autoencoder.
Diff-LMM: Diffusion Teacher-Guided Spatio-Temporal Perception for Video Large Multimodal Models
Jisheng Dang (Sun Yat-sen University), Teng Wang (University of Hong Kong)
ClassificationRecognitionKnowledge DistillationVision Language ModelDiffusion modelVideoMultimodality
🎯 What it does: This paper proposes the Diff-LMM framework, which utilizes intermediate layer features from diffusion models as teacher supervision to enhance spatiotemporal fine-grained perception in video multimodal large models;
DiffECG: Diffusion Model-Powered Label-Efficient and Personalized Arrhythmia Diagnosis
Tianren Zhou (Shandong University), Zhaoyan Shen (Shandong University)
ClassificationAnomaly DetectionNeural Architecture SearchDiffusion modelMultimodalityTime SeriesBiomedical DataElectrocardiogram
🎯 What it does: Proposed a self-supervised learning framework called DiffECG based on diffusion models for efficient detection of arrhythmias and achieved personalized diagnosis
Differentiable Prompt Learning for Vision Language Models
Zhenhan Huang (Rensselaer Polytechnic Institute), Jianxi Gao (Rensselaer Polytechnic Institute)
ClassificationRecognitionKnowledge DistillationTransformerPrompt EngineeringVision Language ModelImage
🎯 What it does: Apply differentiable prompt learning on the pre-trained CLIP model, automatically determining the context length for each layer to improve the performance of downstream tasks in vision-language models.
DiffFERV: Diffusion-based Facial Editing of Real Videos
Xiangyi Chen (Shanghai Jiao Tong University), Li Song (Shanghai Jiao Tong University)
GenerationTransformerPrompt EngineeringDiffusion modelOptical FlowVideoText
🎯 What it does: DiffFERV proposes a diffusion model-based facial video editing framework that achieves precise preservation and consistent editing of facial identity, motion, and background;
DiffSQL: Leveraging Diffusion Model for Zero-Shot Self-Supervised Monocular Depth Estimation
Heyuan Zheng (Northwestern Polytechnical University), Zhiwen Yu (Northwestern Polytechnical University)
Depth EstimationConvolutional Neural NetworkDiffusion modelImage
🎯 What it does: Propose DiffSQL, a self-supervised monocular depth estimation framework that enhances features using diffusion models (Stable Diffusion) and employs an adaptive self-query layer.
Diffuse&Refine: Intrinsic Knowledge Generation and Aggregation for Incremental Object Detection
Jianzhou Wang (Hohai University), Wenhai Wang (Chinese University of Hong Kong)
Object DetectionRepresentation LearningTransformerDiffusion modelImage
🎯 What it does: Propose a framework named DiffKA based on forward and backward diffusion for intrinsic knowledge generation and aggregation, aimed at enhancing category boundaries and suppressing catastrophic forgetting in incremental object detection.
Diffusion Guided Propagation Augmentation for Popularity Prediction
Chaozhuo Li (Beijing University of Posts and Telecommunications), Xi Zhang (Beijing University of Posts and Telecommunications)
Data SynthesisGraph Neural NetworkTransformerDiffusion modelGraphTime Series
🎯 What it does: Propose a generative framework DGPA based on diffusion models to predict content popularity by simulating the propagation process during the early stage of information dissemination.
Diffusion-aware Censored Gaussian Processes for Demand Modelling
Filipe Rodrigues (Technical University of Denmark)
OptimizationDiffusion modelTabularTime Series
🎯 What it does: This paper proposes a Diffusion-aware Censored Gaussian Process (DCGP), which combines Tobit likelihood with a graph diffusion process to more accurately infer true demand from audit observations that are limited by supply constraints;
DiffusionIMU: Diffusion-Based Inertial Navigation with Iterative Motion Refinement
Xiaoqiang Teng (Beijing Technology and Business University), Xiaopeng Zhang (Chinese Academy of Sciences)
Pose EstimationConvolutional Neural NetworkDiffusion modelTime Series
🎯 What it does: Proposes DiffusionIMU, a diffusion model-based inertial navigation framework that refines motion state estimation through an iterative denoising process, directly regressing velocity from raw IMU signals and integrating it to obtain trajectories.
DIIN: Diffusion Iterative Implicit Networks for Arbitrary-scale Super-resolution
Tao Dai (Shenzhen University), Zexuan Zhu (Shenzhen University)
Super ResolutionDiffusion modelImage
🎯 What it does: Propose the Diffusion Iterative Implicit Network (DIIN) for image super-resolution at arbitrary scales.
Directing Mamba to Complex Textures: An Efficient Texture-Aware State Space Model for Image Restoration
Long Peng (University of Science and Technology of China), Zheng-Jun Zha (University of Science and Technology of China)
RestorationSuper ResolutionImage
🎯 What it does: This paper proposes a texture-aware state space model (TA-SSM) combined with multi-directional perception blocks to construct an efficient image restoration framework named TAMambaIR, applicable to tasks such as super-resolution, de-raining, and low-light enhancement.
Dirichlet Process-Based Robust Clustering Using the Median-of-Means Estimator
Supratik Basu (Duke University), Swagatam Das (Indian Statistical Institute)
TabularBiomedical Data
🎯 What it does: This paper proposes an automated clustering algorithm called DP-MoM, which combines the Dirichlet process with median-of-means estimation to robustly determine the number of clusters in noisy or outlier-contaminated data.
Disconfounding Fake News Video Explanation with Causal Inference
Lizhi Chen (Soochow University), Qiaoming Zhu (Soochow University)
Explainability and InterpretabilityTransformerLarge Language ModelVideoTextMultimodality
🎯 What it does: Propose the CIFE framework, which utilizes causal inference to eliminate common causes between video objects and explanation aspects, generating more credible fake news video explanations.
Discrete Budget Aggregation: Truthfulness and Proportionality
Ulrike Schmidt-Kraepelin (TU Eindhoven), Markus Utke (TU Eindhoven)
🎯 What it does: Studies the discrete budget aggregation problem, i.e., designing and analyzing mechanisms that satisfy truthfulness in scenarios where budgets must be allocated as integers to candidate proposals, and explores its limitations in terms of proportionality.
Disentangled and Personalized Representation Learning for Next Point-of-Interest Recommendation
Xuan Rao (University of Electronic Science and Technology of China), Peng Han (University of Electronic Science and Technology of China)
Recommendation SystemRecurrent Neural NetworkTime SeriesSequential
🎯 What it does: Propose a next-location recommendation framework called DPRL, which first decouples POI from its spatial/temporal context and separately learns sequence features, then integrates user preferences for POI and regions through a space-temporal aggregation mechanism, and finally enhances prediction using time queries.
Disentangling Multi-view Representations via Curriculum Learning with Learnable Prior
Kai Guo (Sichuan University), Hao Wang (Sichuan University)
Representation LearningMixture of ExpertsAuto EncoderImage
🎯 What it does: Propose a multi-perspective representation learning framework named CL2P, which leverages curriculum learning to first learn perspective-consistent representations and then perspective-specific representations, while enhancing representation quality through learnable priors, mixture-of-experts layers, and mutual information decoupling.
DisPIM: Distilling PreTrained Image Models for Generalizable Visuo-Motor Control
Haitao Wang (Sun Yat-sen University), Hejun Wu (Sun Yat-sen University)
Knowledge DistillationRobotic IntelligenceTransformerReinforcement LearningImage
🎯 What it does: Propose the DisPIM framework, leveraging pre-trained image models as teachers to distill features into a small learnable encoder, thereby balancing task specificity and generalization ability in visual-motor control tasks.
Distance Preservation Games
Haris Aziz (University of New South Wales), Toby Walsh (University of New South Wales)
Optimization
🎯 What it does: Propose and analyze Distance Preservation Games (DPG), studying the existence and computational complexity of jump stability (Nash equilibrium) and socially optimal location configurations.
Distilling A Universal Expert from Clustered Federated Learning
Zeqi Leng (Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education), Bo Yang (Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education)
Federated LearningSafty and PrivacyKnowledge DistillationMixture of ExpertsImage
🎯 What it does: Building upon cluster federated learning (CFL), the DisUE framework is proposed, which leverages data-free knowledge distillation to distill a unified expert model from multiple cluster expert models, using it as the initialization for the next round of training.
Distributed Cascaded Manifold Hashing Network for Compact Image Set Representation
Xiaxin Wang (Nanjing University of Science and Technology), Xia Wu (Beijing Institute of Technology)
ClassificationRetrievalComputational EfficiencyRepresentation LearningImageBenchmark
🎯 What it does: Proposed a distributed cascading manifold hashing network (DCMHN), which achieves efficient image set classification and retrieval by performing SPD manifold encoding on distributed image sets and learning binary hash codes.
Distribution-Aware Online Learning for Urban Spatiotemporal Forecasting on Streaming Data
Chengxin Wang (National University of Singapore), Beng Chin Ooi
Domain AdaptationComputational EfficiencyGraph Neural NetworkTransformerTime Series
🎯 What it does: Propose a distribution-aware online learning framework named DOL to address progressive distribution shift and location-specific shift in urban spatiotemporal flows, providing an Awake-Hibernate update strategy based on the Streaming Update Mechanism and location-specific learners within the Adaptive ST network.
Divide and Conquer: Coordinating Multiplex Mixture of Graph Learners to Handle Multi-Omics Analysis
Zhihao Wu (Zhejiang University), Haishuai Wang (Zhejiang University)
ClassificationGraph Neural NetworkMixture of ExpertsMultimodalityBiomedical Data
🎯 What it does: Propose a multi-modal graph learning framework MMoG for subtyping classification of multi-omics data;
Dividing Conflicting Items Fairly
Ayumi Igarashi (University of Tokyo), Hirotaka Yoneda (University of Tokyo)
OptimizationGraph
🎯 What it does: This paper studies fair division of indivisible items under graph constraints, proving the existence of maximum EF1 allocations for two agents in any graph, and providing polynomial (for additive) or pseudopolynomial (for monotonic) algorithms; meanwhile, it constructs a counterexample for three agents where no maximum EF1 allocation exists, and proves the NP-hardness of the decision problem for any fixed number of agents n≥3.
DM-POSA: Enhancing Open-World Test-Time Adaptation with Dual-Mode Matching and Prompt-Based Open Set Adaptation
Shiji Zhao (Nanjing University of Aeronautics and Astronautics), Songcan Chen (Nanjing University of Aeronautics and Astronautics)
Domain AdaptationAnomaly DetectionPrompt EngineeringImage
🎯 What it does: Proposed a method called DM-POSA for open-set test-time adaptation, which quickly identifies and handles unknown samples under conditions of semantic drift and covariate drift;
Do Mentioned Items Truly Matter? Enhancing Conversational Recommender Systems with Causal Intervention and Large Language Models
Lingzhi Wang (Harbin Institute of Technology Shenzhen), Kam-Fai Wong (Chinese University of Hong Kong)
Recommendation SystemGraph Neural NetworkTransformerLarge Language ModelTextGraph
🎯 What it does: Propose a hybrid framework that integrates causal intervention with large language models (LLMs) to enhance personalization and diversity in conversational recommendation systems.
Do You Steal My Model? Signature Diffusion Embedded Dual-Verification Watermarking for Protecting Intellectual Property of Hyperspectral Image Classification Models
Yufei Yang (Beijing University of Posts and Telecommunications), Jiahui Qu (Xidian University)
ClassificationSafty and PrivacyConvolutional Neural NetworkTransformerDiffusion modelImage
🎯 What it does: Proposes a dual verification model watermarking method based on sub-pixel space signature diffusion for protecting intellectual property rights of hyperspectral image classification models.
DO-CoLM: Dynamic 3D Conformation Relationships Capture with Self-Adaptive Ordering Molecular Relational Modeling in Language Models
Zhuo Chen (University of Science and Technology of China), Yang Wang (University of Science and Technology of China)
Drug DiscoveryGraph Neural NetworkTransformerLarge Language ModelPrompt EngineeringMultimodalityGraphBiomedical Data
🎯 What it does: Propose a multimodal large language model, DO-CoLM, which integrates 3D molecular conformation information and adaptively adjusts input order for molecular relation learning.
DONIS: Importance Sampling for Training Physics-Informed DeepONet
Shudong Huang (Sichuan University), Jiancheng Lv (Sichuan University)
Computational EfficiencyPhysics Related
🎯 What it does: Propose a two-step importance sampling framework DONIS to accelerate the training of physics-informed DeepONet with unlabeled data.
DPMamba: Distillation Prompt Mamba for Multimodal Remote Sensing Image Classification with Missing Modalities
Yueguang Yang (Xidian University), Wenqian Dong (Xidian University)
ClassificationKnowledge DistillationData-Centric LearningPrompt EngineeringContrastive LearningImageMultimodalityPoint Cloud
🎯 What it does: Proposes the DPMamba framework, achieving remote sensing image classification under missing modalities through knowledge distillation and learnable multimodal prompts (MMAP).
Drafting and Revision: Advancing High-Fidelity Video Inpainting
Zhiliang Wu (Zhejiang University), Yi Yang (Zhejiang University)
RestorationConvolutional Neural NetworkGenerative Adversarial NetworkOptical FlowVideo
🎯 What it does: Proposed a Drafting-and-Revision Completion Network (DRCN), which decomposes video frames into low-frequency semantic structures and high-frequency details using a Laplacian pyramid, completes the draft at low resolution, refines details at high resolution, and finally synthesizes the complete video.
DriftRemover: Hybrid Energy Optimizations for Anomaly Images Synthesis and Segmentation
Siyue Yao (Xi'an Jiaotong-Liverpool University), Eng Gee Lim (Xi'an Jiaotong-Liverpool University)
SegmentationData SynthesisAnomaly DetectionTransformerDiffusion modelImage
🎯 What it does: By leveraging rough masks and anomaly class information, using Stable Diffusion combined with attention mapping to automatically generate anomaly images and their precise mask labels, thus alleviating the problem of scarce real anomaly samples.
DToMA: Training-free Dynamic Token MAnipulation for Long Video Understanding
Bowen Yuan (Nanjing University of Posts and Telecommunications), Bing-Kun Bao (Nanjing University of Posts and Telecommunications)
RecognitionComputational EfficiencyTransformerVideoMultimodalityBenchmark
🎯 What it does: The paper proposes a training-free dynamic token manipulation method called DToMA, aiming to improve efficiency and understanding in long video comprehension tasks.
Dual Encoder Contrastive Learning with Augmented Views for Graph Anomaly Detection
Nannan Wu (Tianjin University), Yiming Zhao (Tianjin University)
Anomaly DetectionGraph Neural NetworkTransformerContrastive LearningGraph
🎯 What it does: Propose the DECLARE framework, integrating dual encoders (GCN + Graph Transformer), multi-view contrastive learning, structural and attribute reconstruction, and anomaly scoring to achieve graph anomaly detection.
Dual Robust Unbiased Multi-View Clustering for Incomplete and Unpaired Information
Liang Zhao (Dalian University of Technology), Bo Xu (Dalian University of Technology)
Representation LearningContrastive LearningImageTextMultimodality
🎯 What it does: Propose a deep multi-view clustering method called DRUMVC, which can simultaneously address partial viewpoint alignment (PVP) and sample sparsity (PSP) problems.
Dual-Agent Reinforcement Learning for Automated Feature Generation
Wanfu Gao (Jilin University), Kunpeng Liu (Portland State University)
Data-Centric LearningTransformerReinforcement LearningTabular
🎯 What it does: Propose a dual-agent reinforcement learning framework, DARL, for automatically generating and selecting features, while enhancing state representation through self-attention.
Dual-Balancing for Physics-Informed Neural Networks
Chenhong Zhou (Hong Kong Baptist University), Ching Eng Png (Agency for Science, Technology and Research)
Physics Related
🎯 What it does: Propose a dual-balanced PINN (DB-PINN) that dynamically adjusts loss weights to address training imbalance between PDE residual loss and boundary/initial condition loss, as well as between different conditions.
Dual-level Fuzzy Learning with Patch Guidance for Image Ordinal Regression
Chunlai Dong (Zhejiang University), Jian Wu (University Of Notre Dame)
ClassificationTransformerImageBiomedical Data
🎯 What it does: This work proposes a dual-layer fuzzy learning and patch-guided framework (DFPG), which achieves precise image ordinal regression grading by leveraging patch-level pseudo labels and fuzzy logic;
Dual-Perspective United Transformer for Object Segmentation in Optical Remote Sensing Images
Yanguang Sun (Nanjing University of Science and Technology), Lei Luo (Nanjing University of Science and Technology)
SegmentationTransformerImage
🎯 What it does: This paper proposes a Dual-Perspective Unified Transformer (DPU-Former) for target segmentation in optical remote sensing images, and presents a complete encoder-decoder architecture.
DualCast: A Model to Disentangle Aperiodic Events from Traffic Series
Xinyu Su (University of Melbourne), Jianzhong Qi (University of Melbourne)
Graph Neural NetworkTransformerTime Series
🎯 What it does: This paper proposes the DualCast framework, which decomposes traffic sequences into periodic and environmental (non-periodic) signals using a dual-branch structure, thereby improving prediction accuracy.
DUQ: Dual Uncertainty Quantification for Text-Video Retrieval
Xin Liu (Southwestern University of Finance and Economics), Yee-Hong Yang (University of Alberta)
RetrievalTransformerVision Language ModelMultimodality
🎯 What it does: Studied the problem of uncertainty in text-video retrieval, proposing the Dual Uncertainty Quantification (DUQ) framework to improve retrieval performance.
Dyn-D^2P: Dynamic Differentially Private Decentralized Learning with Provable Utility Guarantee
Zehan Zhu (Zhejiang University), Jinming Xu (Zhejiang University)
Federated LearningSafty and PrivacyConvolutional Neural NetworkImage
🎯 What it does: Proposes Dyn-DP, a differential privacy decentralized learning algorithm that dynamically adjusts gradient clipping and noise levels in time-varying directed networks, balancing privacy and accuracy.
Dynamic Anchor-based Ensemble Clustering via Hypergraph Reconstruction
Jiaxuan Xu (Sichuan University), Liang Du (Shanxi University)
Computational EfficiencyImageTabular
🎯 What it does: This paper proposes an ensemble clustering method called YACHT, which combines dynamic anchor learning with hypergraph reconstruction to improve clustering accuracy without relying on original features.
Dynamic and Adaptive Feature Generation with LLM
Xinhao Zhang (Portland State University), Kunpeng Liu (Portland State University)
ClassificationTransformerLarge Language ModelTabularChain-of-Thought
🎯 What it does: Proposed a dynamic adaptive feature generation framework called LFG based on large language models (LLMs), which leverages an LLM agent to generate, evaluate, and optimize feature sets through multi-round iterations.
Dynamic Higher-Order Relations and Event-Driven Temporal Modeling for Stock Price Forecasting
Kijeong Park (Gwanak Lab Co., Ltd.), Jong-June Jeon (University of Seoul)
Graph Neural NetworkTransformerTime SeriesFinance Related
🎯 What it does: Proposed a stock price prediction model that integrates the Hawkes process with dynamic hypergraphs
Dynamic Multiple High-order Correlations Fusion with Noise Filtering for Incomplete Multi-view Noisy-label Learning
Kaixiang Wang (Nanjing University of Finance and Economics), Fan Yang (Nanjing University of Finance and Economics)
Representation LearningGraph Neural NetworkImageMultimodality
🎯 What it does: Propose a dynamic multi-hypergraph neural network and a noise filtering matrix to address the challenges of view missingness and label noise in incomplete multi-view multi-label learning.
Dynamic Network Discovery via Infection Tracing
Ben Bals (Centrum Wiskunde & Informatica), George Skretas (Hasso Plattner Institute)
OptimizationExplainability and InterpretabilityGraph
🎯 What it does: Proposed and studied a model for recovering temporal graphs from infection trajectories, and designed the DiscoveryFollow algorithm based on δ-edge-connected components;
Dynamic Replanning for Improved Public Transport Routing
Abdallah Abuaisha (Monash University), Mark Wallace (Monash University)
OptimizationComputational EfficiencyTabular
🎯 What it does: This paper proposes a dynamic replanning method for real-time response to delays in public transportation route planning;
Dynamic Seed-GrowthCM: A Dynamic Benefit-Oriented Algorithm for Core Maximization on Large Graphs
Dongyuan Ma (Tianjin University), Xin Huang (Hong Kong Baptist University)
OptimizationGraph
🎯 What it does: Propose a core maximization algorithm called Dynamic Seed-GrowthCM, which dynamically estimates benefits based on λ-shell components and inserts edges in graphs using complete growth and partial growth strategies to maximize the k-core size.