AAAI 2026 Papers — Page 11
AAAI Conference on Artificial Intelligence · 4149 papers
DOS: Directional Object Separation in Text Embeddings for Multi-Object Image Generation
Dongnam Byun (Seoul National University), Wonjong Rhee (Seoul National University)
GenerationPrompt EngineeringVision Language ModelDiffusion modelImageTextBenchmark
🎯 What it does: Propose the DOS method, which improves the success rate of multi-object image generation by modifying CLIP text embeddings in text generation models.
DOS: Distilling Observable Softmaps of Zipfian Prototypes for Self-Supervised Point Representation
Mohamed Abdelsamad (Bosch Center for Artificial Intelligence), Abhinav Valada (Bosch Center for Artificial Intelligence)
Autonomous DrivingKnowledge DistillationRepresentation LearningTransformerPoint Cloud
🎯 What it does: Propose DOS, a self-supervised 3D point cloud learning framework based on observable softmaps, which achieves high-quality semantic features through teacher-student structure with soft mapping distillation on visible points.
Double Rounding: Nearly Lossless Adaptive Bit Switching for QAT
Haiduo Huang (Xi'an Jiaotong University), Pengju Ren (Xi'an Jiaotong University)
ClassificationObject DetectionSegmentationConvolutional Neural NetworkImageText
🎯 What it does: Proposed the Double Rounding quantization method, combined with Adaptive Learning Rate Scaling (AdaScale) and Hessian-based Mixed Precision Search (HessBit), to achieve joint training of multi-precision and mixed-precision models in one go, enabling nearly lossless bit-width switching.
Doubly Debiased Test-Time Prompt Tuning for Vision-Language Models
Fei Song (National Key Laboratory of Space Integrated Information System, Institute of Software, Chinese Academy of Sciences), Jiangmeng Li (National Key Laboratory of Space Integrated Information System, Institute of Software, Chinese Academy of Sciences)
ClassificationKnowledge DistillationTransformerPrompt EngineeringVision Language ModelImageMultimodalityRetrieval-Augmented Generation
🎯 What it does: Proposes a dual debiasing test-time prompt tuning method called D2TPT, aiming to enhance the generalization ability of vision-language models on unlabeled test data by mitigating prompt optimization bias through dynamic retrieval and reliability-aware mechanisms.
DP-GenG: Differentially Private Dataset Distillation Guided by DP-Generated Data
Shuo Shi (Zhejiang University), Tianyu Du (Hangzhou Dianzi University)
Safty and PrivacyKnowledge DistillationImage
🎯 What it does: Propose a dataset distillation framework DP-GENG that leverages differentially private generated data to guide the process.
DP-NCB: Privacy Preserving Fair Bandits
Dhruv Sarkar (Indian Institute of Technology Kharagpur), Sayak Ray Chowdhury (Indian Institute of Technology Kanpur)
Safty and PrivacyReinforcement Learning
🎯 What it does: Proposed a multi-armed bandit algorithm DP-NCB that simultaneously satisfies differential privacy (DP) and fairness (measured by Nash regret), providing theoretical guarantees and experimental validation under both global and local DP models.
DPRM: A Dual Implicit Process Reward Model in Multi-Hop Question Answering
Xinyi Wang (National University of Defense Technology), Minlie Huang (National University of Defense Technology)
OptimizationReinforcement Learning from Human FeedbackTransformerLarge Language ModelTextGraphRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: This paper proposes a Dual Implicit Process Reward Model (DPRM) to simultaneously evaluate chain-of-thought (CoT) and knowledge graph (KG) paths in multi-hop question answering tasks, achieving collaborative reasoning through mutual evaluation between the two.
DQT: Dynamic Quantization Training via Dequantization-Free Nested Integer Arithmetic
Hazem Hesham Yousef Shalby (Politecnico di Milano), Manuel Roveri (EssilorLuxottica)
ClassificationComputational EfficiencyConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a dynamic quantization training framework called DQT, which utilizes nested integer representation to achieve dynamic mixed-precision inference without dequantize.
DR.Experts: Differential Refinement of Distortion-Aware Experts for Blind Image Quality Assessment
Bohan Fu (Beijing Institute of Technology), Runze Hu (Beijing Institute of Technology)
TransformerMixture of ExpertsVision Language ModelImage
🎯 What it does: Proposes DR.Experts, a blind image quality assessment framework that leverages distortion priors extracted by DA-CLIP, refines distortion features through the Distortion-Saliency Differential Module, and fuses them according to distortion importance in the Dynamic Distortion Weighting Module.
DRAFT-RL: Multi-Agent Chain-of-Draft Reasoning for Reinforcement Learning-Enhanced LLMs
Yuanhao Li, Wei Tan (Beijing University Of Posts And Telecommunications)
Reinforcement Learning from Human FeedbackReinforcement LearningAgentic AITextChain-of-Thought
🎯 What it does: Propose the DRAFT-RL framework, integrating multi-agent reinforcement learning with Chain-of-Draft (CoD) reasoning, enabling each agent to generate multiple concise reasoning drafts and selecting/learning through peer evaluation and reward models.
DragNeXt: Rethinking Drag-Based Image Editing
Yuan Zhou (Nanyang Technological University), Hanwang Zhang (Hefei University of Technology)
Diffusion modelImageBenchmark
🎯 What it does: Proposed a region-level geometric transformation-based Drag-Based Image Editing (DBIE) framework named DragNeXt, utilizing Latent Space Optimization (LRO) and Progressive Backward Self-Intervention (PBSI) to achieve image drag editing.
Dream-IF: Dynamic Relative EnhAnceMent for Image Fusion
Xingxin Xu (Tianjin University), Pengfei Zhu (National University of Defense Technology)
RestorationObject DetectionTransformerPrompt EngineeringImageMultimodality
🎯 What it does: Propose Dream-IF by jointly modeling multi-modal image fusion and enhancement, leveraging dynamic relative enhancement to improve fusion quality.
DreamRunner: Fine-Grained Compositional Story-to-Video Generation with Retrieval-Augmented Motion Adaptation
Zun Wang (University Of North Carolina Chapel Hill), Mohit Bansal (University Of North Carolina Chapel Hill)
GenerationTransformerLarge Language ModelDiffusion modelVideoTextRetrieval-Augmented Generation
🎯 What it does: This paper proposes a story-to-video generation framework called DREAMRUNNER, which integrates LLM hierarchical planning, retrieval-enhanced motion adaptation, and spatiotemporal region 3D attention. It enables fine-grained control over multiple objects and events within a single scene and generates coherent multi-scene videos.
DRFGD: Disentangled Representation-Focused Generative Defense for Attack-Tolerant Cross-Modal Hashing
Zhongqing Yu (Huaqiao University), Pan Zhou (Hong Kong Baptist University)
RetrievalAdversarial AttackConvolutional Neural NetworkGenerative Adversarial NetworkMultimodality
🎯 What it does: Propose a generative defense framework named DRFGD based on disentangled representation to enhance the robustness of cross-modal hashing against adversarial attacks;
Drift Estimation for Diffusion Processes Using Neural Networks Based on Discretely Observed Independent Paths
Yuzhen Zhao (Sorbonne Universite), Marc Hoffmann (Sorbonne Universite)
Time SeriesStochastic Differential Equation
🎯 What it does: Nonparametric estimation of the drift function in time-homogeneous diffusion processes using deep ReLU neural networks, with non-asymptotic convergence rates provided.
Drift-aware Collaborative Assistance Mixture of Experts for Heterogeneous Multistream Learning
En Yu (Australian Artificial Intelligence Institute), Guangquan Zhang (Australian Artificial Intelligence Institute)
Domain AdaptationTransformerMixture of ExpertsTabularTime Series
🎯 What it does: This study proposes a dynamic collaborative mixture of experts framework called CAMEL, designed for efficient learning and prediction in real-time, evolving multi-stream environments.
DRIFT: Difference-Aware Reinforcement Through Iterative Fine-Tuning for Language Model
Wenjie Liao (OPPO Mobile Telecommunications Corp), Haonan Lu (OPPO Mobile Telecommunications Corp)
GenerationTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningText
🎯 What it does: Propose DRIFT, a self-play fine-tuning framework that uses differential training to enhance the generation quality of large language models.
Drifting Away from Truth: GenAI-Driven News Diversity Challenges LVLM-Based Misinformation Detection
Fanxiao Li, Wei Zhou (Yunnan University)
Data SynthesisAnomaly DetectionLarge Language ModelVision Language ModelDiffusion modelImageTextMultimodalityBenchmark
🎯 What it does: Studied the multi-level drift caused by the diversity of news generated by generative AI, and evaluated the robustness of large vision-language models in detecting multimodal misinformation, constructing a benchmark named DRIFTBENCH.
Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning
Yue Li (University of Science and Technology of China), Xinhai Zhao (Huawei Noah's Ark Lab)
Autonomous DrivingOptimizationTransformerSupervised Fine-TuningReinforcement LearningVision Language ModelImageTextSequentialChain-of-Thought
🎯 What it does: Propose the Drive-R1 model in autonomous driving scenarios, integrating vision-language models with planning, achieving interpretable trajectory prediction from visual inputs through supervised learning and reinforcement learning.
DriveFlow: Rectified Flow Adaptation for Robust 3D Object Detection in Autonomous Driving
Hongbin Lin (Chinese University of Hong Kong), Zhen Li (Chinese University of Hong Kong)
Autonomous DrivingTransformerDiffusion modelFlow-based ModelRectified FlowImageVideoOrdinary Differential Equation
🎯 What it does: Proposes DriveFlow, a frequency decomposition method based on a pre-trained text-to-image flow model, to enhance the robustness of visual 3D object detection during training.
DriveLiDAR4D: Sequential and Controllable LiDAR Scene Generation for Autonomous Driving
Kaiwen Cai (Li Auto Inc), Xianpeng Lang (Li Auto Inc)
Autonomous DrivingConvolutional Neural NetworkTransformerDiffusion modelMultimodalityPoint CloudSequential
🎯 What it does: Designed DriveLiDAR4D, a 4D LiDAR scene generation pipeline that leverages multimodal conditions (road sketches, scene descriptions, object priors) and an isometric projection spatiotemporal diffusion model LiDAR4DNet, enabling controllable generation of foreground, background, and object details, as well as temporally consistent sequences.
DriveSuprim: Towards Precise Trajectory Selection for End-to-End Planning
Wenhao Yao (Fudan University), Zuxuan Wu (NVIDIA)
Autonomous DrivingOptimizationKnowledge DistillationTransformer
🎯 What it does: Propose the DriveSuprim method, which improves the trajectory selection accuracy of end-to-end planning through coarse-to-fine filtering, rotation data augmentation, and self-distillation.
Driving with Advice: Large Model as Motion Advisor for Joint Planning
Junyin Wang (Wuhan University of Technology), Shengwu Xiong (VOYAH Automobile Technology Co., Ltd.)
Autonomous DrivingOptimizationExplainability and InterpretabilityComputational EfficiencyKnowledge DistillationRepresentation LearningData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningVision Language ModelDiffusion modelImageMultimodalityPoint Cloud
🎯 What it does: Propose a closed-loop framework, Driving with Advice, which integrates a vision-language model (VLM) as a motion advisor to achieve collaboration between high-level semantic reasoning and low-level trajectory planning in end-to-end autonomous driving; transfer large model driving reasoning to small models through semantic intent pretraining (SIP); implement interpretable decision-making via a discrete action space (direction × speed); and employ group relative policy optimization (GRPO) for multi-objective reinforcement learning on policy and diffusion-based trajectory generation.
DRMD: Deep Reinforcement Learning for Malware Detection Under Concept Drift
Shae McFadden (King's College London), Fabio Pierazzi (Alan Turing Institute)
Anomaly DetectionReinforcement LearningTabular
🎯 What it does: This paper reformulates the Android malware detection problem as a first-order Markov Decision Process (MD-MDP) and trains a DRMD agent using deep reinforcement learning (PPO), integrating classification, active learning, and rejection decisions to achieve adaptive detection under concept drift.
Dropout Prompt Learning: Towards Robust and Adaptive Vision-Language Models
Biao Chen (University of Electronic Science and Technology of China), Yuchen Wang (University of Electronic Science and Technology of China)
ClassificationRepresentation LearningTransformerPrompt EngineeringVision Language ModelContrastive LearningMultimodalityBenchmark
🎯 What it does: This paper proposes the Dropout Prompt Learning (DroPLe) framework, applying Dropout at the token level in vision-language models (VLMs), designing importance-weighted token dropout (IWTD) and residual entropy regularization to enhance the model's robustness and generalization ability under low-data, long-tail, and out-of-distribution scenarios.
DRSoRec: Dual-Rectification of Social Networks for Recommendation
Liangxun Yang (Nanjing University of Aeronautics and Astronautics), Yicong Li (Nanjing University of Aeronautics and Astronautics)
Recommendation SystemGraph Neural NetworkMixture of ExpertsContrastive LearningGraph
🎯 What it does: This paper proposes the DRSoRec dual correction framework, which denoises and completes the original social network through two parallel branches: invariant social rationality discovery and adaptive social connection refinement. The corrected social information is then fused with collaborative information generated by LightGCN to enhance social recommendation performance.
DS-ATGO: Dual-Stage Synergistic Learning via Forward Adaptive Threshold and Backward Gradient Optimization for Spiking Neural Networks
Jiaqiang Jiang (Zhejiang University of Technology), Rui Yan (Zhejiang University of Technology)
ClassificationSpiking Neural NetworkImageVideo
🎯 What it does: Introducing dual-phase collaborative learning in direct SNN training: forward adaptive threshold (AT) and backward gradient optimization (TGO)
DS-ProGen: A Dual-Structure Deep Language Model for Functional Protein Design
Yanting Li (Chinese University of Hong Kong), Yu Li (Chinese University of Hong Kong)
Protein Structure PredictionGraph Neural NetworkTransformerLarge Language ModelBiomedical DataBenchmark
🎯 What it does: Propose DS-ProGen, a dual-modal deep language model that integrates protein backbone geometry and surface chemical features for inverse protein folding and functional protein design.
DSAP: Enhancing Generalization in Goal-Conditioned Reinforcement Learning
Yiming Wang (University of Macau), Leong Hou U (University of Macau)
Reinforcement LearningGraph
🎯 What it does: Propose the DSAP (Deconfounded State Abstraction for Policy learning) framework in goal-conditioned reinforcement learning, which learns a causal graph as a proxy for confounding variables, employs backdoor adjustment to achieve state abstraction and policy learning, thereby enhancing generalization capabilities for new goals and environments.
DSCF: Dual-Source Counterfactual Fusion for High-Dimensional Combinatorial Interventions
Jitong Dou (Beijing Institute of Technology), Yurong Cheng (Beijing Institute of Technology)
Mixture of ExpertsSequential
🎯 What it does: Proposes Dual-Source Counterfactual Fusion (DSCF), achieving unbiased counterfactual prediction in high-dimensional composite intervention scenarios by jointly learning observed data and proxy counterfactual samples obtained through matching.
DSCodeBench: A Realistic Benchmark for Data Science Code Generation
Shuyin Ouyang (King's College London), Jie M. Zhang (King's College London)
Large Language ModelTextBenchmark
🎯 What it does: This paper proposes DSCodeBench—a code generation benchmark based on real GitHub data science projects, containing 1,000 tasks involving 10 major Python data science libraries;
DSFedMed: Dual-Scale Federated Medical Image Segmentation via Mutual Distillation Between Foundation and Lightweight Models
Hanwen Zhang (Peking University), Guibo Luo (Peking University)
SegmentationData SynthesisFederated LearningKnowledge DistillationTransformerDiffusion modelBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: Propose DSFedMed, a dual-scale federated learning framework that collaborates to train a server-side large foundational model and client-side lightweight model through mutual knowledge distillation, achieving high accuracy in medical image segmentation with low communication/inference costs.
DSP-PCQA: Integrating Multiple Perception Preferences for Point Cloud Quality Assessment
Mingxuan Li (Beijing Institute of Technology), Xiaohui Chu (Beijing Institute of Technology)
Convolutional Neural NetworkTransformerSupervised Fine-TuningPoint Cloud
🎯 What it does: Propose a dual-stream perception point cloud quality assessment framework, DSP-PCQA, which simulates two human cognitive channels: technical rationality and semantic perception, addressing the issues of distortion and semantic conflicts caused by traditional single-view assessment methods.
DTTNet: Improving Video Shadow Detection via Dark-Aware Guidance and Tokenized Temporal Modeling
Zhicheng Li (China University of Mining and Technology), Yong Zhou (China University of Mining and Technology)
SegmentationTransformerVision Language ModelContrastive LearningVideoMultimodality
🎯 What it does: Propose DTTNet, combining dark text prior and tokenized temporal modeling to achieve video shadow detection.
Dual Coding Theory in Action: Language-Assisted Human Pose Estimation in Videos
Sifan Wu (Jilin University), Yingying Jiao (Zhejiang University of Technology)
Pose EstimationTransformerLarge Language ModelVision Language ModelContrastive LearningVideoTextMultimodality
🎯 What it does: Proposed a video human pose estimation framework LAPose based on dual-encoding theory, which enhances the robustness of pose estimation by generating video text descriptions through multimodal large language models and achieving coarse-to-fine hierarchical alignment between visual and textual modalities.
Dual Graph Disambiguation for Multi-Instance Partial-Label Learning
Zhen Zhu (Beijing Jiaotong University), Yining Sun (Yixiaomo.Inc)
ClassificationGraph Neural NetworkImageBiomedical DataBenchmarkAudio
🎯 What it does: This paper proposes the DualG framework, which simultaneously constructs package-level and instance-level dual graphs in multi-instance partial label learning to achieve joint optimization of feature learning and label disambiguation.
Dual Mamba for Node-Specific Representation Learning: Tackling Over-Smoothing with Selective State Space Modeling
Xin He (Jilin University), Xin Wang (Jilin University)
ClassificationRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: Proposes Dual Mamba-enhanced Graph Convolutional Network (DMbaGCN), which jointly models hierarchical evolution of node representations and global dependencies through two modules: Local State-Evolution Mamba (LSEMba) and Global Context-Aware Mamba (GCAMba), mitigating over-smoothing in deep GNNs.
Dual-Branch Asymmetric Discrepancy Learning Based on Fake Image Pattern-Coexistence for AI-Generated Image Detection
Chunli Song (Beijing University of Posts and Telecommunications), Shuwu Zhang (Beijing University of Posts and Telecommunications)
Anomaly DetectionConvolutional Neural NetworkDiffusion modelGenerative Adversarial NetworkImageBenchmark
🎯 What it does: This paper proposes a dual-branch asymmetric difference learning framework (DADL), which improves the detection performance of AI-generated images by amplifying inconsistencies within forged images.
Dual-Channel Learning Framework for Zero-Shot CircRNA-miRNA Interaction Prediction via State Space Modeling
Mengmeng Wei (China University of Mining and Technology), Haicheng Yi (City University of Hong Kong (Dongguan))
Drug DiscoveryGraph Neural NetworkTransformerGraphBiomedical Data
🎯 What it does: Proposed a dual-channel learning framework called ZeroStem based on state-space modeling for zero-shot CircRNA-miRNA interaction prediction.
Dual-Geometry Graph Network: Unifying Local and Global Priors for Few-Shot Learning
Zheng Han (University of Science and Technology Beijing), Xu-Cheng Yin
Meta LearningGraph Neural NetworkImageBiomedical Data
🎯 What it does: Propose the Dual-Geometry Graph Network (DGGN), which integrates local Ollivier-Ricci curvature and global resistance embedding to jointly construct task graphs and realize geometric priors within graph neural networks;
Dual-Horizon Interest Model for Unified Search and Recommendation
Wenhao Zhu (Wuhan University), Hao Wang (Wuhan University)
Recommendation SystemMixture of ExpertsContrastive LearningText
🎯 What it does: Propose the DHIM model, unifying search and recommendation, explicitly modeling and cross-scenario fusing user long-term and short-term interests.
Dual-Kernel Graph Community Contrastive Learning
Xiang Chen (Yunnan University), Liang Duan (Tianjin University of Science and Technology)
Computational EfficiencyKnowledge DistillationRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Propose a dual-kernel graph community contrastive learning framework that leverages community structure and multi-kernel learning, simultaneously incorporating node-level and community-level information, while achieving efficient inference through decoupled GNN and knowledge distillation.
Dual-Path Knowledge-Augmented Contrastive Alignment Network for Spatially Resolved Transcriptomics
Wei Zhang (City University of Hong Kong), Xinyue Li (City University of Hong Kong)
Knowledge DistillationRepresentation LearningLarge Language ModelContrastive LearningMultimodalityBiomedical Data
🎯 What it does: Propose DKAN dual-path knowledge-enhanced contrastive alignment network for predicting spatial gene expression using H&E sections;
Dual-Perspective Disentanglement: Learning Symmetric Group-Aware Representations for Cross-Domain Recommendation
Borui Wu (Jilin University), Yuanbo Xu (Jilin University)
Recommendation SystemGraph Neural NetworkAuto EncoderGraph
🎯 What it does: Proposes a dual-perspective decomposition cross-domain recommendation framework called DPGCDR, which dynamically clusters users and items into groups/topics and then performs symmetric decomposition on specific and shared factors before fusing them.
Dual-Phase Visual-Language Pretraining and Adaptation for Long-Tailed Multi-Label Recognition
Yongcheng Li (Tongji University), Cairong Zhao (Tongji University)
RecognitionTransformerLarge Language ModelVision Language ModelContrastive LearningMultimodality
🎯 What it does: Propose a two-stage vision-language pre-training and adaptation framework DP-VLPA, which generates rich descriptions for tail classes using large language models (LLMs) and improves long-tailed multi-label recognition performance in the second stage through dynamic query reweighting and co-occurrence aware loss.
Dual-Seed Evolutionary Algorithm for Noise Optimization in Diffusion Models
Yuzheng Tan (Xidian University), Guangneng Hu (Xidian University)
GenerationOptimizationDiffusion modelImageBenchmark
🎯 What it does: Propose an evolution-based seed optimization framework (SOE) that globally searches and locally refines the initial noise of text-to-image diffusion models without training, enhancing the semantic consistency and visual quality of generated images.
Dual-stream Relation-modeling Disentanglement for Cloth-Changing Person Re-Identification
Shijuan Huang, Zhao Lv (Huazhong University Of Science And Technology)
RecognitionRetrievalTransformerVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Propose the DRDnet framework, utilizing dual-stream decomposition and relationship modeling to achieve person re-identification under clothing changes, overcoming the issues of traditional auxiliary modalities being overly dependent on single modes and label inconsistencies.
Dual-Teacher Interactive Knowledge Distillation Network for Text-to-Visible & Infrared Person Retrieval
Chenglong Li (Anhui University), Aihua Zheng (Anhui University)
RetrievalKnowledge DistillationTransformerVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Propose a dual-teacher interactive knowledge distillation network (DIKDNet), which enhances the accuracy and robustness of visible light (RGB) and infrared (TIR) person re-identification under text retrieval by fusing RGB and TIR images with text embeddings.
Dual-View Inference Attack: Machine Unlearning Amplifies Privacy Exposure
Lulu Xue (Huazhong University of Science and Technology), Leo Yu Zhang (Griffith University)
Safty and PrivacyAdversarial AttackConvolutional Neural NetworkImage
🎯 What it does: Proposes a dual-view inference attack (Dual-View Inference Attack, DVIA), which infers the membership status of retained data by leveraging output differences between pre- and post-unlearning models, thereby revealing privacy leakage risks caused by machine unlearning.
DualFete: Revisiting Teacher-Student Interactions from a Feedback Perspective for Semi-supervised Medical Image Segmentation
Le Yi (Sichuan University), Zizhou Wang (Sichuan University)
SegmentationMeta LearningBiomedical Data
🎯 What it does: Proposes a feedback-based teacher-student framework to correct error propagation in semi-supervised medical image segmentation.
DualScope: Capturing Critical Spatial and Temporal Cues for Distracted Driving Activity Recognition
Zhijie Qiu (Hong Kong University of Science and Technology), Wei Ma (Beijing University of Technology)
ClassificationRecognitionAutonomous DrivingKnowledge DistillationTransformerVideo
🎯 What it does: Propose the DualScope framework, which simultaneously focuses on spatial key regions and temporal dynamic features for distracted driving behavior recognition.
DualSpeechLM: Towards Unified Speech Understanding and Generation via Dual Speech Token Modeling with Large Language Models
Yuanyuan Wang (Chinese University of Hong Kong), Xixin Wu (Chinese University of Hong Kong)
RecognitionGenerationTransformerLarge Language ModelSupervised Fine-TuningAuto EncoderTextAudio
🎯 What it does: This study proposes the DualSpeechLM model, achieving the unification of speech understanding and generation by bridging the modal gap between text and speech through a Understanding-Driven Speech Tokenizer (USTokenizer) and a dual token modeling framework.
DuGI-MAE: Improving Infrared Mask Autoencoders via Dual-Domain Guidance
Yinghui Xing (Northwestern Polytechnical University), Di Xu (Huawei)
Object DetectionSegmentationRepresentation LearningTransformerAuto EncoderImage
🎯 What it does: Proposes DuGI-MAE, a self-supervised pre-training model for infrared images, combining entropy-based deterministic masking and dual-domain guidance modules to enhance representation learning for infrared images.
DuoKD: Dual Knowledge Distillation from Large Language Models for Robust Graph Neural Networks
Cuiying Huo (Tianjin University), Di Jin (Tianjin University)
Knowledge DistillationGraph Neural NetworkLarge Language ModelPrompt EngineeringContrastive LearningTextGraph
🎯 What it does: This paper proposes DuoKD, a bidirectional knowledge distillation framework that extracts semantically similar positive knowledge and semantically dissimilar negative knowledge from large language models (LLMs) simultaneously, converting them into supervision signals for the opinion layer and rationality layer, supplemented by structured rationality-guided information passing in GNNs.
DUP: Detection-guided Unlearning for Backdoor Purification in Language Models
Man Hu (Beijing Electronic Science and Technology Institute), Shuai Zhao (Nanyang Technological University)
Safty and PrivacyKnowledge DistillationAdversarial AttackTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposed the DUP framework, combining feature-level backdoor detection with parameter-efficient unlearning (LoRA + knowledge distillation) for backdoor purification in language models.
Duplex Rewards Optimization for Test-Time Composed Image Retrieval
Haoliang Zhou (Tianjin University Of Technology), Changsheng Xu (Chinese Academy Of Sciences)
RetrievalTransformerReinforcement LearningVision Language ModelContrastive LearningMultimodality
🎯 What it does: This paper proposes the Test-Time Composed Image Retrieval (TT-CIR) framework and designs the TT-RLDR (Test-Time Reinforcement Learning with Duplex Rewards) method, which leverages reinforcement learning during testing to adapt pre-trained vision-language models and enhance compositional image retrieval performance.
DW-DGAT: Dynamically Weighted Dual Graph Attention Network for Neurodegenerative Disease Diagnosis
Chengjia Liang (Shenzhen University), Zhongwei Huang (Hubei University of Technology)
ClassificationGraph Neural NetworkTransformerImageMultimodalityTabularMagnetic Resonance ImagingAlzheimer's Disease
🎯 What it does: Proposed a dynamic weighted dual-graph attention network (DW-DGAT) for fusing multimodal, multi-structure neuroimaging and phenotypic data to enable early diagnosis of Parkinson's disease and Alzheimer's disease.
DWTSG: Parameter-Efficient Fine-Tuning of Large Pre-trained Models via Discrete Wavelet Transform and Subband Guidance
Chengwei Sun, Yang Yang (University Of Electronic Science And Technology Of China)
Computational EfficiencyKnowledge DistillationSupervised Fine-TuningImageText
🎯 What it does: Propose a parameter-efficient fine-tuning framework called DWTSG based on discrete wavelet transform (DWT), which utilizes a subband energy selection strategy to update only the coefficients with the strongest structural expressiveness, achieving efficient adaptation of large pre-trained models.
DyC-STG: Dynamic Causal Spatio-Temporal Graph Network for Real-time Data Credibility Analysis in IoT
Guanjie Cheng (Zhejiang University), Shuiguang Deng (Zhejiang University)
Anomaly DetectionGraph Neural NetworkTransformerMultimodalityTime Series
🎯 What it does: Propose the DyC-STG framework for real-time IoT data credibility assessment, integrating event-driven dynamic graphs with causal Transformers;
Dynamic Agent Grouping ECBS: Scaling Windowed Multi-Agent Path Finding with Completeness Guarantees
Tiannan Zhang (Carnegie Mellon University), Maxim Likhachev (Carnegie Mellon University)
OptimizationBenchmark
🎯 What it does: This paper proposes Dynamic Agent Grouping ECBS (DAG-ECBS), which uses a bounded suboptimal solver within the Windowed Complete MAPF framework to solve windowed multi-agent pathfinding problems while maintaining completeness.
Dynamic Cognitive Planning for Cognitive-Functional Dialogue: A Case Study in Emotional Support Conversation
Jiaqi Liu (Beijing Normal University), Wenbin Jiang (Beijing Normal University)
TransformerLarge Language ModelTextSequential
🎯 What it does: Proposed a dynamic cognitive planning method (DyCoP) to generate dialogue guidance plans in real-time based on the user's cognitive psychological evolution in emotional support dialogues, achieving more effective cognitive functional conversations.
Dynamic Deep Graph Learning for Incomplete Multi-View Clustering with Masked Graph Reconstruction Loss
Zhenghao Zhang (University of Chinese Academy of Sciences), Jungang Xu (University of Chinese Academy of Sciences)
Representation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposes DGIMVCM, a dynamic deep graph learning framework for incomplete multi-view clustering, which compensates for missing views and performs clustering through global graph fusion, GCN embedding layer, GAT encoder, mask graph reconstruction loss, and pseudo-label self-supervised module.
Dynamic Deep Prompt Optimization for Defending Against Jailbreak Attacks on LLMs
Doniyorkhon Obidov (Michigan Technological University), Kaichen Yang (Michigan Technological University)
Safty and PrivacyAdversarial AttackLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextBenchmark
🎯 What it does: Proposes a defense method based on dynamic deep hint optimization (DDPO), which generates input-dependent defense embeddings using intermediate features of the LLM itself and injects them into subsequent layers;
Dynamic Gaussian Scene Reconstruction from Unsynchronized Videos
Zhixin Xu (Tsinghua University), Bin Wang (Tsinghua University)
GenerationOptimizationGaussian SplattingVideo
🎯 What it does: Proposed a pluggable coarse-to-fine temporal alignment module to address the problem of 4D high-quality Gaussian reconstruction caused by asynchrony in multi-view videos.
Dynamic Geometric Equivariant Network for Full-Atom Antibody Design
Weihong Huang (Wuhan University), Juan Liu (Wuhan University)
Drug DiscoveryProtein Structure PredictionGraph Neural NetworkBiomedical Data
🎯 What it does: Proposed and implemented an end-to-end all-atom antibody design model called DGENet, which can simultaneously predict the amino acid sequence and three-dimensional structure of antibodies, and perform affinity optimization at the antibody-antigen binding site.
Dynamic Semantic Tokenization for Time Series via Elastic Sampling on Physics-aware Perception
Huaizhang Liao (National University of Defense Technology), Yongxiang Liu (National University of Defense Technology)
ClassificationRepresentation LearningRecurrent Neural NetworkTransformerContrastive LearningTime SeriesBiomedical DataPhysics Related
🎯 What it does: Investigated physics-aware semantic segmentation for time series, proposing the PATK framework to achieve elastic adaptive segmentation.
Dynamic Sparsity: Challenging Common Sparsity Assumptions for Learning World Models in Robotic Reinforcement Learning Benchmarks
Muthukumar Pandaram (University of Innsbruck), Justus Piater (University of Innsbruck)
Robotic IntelligenceReinforcement LearningWorld ModelBenchmark
🎯 What it does: Empirical analysis using the real dynamics and Jacobian from MuJoCo Playground to evaluate global, state-dependent, and temporal sparsity, and to test the performance of simple MLPs in capturing these sparse structures.
Dynamic Weight Adaptation in Spiking Neural Networks Inspired by Biological Homeostasis
Yunduo Zhou (Dalian University of Technology), Xin Yang (Dalian University of Technology)
Robotic IntelligenceSpiking Neural NetworkReinforcement LearningSequential
🎯 What it does: Propose a dynamic weight adaptation mechanism (DWAM) based on BCM theory, enabling spiking neural networks to spontaneously achieve neuronal steady-state regulation during the inference phase;
Dynamic-Static Collaboration for Unsupervised Domain Adaptive Video-Based Visible-Infrared Person Re-Identification
Jiaxu Leng (Chongqing University of Posts and Telecommunications), Xinbo Gao (Chongqing University of Posts and Telecommunications)
RecognitionDomain AdaptationTransformerContrastive LearningVideo
🎯 What it does: This paper studies the task of unsupervised domain adaptation for video visible-infrared person re-identification (UDA-VVIReID) and proposes the DSC framework based on dynamic-static collaboration.
Dynamic-Static Synergistic Selection Method for Candidate Code Solutions with Generated Test Cases
Ren-Biao Liu (Nanjing University), Ming Li (Nanjing University)
AI Code AssistantLarge Language ModelText
🎯 What it does: Propose a dynamic-static collaborative selection method (DS3), which first performs AST filtering and static analysis on the code and test cases generated by LLMs, and then selects the optimal code implementation by scoring candidate code through dynamic execution consistency.
DynamicEarth: How Far Are We from Open-Vocabulary Change Detection?
Kaiyu Li, Zhi Wang (Xi'an Jiaotong University)
ClassificationObject DetectionSegmentationTransformerPrompt EngineeringVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Proposed the Open-Vocabulary Change Detection (OVCD) task and designed two unsupervised frameworks, M-C-I and I-M-C, based on pre-trained foundation models, for detecting changes of arbitrary categories in remote sensing images.
DynamicRTL: RTL Representation Learning for Dynamic Circuit Behavior
Ruiyang Ma (Peking University), Guojie Luo (Nanjing University of Aeronautics and Astronautics)
Representation LearningRecurrent Neural NetworkGraph Neural NetworkGraphTime Series
🎯 What it does: A dynamic behavior-based RTL-level circuit representation learning model, DR-GNN, was researched and implemented, along with the construction of the first dynamic circuit dataset containing 6,300 Verilog designs and 63,000 simulation trajectories.
DynaQuant: Dynamic Mixed-Precision Quantization for Learned Image Compression
Youneng Bao (Shenzhen University), Yongsheng Liang (Shenzhen University)
CompressionImage
🎯 What it does: Propose DynaQuant, a dynamic mixed-precision quantization framework for learned image compression, achieving content-aware quantization parameters and dynamic bit-width selection;
DySy-Det: A Synergistic Framework with Dynamic Reconstruction-Path Consistency for AI-Generated Image Detection
Fanli Jin (Zhejiang University), Zhisheng Yan (George Mason University)
Anomaly DetectionTransformerVision Language ModelDiffusion modelImage
🎯 What it does: Propose a unified framework named DySy-Det for detecting AI-generated images, integrating multi-dimensional features such as semantic consistency, local reconstruction error, and dynamic generation trajectory consistency.
E-Logic Prompt: Unified Energy-Logic Framework for Continual Visual Question Answering
Jiayao Tan (Tianjin University), Liang Wan (Tianjin University)
Prompt EngineeringVision Language ModelScore-based ModelMultimodality
🎯 What it does: This paper proposes a unified energy logic prompt framework (E-Logic Prompt) based on energy models to address the knowledge forgetting problem in continuous visual question answering.
E-MaT:Event-oriented Mamba for Egocentric Point Tracking
Han Han (University of Science and Technology of China), Zheng-jun Zha (University of Science and Technology of China)
Object TrackingVideoMultimodality
🎯 What it does: Proposed the E-MaT framework, utilizing an event camera to achieve first-person perspective point tracking, comprising an event-oriented Mamba encoder and a motion-adaptive suppression module.
E³SAM2: Entropy-Aware and Edge-Guided Adaptation of SAM2 for Echocardiography Video Segmentation
Long Zheng (Guizhou University), Shuyun Li (Guizhou University)
SegmentationTransformerSupervised Fine-TuningVideoBiomedical DataUltrasound
🎯 What it does: Propose a lightweight E3SAM2 framework for cardiac ultrasound video segmentation, combining entropy-guided attention, entropy regularization, and edge supervision;
EA-VAE: Learning to Reconstruct Dysarthric Speech via Variational Autoencoder with Encoding Alignment
Daipeng Zhang (Tianjin University), Jianguo Wei (Tianjin University)
RestorationRecurrent Neural NetworkTransformerAuto EncoderAudio
🎯 What it does: Proposed the EA-VAE model, which uses a variational autoencoder combined with encoding alignment, distribution alignment, and duration alignment modules to directly map disordered speech to intelligible normal speech.
EAGLE: Episodic Appearance- and Geometry-aware Memory for Unified 2D-3D Visual Query Localization in Egocentric Vision
Yifei Cao (Dalian University of Technology), Xun Tu (Dalian University of Technology)
Object TrackingSegmentationRetrievalOptimizationMeta LearningVideo
🎯 What it does: Proposed the EAGLE framework for achieving unified 2D-3D visual query localization in first-person perspective.
EARG-Net: Edge-Aware Reconstruction-Guided Network for Image Manipulation Detection and Localization
Yanpu Yu (University Of Science And Technology Of China), Nenghai Yu (University Of Science And Technology Of China)
SegmentationAnomaly DetectionTransformerAuto EncoderImage
🎯 What it does: Proposed an edge-aware reconstruction-guided network (EARG-Net), which achieves image tampering detection and localization by masking suspected tampered regions, reconstructing using a pre-trained image inpainting model, and then extracting fine-grained forgery traces through reconstruction residuals.
Early Warning of Intraoperative Adverse Events via Transformer-Driven Multi-Label Learning
Xueyao Wang (Chengdu Institute of Computer Application, Chinese Academy of Sciences), Yu Yao (China Zhenhua Research Institute Co., Ltd.)
Anomaly DetectionTransformerTime SeriesBiomedical DataElectronic Health Records
🎯 What it does: This paper proposes an early warning method for multi-label intraoperative adverse events (including hypotension, low sedation, arrhythmia, hypoxemia, hypothermia, and low carbon dioxide emission), and constructs the first multi-label dataset, MuAE;
Earth-Adapter: Bridge the Geospatial Domain Gaps with a Frequency-Guided Mixture of Adapters
Xiaoxing Hu (Beijing Institute of Technology), Xue Yang (Peking University)
SegmentationDomain AdaptationTransformerSupervised Fine-TuningMixture of ExpertsImage
🎯 What it does: This paper proposes Earth-Adapter, a parameter-efficient fine-tuning method specifically designed for remote sensing segmentation tasks, aiming to address the issue of performance degradation caused by traditional PEFT methods being easily disturbed by noise and artifacts in remote sensing images.
EarthCrafter: Scalable 3D Earth Generation via Dual-Sparse Latent Diffusion
Shang Liu (DAMO Academy, Alibaba Group), Fan Wang (DAMO Academy, Alibaba Group)
GenerationData SynthesisConvolutional Neural NetworkTransformerDiffusion modelFlow-based ModelGaussian SplattingImageMesh
🎯 What it does: Propose the EarthCrafter framework, combined with the Aerial-Earth3D dataset to achieve large-scale 3D Earth generation
Easy for Children, Hard for AI: The Limits of Multimodal LLMs in Early Childhood Learning
Jingping Liu (Sun Yat-sen University), Huacan Wang (University of Chinese Academy of Sciences)
Supervised Fine-TuningVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Proposed and constructed ChildBench — a multi-modal benchmark for evaluating models' cognitive abilities in early childhood learning, including spatial reasoning, visual reasoning, visual discrimination, counting, and visual tracking;
Easy to Learn, Yet Hard to Forget: Towards Robust Unlearning Under Bias
JuneHyoung Kwon (Chung-Ang University), YoungBin Kim (Chung-Ang University)
Explainability and InterpretabilityComputational EfficiencyImage
🎯 What it does: This paper investigates the failure of machine unlearning in biased models, proposing a three-stage CUPID framework to achieve robust category-level forgetting.
EasyText: Controllable Diffusion Transformer for Multilingual Text Rendering
Runnan Lu (National University of Singapore), Yiren Song (National University of Singapore)
GenerationTransformerDiffusion modelImageText
🎯 What it does: Propose the EasyText framework, which achieves controllable multilingual text rendering based on Diffusion Transformer; combines character positioning encoding with position interpolation to support text generation at arbitrary positions and shapes.
EC-MVSNet: Enhanced Cascaded Multi-View Stereo with Cross-Scale Relevance Integration
Shaoqian Wang (North China Electric Power University), Yuchao Dai (Northwestern Polytechnical University)
Depth EstimationConvolutional Neural NetworkImageBenchmark
🎯 What it does: This paper proposes EC-MVSNet, an enhanced cascaded multi-view stereo framework, which improves depth estimation accuracy by leveraging cross-scale feature joint cost volume construction, probabilistic guidance enhancement, and monocular feature refinement.
EccoMamba: Enhanced Cross-hierarchical Continuity Orthogonal Mamba for Medical Image Segmentation
Junlin Xu, Yajie Meng (Hainan University)
SegmentationConvolutional Neural NetworkBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: Proposes a medical image segmentation model named EccoMamba based on a U-shaped encoder-decoder architecture, addressing the sequential and single-scale limitations of the Mamba structure when processing 2D/3D medical images.
ECD: Evidence-guided Contrastive Decoding in Retrieval-Augmented Generation with Accurate Knowledge Reference Adjustment
Yize Sui (National University of Defense Technology), Wenjing Yang (National University of Defense Technology)
GenerationRetrievalTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: Propose an evidence-guided contrastive decoding method (ECD) based on the Dirichlet distribution for dynamically balancing the dependency between parameter knowledge and retrieval context in retrieval-augmented generation (RAG) systems;
EchoBat: Echo-Vision Enhancement and Echo-Layered Sampling for Video LLMs Hallucination Mitigation
Shuai Liu (Xi'an Jiaotong University), Chenhao Lin (Xi'an Jiaotong University)
OptimizationLarge Language ModelSupervised Fine-TuningVision Language ModelVideoTextRetrieval-Augmented GenerationAudio
🎯 What it does: This paper proposes EchoBat, which enhances video understanding using audio and constructs high-quality preference pairs through key frame selection and reverse video generation of adversarial samples, significantly reducing hallucinations in video LLMs via DPO fine-tuning.
EchoEdit: Consistent Multi-Hop Question Answering via Ripple Control in Knowledge Editing
Jinwei Shi, Tieke He (Nanjing University)
Graph Neural NetworkLarge Language ModelTextGraphChain-of-Thought
🎯 What it does: This paper proposes EchoEdit, an impact-control-based knowledge editing framework that maintains coherence and consistency in multi-hop reasoning without retraining the model.
EchoGen: Cycle-Consistent Learning for Unified Layout-Image Generation and Understanding
Kai Zou (University of Science and Technology of China), Bin Liu (Tongji University)
Object DetectionGenerationTransformerReinforcement LearningAuto EncoderImage
🎯 What it does: Proposed the EchoGen unified framework, which jointly addresses layout-to-image generation and image localization tasks, achieving high-quality, layout-accurate image synthesis and precise image localization within the same model.
Echoless Label-Based Pre-computation for Memory-Efficient Heterogeneous Graph Learning
Jun Hu (National University of Singapore), Bingsheng He (Institute of Automation, Chinese Academy of Sciences)
Computational EfficiencyRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: Propose a precomputation framework named Echoless-LP to address the echo leakage problem caused by label precomputation in heterogeneous graph learning;
EchoMimicV3: 1.3B Parameters Are All You Need for Unified Multi-Modal and Multi-Task Human Animation
Rang Meng (Ant Group), Chenguang Ma (Ant Group)
GenerationTransformerDiffusion modelVideoTextMultimodalityAudio
🎯 What it does: Propose EchoMimicV3, a unified multi-task, multi-modal human animation framework with 1.3B parameters, capable of performing multiple animation tasks such as text-to-video, image-to-video, and audio-driven lip-sync within the same model.
EcoAgent: An Efficient Device-Cloud Collaborative Multi-Agent Framework for Mobile Automation
Biao Yi (Zhejiang University), Fan Wu (Hong Kong Polytechnic University)
Safty and PrivacyComputational EfficiencyAI Code AssistantTransformerLarge Language ModelVision-Language-Action ModelMultimodalityBenchmarkChain-of-Thought
🎯 What it does: Propose EcoAgent, a closed-loop device-cloud collaborative multi-agent framework for achieving privacy-friendly, efficient, and responsive mobile automation;
ECPv2: Fast, Efficient, and Scalable Global Optimization of Lipschitz Functions
Fares Fourati (KAUST), Vaneet Aggarwal (Purdue University)
OptimizationBenchmark
🎯 What it does: Proposed the ECPv2 algorithm for global optimization under unknown Lipschitz constants, improving the efficiency and scalability of ECP.
Edge Consistency for 4D Gaussian Splatting in Dynamic Scene Rendering
Boya Shi (Shanghai Jiao Tong University), Yi Xiaodong
GenerationNeural Radiance FieldGaussian SplattingOptical FlowVideo
🎯 What it does: Propose Edge4DGS, which utilizes a hybrid geometric representation combining Gaussian spheres and convex hulls, along with continuous-time edge consistency regularization, to achieve real-time rendering of dynamic scenes from sparse monocular inputs.
Edge Self-Adversarial Augmentation Enhances Graph Contrastive Learning Against Neighborhood Inconsistency
Chunchun Chen (Tongji University), Wei Ye (Tongji University)
Representation LearningAdversarial AttackGraph Neural NetworkGenerative Adversarial NetworkContrastive LearningGraph
🎯 What it does: Propose an edge self-adversarial enhanced graph contrastive learning framework named EDA-GCL, which can generate two adversarial views with neighborhood inconsistency under unsupervised settings and perform contrastive learning using them.
Edge-Binary Public Goods Games
Thekla Hamm (Eindhoven University of Technology), Paloma T. de Lima (Norwegian School of Economics)
OptimizationGraph
🎯 What it does: This paper proposes a novel Edge-Binary Public Goods Game (EBPG) and investigates the computational complexity of the existence of pure Nash equilibria under different graph structures.
Edge-Centric Relational Reasoning for 3D Scene Graph Prediction
Yanni Ma (Sun Yat-sen University), Martin R. Oswald (University of Amsterdam)
Object DetectionRecurrent Neural NetworkGraph Neural NetworkPoint Cloud
🎯 What it does: This paper proposes a relation reasoning framework called LEO, which transitions from edges to objects for 3D scene graph prediction;