AAAI 2026 Papers — Page 8
AAAI Conference on Artificial Intelligence · 4149 papers
CoT-VLNBench: A Benchmark for Visual Chain-of-Thought Reasoning in Vision-Language-Navigation Robots
Xiao Zhao (Tencent), Kuifeng Su (Tencent)
Explainability and InterpretabilityRobotic IntelligenceLarge Language ModelSupervised Fine-TuningVision Language ModelImageVideoTextMultimodalityPoint CloudBenchmarkChain-of-Thought
🎯 What it does: This paper proposes CoT-VLNBench, the first large-scale navigation benchmark for visual chain-of-thought (CoT) reasoning in quadruped robots, and trains the CoT-VLN model based on this benchmark, achieving interpretable multi-step navigation in complex indoor and outdoor environments.
CounterBench: Evaluating and Improving Counterfactual Reasoning in Large Language Models
Yuefei Chen (Rutgers University), Ruixiang Tang (Rutgers University)
TransformerLarge Language ModelTextBenchmarkChain-of-Thought
🎯 What it does: Propose the CounterBench dataset to evaluate the counterfactual reasoning capabilities of LLMs, and introduce the CoIn reasoning framework to enhance performance
Counterfactual eXplainable AI (XAI) Method for Deep Learning-Based Multivariate Time Series Classification
Alan Gabriel Paredes Cetina (SnT University of Luxembourg), Sylvain Kubler (SnT University of Luxembourg)
ClassificationExplainability and InterpretabilityConvolutional Neural NetworkTime Series
🎯 What it does: Proposed a multi-objective adversarial counterfactual explanation method for time series called CONFETTI, which can significantly reduce the modification of time series while maintaining confidence and enhancing proximity.
Counterfactual Fairness with Imperfect Causal Graphs
Cong Su (Shandong University), Guoxian Yu (Shandong University)
OptimizationExplainability and InterpretabilityGraphTabular
🎯 What it does: Propose the CF-ICG framework to estimate and guarantee counterfactual fairness for partially known causal graphs (CPDAGs).
Counterfactual Planning for Generalizable Agents’ Actions
Jiarun Fu (Beijing Institute of Technology), Junyu Zhang (Beijing Institute of Technology)
Explainability and InterpretabilityTransformerLarge Language ModelReinforcement LearningPrompt EngineeringWorld Model
🎯 What it does: Propose the Counterfactual Planning framework, modeling the action planning of LLM-driven agents as a structural causal model, and enhancing generalization and adaptability in complex dynamic environments by dynamically inferring environmental confounding factors.
Counterfactual Question Generation Uncovering Learner Contradictions
Bo Zhang (Nanjing Normal University), Junsheng Zhou (Nanjing Normal University)
GenerationExplainability and InterpretabilityTransformerLarge Language ModelPrompt EngineeringTextGraphBenchmarkRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose the GapProbe framework for generating adversarial follow-up questions (CFQ) by interacting Large Language Models (LLMs) with knowledge graphs (KGs) to reveal hidden contradictions in learners' answers and promote reflection.
Counterfactual-based Cognitive Alignment In-Context Learning for Relation Extraction
Qibin Li (Dalian University Of Technology), Nianmin Yao
RecognitionTransformerLarge Language ModelText
🎯 What it does: Propose the Counterfactual-based Cognitive Alignment (CCA) framework, which leverages human counterfactual reasoning and cognitive structure identification to generate and align the example space, thereby enhancing the In-Context Learning (ICL) performance of large language models (LLMs) in relation extraction tasks.
Counterfactual-Driven Zero-Shot Classifier Expansion
Xiangyu Wang (University of Science and Technology of China), Huanhuan Chen (University of Science and Technology of China)
ClassificationTransformerLarge Language ModelVision Language ModelAuto EncoderContrastive LearningImageTextMultimodalityBenchmark
🎯 What it does: Developed a framework for image-free zero-shot classifier expansion based on counterfactual causal reasoning, which leverages large language models to automatically generate factual and counterfactual textual descriptions. It explicitly infers classification boundaries in the weight space through symmetric attention mutual purification, dual decoder alignment networks, and separation loss, enabling inference for new categories without images.
Covariance Scattering Transforms
Andrea Cavallo (Delft University of Technology), Elvin Isufi (Delft University of Technology)
Representation LearningBiomedical DataAlzheimer's Disease
🎯 What it does: Proposed Covariance Scattering Transforms (CST), a training-free deep network that achieves data representation through local wavelet filtering of covariance spectra.
Coverage-Constrained Human-AI Cooperation with Multiple Experts
Zheng Zhang (University of Surrey), Gustavo Carneiro (University of Surrey)
ClassificationMixture of ExpertsImageBiomedical Data
🎯 What it does: This paper proposes the CL2DC method, achieving a coverage-constrained human-AI collaborative classification system that integrates learning-based delay (L2D) and compensation (L2C) in a multi-expert environment.
COVR: Collaborative Optimization of VLMs and RL Agent for Visual-Based Control
Canming Xia (Sun Yat-sen University), Luntong Li (Peking University)
Autonomous DrivingOptimizationRobotic IntelligenceSupervised Fine-TuningReinforcement LearningVision Language ModelImage
🎯 What it does: This paper proposes the COVR framework, which leverages the collaboration between vision-language models (VLM) and visual reinforcement learning (RL): first, the interaction data generated by RL is used to fine-tune the VLM (with the EDDF and RALW modules filtering high-quality trajectories and weighting losses), and then the action guidance information generated by the fine-tuned VLM is used as a regularization constraint to guide RL policy learning, achieving bidirectional improvement between VLM and RL throughout the process.
CP-CLIP: Customized Parameter Generation for Open-vocabulary Semantic Segmentation
Zelin Peng (Shanghai Jiao Tong University), Wei Shen (Shanghai Jiao Tong University)
SegmentationTransformerSupervised Fine-TuningVision Language ModelImage
🎯 What it does: Propose CP-CLIP, a fine-tuning method for open-vocabulary semantic segmentation. Without freezing the pre-trained CLIP weights, it generates per-image custom parameters using a custom parameter generator, achieving pixel-level text alignment.
CP-FREEZER: Latency Attacks Against Vehicular Cooperative Perception
Chenyi Wang (University of Arizona), Ming Li (University of Arizona)
Autonomous DrivingComputational EfficiencyAdversarial AttackPoint CloudBenchmark
🎯 What it does: Designed and implemented CP-FREEZER, a delay attack targeting vehicle cooperative perception (Cooperative Perception, CP), which exploits adversarial perturbations on shared V2V information to maximize the execution time of CP algorithms, thereby degrading the availability of perception systems.
CP-Router: An Uncertainty-Aware Router Between LLM and LRM
Jiayuan Su (Zhejiang University), Hongwei Wang (University Of Illinois Chicago)
Computational EfficiencyTransformerTextBenchmark
🎯 What it does: Developed a training-free, model-agnostic routing framework called CP-Router, which dynamically selects between LLM and LRM based on the size of the prediction set from Conformal Prediction;
CP-Search: A Chain Progressive Search Training Framework Incentivizing the Cognitive Behaviors for Searching in LLMs
Zehua Wang (Harbin Institue of Technology (Shenzhen)), Buzhou Tang (Harbin Institue of Technology (Shenzhen))
RetrievalOptimizationTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextChain-of-Thought
🎯 What it does: This study proposes CP-Search, a two-phase training framework designed to enhance the retrieval capabilities of large language models in multi-hop reasoning scenarios;
CPOStream: Collaborating Prediction and Observation for Flicker-Free Streamable Free-Viewpoint Video with 3DGS
Zhenyu Bao (Peking University), Kanglin Liu (Pengcheng Laboratory)
GenerationGaussian SplattingVideo
🎯 What it does: Proposes the CPOStream framework, which dynamically freezes or activates 3D Gaussians through a prediction and observation module to achieve flicker-free rendering in real-time streaming free-viewpoint videos.
CR³: Boosting Compositional Reasoning in MLLMs Through Rule-Based Reinforcement Learning
Shun Qian (Harbin Institute of Technology), Baoxun Wang (Tencent)
Supervised Fine-TuningReinforcement LearningContrastive LearningMultimodality
🎯 What it does: Propose the CR³ framework, which enhances the compositional reasoning ability of multimodal large language models (MLLMs) through rule-based reinforcement learning.
CRAF: A Clinical Reasoning-Adaptive Framework via Reinforcement Learning for Similar Case Retrieval
Jie Lin (Xiamen University), Liansheng Wang (Xiamen University)
RetrievalTransformerLarge Language ModelReinforcement LearningBiomedical Data
🎯 What it does: Proposed the CRAF framework for clinical similar case retrieval, achieving query rewriting through reinforcement learning to generate diagnostic reasoning paths.
Creating Blank Canvas Against AI-enabled Image Forgery
Qi Song (Hong Kong Baptist University), Renjie Wan (Hong Kong Baptist University)
Anomaly DetectionSafty and PrivacyAdversarial AttackTransformerImage
🎯 What it does: Proposed an active image forgery localization method that converts the original image into a 'blank canvas' through frequency domain adaptive adversarial perturbation, making the SAM model unable to recognize image content in a seamless state; when the image is maliciously tampered, SAM can detect anomalous regions.
CreBench: Human-Aligned Creativity Evaluation from Idea to Process to Product
Kaiwen Xue (Beijing University of Posts and Telecommunications), Jiayi Cen (University of Southampton)
TransformerLarge Language ModelSupervised Fine-TuningVision Language ModelMultimodalityBenchmark
🎯 What it does: Constructed the CreBench evaluation benchmark and the CreMIT multimodal instruction-tuning dataset, and trained the CreExpert model on this foundation for multi-dimensional assessment of human creativity.
Credal Ensemble Distillation for Uncertainty Quantification
Kaizheng Wang (KU Leuven), Hans Hallez (KU Leuven)
ClassificationExplainability and InterpretabilityKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: This paper proposes the Credal Ensemble Distillation (CED) framework, which compresses Deep Ensemble (DE) into a single model called CREDIT. CREDIT forms a credible set by outputting class probability intervals, achieving simultaneous quantification of uncertainty;
CRISP: Curriculum-Inducing Primitive Informed Subgoal Prediction for Boosting Hierarchical Reinforcement Learning
Utsav Singh (IIT Kanpur), Vinay P. Namboodiri (University of Bath)
Robotic IntelligenceReinforcement LearningGenerative Adversarial Network
🎯 What it does: This paper proposes the CRISP framework, which generates reachable subgoals by adaptively parsing expert demonstrations, supplemented with inverse reinforcement learning regularization to stabilize hierarchical reinforcement learning.
CroPS: Improving Dense Retrieval with Cross-Perspective Positive Samples in Short-Video Search
Ao Xie (Kuaishou Technology), Han Li (Kuaishou Technology)
RetrievalRecommendation SystemLarge Language ModelContrastive LearningVideo
🎯 What it does: Propose the CroPS framework, which enriches positive samples in short video search by leveraging query rewriting, recommendation system interaction, and external knowledge generated by large language models, and trains a dual-tower retriever using hierarchical label assignment and H-InfoNCE loss.
Cross Modal Fine-grained Alignment via Granularity-aware and Region-uncertain Modeling
Jiale Liu (South China Normal University), Yuncheng Jiang (Harbin Institute of Technology)
RetrievalTransformerVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Propose the GRM framework, employing significance-aware and granularity-aware adapters, region prompts, and Gaussian Mixture Uncertainty Modeling to achieve cross-modal fine-grained image-text alignment.
Cross-Domain Few-Shot Learning via Multi-View Collaborative Optimization with Vision-Language Models
Dexia Chen (Sun Yat-sen University), Ruixuan Wang (Sun Yat-sen University)
Domain AdaptationMeta LearningImageTextBenchmark
🎯 What it does: Propose a cross-domain few-shot learning framework named CoMuCo, which utilizes dual expert modules to achieve multi-perspective feature co-optimization.
Cross-domain Joint Learning with Prototype-guided Mixture-of-Experts for Infrared Moving Small Target Detection
Weiwei Duan (University of Electronic Science and Technology of China), Mao Ye (University of Electronic Science and Technology of China)
Object DetectionConvolutional Neural NetworkMixture of ExpertsContrastive LearningImage
🎯 What it does: Propose a cross-domain joint learning framework named CoMoE to address the limitations of single-domain learning in infrared small target detection, constructing a detector that is generalizable across multiple datasets.
Cross-Field Interface-Aware Neural Operators for Multiphase Flow Simulation
Zhenzhong Wang (Xiamen University), Min Jiang (Xiamen University)
Convolutional Neural NetworkTransformerPhysics Related
🎯 What it does: Propose an Interface Information-Aware Neural Operator (IANO) specifically designed for high-precision, low-cost simulation of multiphase flows.
Cross-Granularity Hypergraph Retrieval-Augmented Generation for Multi-hop Question Answering
Changjian Wang (Mashang Consumer Finance Co Ltd), Ning Jiang (Mashang Consumer Finance Co Ltd)
GenerationRetrievalGraph Neural NetworkLarge Language ModelTextGraphRetrieval-Augmented Generation
🎯 What it does: Proposed a cross-grained hypergraph retrieval augmented generation (HGRAG) framework, utilizing entity hypergraphs, hypergraph diffusion retrieval, and retrieval augmented modules to jointly exploit structural and semantic information from fine-grained entities and coarse-grained paragraphs for multi-hop question answering retrieval and answer generation.
Cross-modal Prompting for Balanced Incomplete Multi-modal Emotion Recognition
Wen-Jue He (Harbin Institute of Technology), Zheng Zhang (Harbin Institute of Technology)
RecognitionPrompt EngineeringMultimodality
🎯 What it does: This paper proposes a cross-modal prompting (ComP) framework that enhances the robustness of incomplete multimodal sentiment recognition by leveraging cross-modal knowledge propagation and prompt generation.
Cross-modal Proxy Evolving for OOD Detection with Vision-Language Models
Hao Tang (Hong Kong Polytechnic University), Jing Qin (National University of Singapore)
Anomaly DetectionTransformerVision Language ModelContrastive LearningImageText
🎯 What it does: Proposed the CoEvo framework, leveraging a pre-trained vision-language model (CLIP) without training samples, bidirectionally dynamically co-evolving text and visual agents to achieve zero-shot OOD detection.
Cross-Modal Unlearning via Influential Neuron Path Editing in Multimodal Large Language Models
Kunhao Li (South China University Of Technology), Jason Xue (Southeast University)
Explainability and InterpretabilityTransformerLarge Language ModelMultimodalityBenchmark
🎯 What it does: Proposed a cross-modal machine unlearning method called MIP-Editor, which can achieve targeted forgetting of specific knowledge in multimodal large language models by locating and editing path neurons that influence the model, while maintaining the model's overall capabilities.
Cross-Sample Augmented Test-Time Adaptation for Personalized Intraoperative Hypotension Prediction
Kanxue Li, Baosheng Yu (Yunnan United Vision Technology Company Limited)
RetrievalDomain AdaptationAnomaly DetectionTransformerAuto EncoderContrastive LearningTime SeriesBiomedical DataElectronic Health Records
🎯 What it does: Propose the CSA-TTA framework, achieving test-time personalized prediction of intraoperative hypotension (IOH) through cross-sample augmentation, coarse-to-fine retrieval, and multi-task optimization.
Cross-Scale Collaboration between LLMs and Lightweight Sequential Recommenders with Domain-Specific Latent Reasoning
Yipeng Zhang (Tsinghua University), Wenwu Zhu (Tencent Inc)
Recommendation SystemTransformerLarge Language ModelTextSequential
🎯 What it does: Propose the CoderRec framework, which combines semantic IDs generated by large language models (LLMs) with a lightweight sequential recommender through cross-scale model collaboration, and introduces domain-specific latent reasoning to enhance sequential recommendation performance.
Cross-Space Synergy: A Unified Framework for Multimodal Emotion Recognition in Conversation
Xiaosen Lyu (Huaqiao University), Jing Wang (Huaqiao University)
RecognitionOptimizationRepresentation LearningTransformerMultimodality
🎯 What it does: This paper proposes the Cross-Space Synergy (CSS) framework for multi-modal dialogue emotion recognition, synergistically coupling representation learning and gradient optimization.
Cross-temporal 3D Gaussian Splatting for Sparse-view Guided Scene Update
Zeyuan An (Beihang University), Xiaohui Liang (Beihang University)
RestorationGenerationOptimizationGaussian SplattingImagePoint Cloud
🎯 What it does: This paper proposes a cross-temporal scene update framework based on 3D Gaussian light scattering, which can update or recover 3D scenes at different time points by leveraging sparse viewpoints and historical prior information.
Cross-view Anchor Graph Learning and Factorization for Incomplete Multi-view Clustering
Xinxin Wang (Shenzhen University), Yicong Zhou (Macau Polytechnic University)
Representation LearningGraph Neural NetworkMultimodality
🎯 What it does: Proposes a cross-view anchor graph learning and decomposition (AGLF) method for handling incomplete multi-view clustering problems, directly performing subgraph learning and generating soft labels through the anchor graph without requiring post-processing.
Cross-View Progressive Feature Filtering for Multi-View Graph Clustering in Remote Sensing
Bowen Liu (Hainan University), Miao Yu (Hainan Blockchain Technology Engineering Research Center)
Representation LearningGraph Neural NetworkContrastive LearningMultimodalityGraph
🎯 What it does: Propose a cross-perspective progressive feature filtering multi-view graph clustering framework, CF-MVGC, for unsupervised clustering of remote sensing data.
CrossCheck-Bench: Diagnosing Compositional Failures in Multimodal Conflict Resolution
Baoliang Tian (ByteDance), Minghui Qiu (ByteDance)
Prompt EngineeringVision Language ModelMultimodalityBenchmark
🎯 What it does: Designed and constructed the CrossCheck-Bench benchmark, which includes three cognitive levels (perception, integration, reasoning) and seven atomic capabilities, generating 15,000 conflicting question-answer pairs derived from real e-commerce listings.
CrossCut: Cross-Patch Aware Interactive Segmentation for Remote Sensing Images
Zheng Lin, Bojian Zhang (Tsinghua University)
SegmentationTransformerPrompt EngineeringImage
🎯 What it does: Proposes a cross-patch aware interactive segmentation framework called CrossCut for target segmentation in remote sensing images.
CrossVid: A Comprehensive Benchmark for Evaluating Cross-Video Reasoning in Multimodal Large Language Models
Jingyao Li (Xiaohongshu Inc), Yao Hu (Xiaohongshu Inc)
Large Language ModelPrompt EngineeringVision Language ModelVideoTextMultimodalityBenchmark
🎯 What it does: Proposed the CrossVid benchmark to evaluate the capabilities of multimodal large language models in cross-video reasoning (CVR) tasks.
CrystalDiT: Simple Diffusion Transformers for Crystal Generation
Xiaohan Yi (Tsinghua University), Peilin Zhao (Shanghai Jiao Tong University)
GenerationData SynthesisTransformerDiffusion modelScore-based ModelGraphTabularPhysics Related
🎯 What it does: Proposed CrystalDiT, a unified diffusion Transformer for generating crystal structures with physical feasibility and novelty.
CSP4SDG: Constraint and Information-Theory Based Role Identification in Social Deduction Games with LLM-Enhanced Inference
Kaijie Xu (McGill University), Simon Mark Lucas (Queen Mary University of London)
Explainability and InterpretabilityLarge Language ModelText
🎯 What it does: Proposed a training-free, constraint-based probabilistic framework called CSP4SDG for character inference in social deduction games, which integrates LLM to automatically extract language-agnostic constraints and update character posterior distributions in real-time.
CtrlFuse: Mask-Prompt Guided Controllable Infrared and Visible Image Fusion
Yiming Sun (Southeast University), Pengfei Zhu (Southeast University)
Object DetectionSegmentationTransformerPrompt EngineeringImage
🎯 What it does: Proposed a controllable infrared-visible image fusion framework called CtrlFuse based on mask prompts, which can dynamically control the fusion process according to user-specified masks while simultaneously enhancing downstream semantic segmentation performance.
CTX-Coder: Cross-Attention Architectures Empower LLMs for Long-Context Vulnerability Detection
Jujie Wang (Beijing University of Posts and Telecommunications), Minjiao Yang (Beijing University of Posts and Telecommunications)
AI Code AssistantTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposed the CTX-Coder framework, which leverages cross-attention to integrate context function embedding vectors with target functions, thereby enhancing vulnerability detection performance in long contexts.
Cubing for Tuning
Haoze Wu (Amherst College), Nina Narodytska (Stanford University)
OptimizationBenchmark
🎯 What it does: This paper proposes an online adaptive strategy optimization method called TACO, which learns the optimal solving strategy by generating subproblems from a single problem itself.
CueBench: Advancing Unified Understanding of Context-Aware Video Anomalies in Real-World
Yating Yu (Northwestern Polytechnical University), Jiajun Zhang (Northwestern Polytechnical University)
Anomaly DetectionSupervised Fine-TuningReinforcement LearningVision Language ModelVideoBenchmark
🎯 What it does: Proposed the CUEBENCH benchmark, constructed a five-layer hierarchical dictionary containing absolute and conditional anomalous events, and provided a unified evaluation framework with multi-task (recognition, detection, localization, prediction) implementations for video anomaly understanding; simultaneously developed the CUE-R1 model, utilizing SFT+RFT RL fine-tuning to achieve unified generative anomaly understanding.
Cueing Without Gapping: Cuer-Independent Cued Speech Recognition Powered by Cross-Cuer Invariant Modeling
Fengji Ma (Hong Kong University of Science and Technology), Li Liu (Hong Kong University of Science and Technology)
RecognitionPose EstimationConvolutional Neural NetworkGraph Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelContrastive LearningVideoTextMultimodality
🎯 What it does: For cued speech recognition for hearing-impaired individuals, this paper proposes an ACSR system based on multimodal seamless alignment and language model decoding.
CyC3D: Fine-grained Controllable 3D Generation via Cycle Consistency Regularization
Hongbin Xu (South China University of Technology), Ming Li (Fudan University)
GenerationData SynthesisVision Language ModelNeural Radiance FieldMultimodalityPoint CloudMesh
🎯 What it does: Propose the CyC3D framework, incorporating conditional and view cycle consistency regularization during training, significantly enhancing the controllability and fine-grained control of 3D generation.
Cyto-SSL: A Self-Supervised Pretraining Framework for Cytology Foundation Model
Yiming Zhang (Beijing Institute of Technology), Bin Hu (Beijing Institute of Technology)
ClassificationRecognitionKnowledge DistillationRepresentation LearningConvolutional Neural NetworkTransformerContrastive LearningBiomedical Data
🎯 What it does: This paper proposes Cyto-SSL, a self-supervised pre-training framework specifically designed for cytological images, aimed at building a foundational model for cytology;
D-FCGS: Feedforward Compression of Dynamic Gaussian Splatting for Free-Viewpoint Videos
Wenkang Zhang (Shanghai Jiao Tong University), Zhengxue Cheng (Shanghai Jiao Tong University)
CompressionGaussian SplattingVideo
🎯 What it does: This paper proposes a forward compression framework named D-FCGS for scene-agnostic compression of dynamic 3D Gaussian Splatting, to support free-viewpoint video (FVV) applications.
D-GARA: A Dynamic Benchmarking Framework for GUI Agent Robustness in Real-World Anomalies
Sen Chen (Tongji University), Zheng Wang (Guangdong Laboratory of Artificial Intelligence and Digital Economy)
Anomaly DetectionBenchmark
🎯 What it does: This paper proposes the D-GARA framework for dynamically evaluating the robustness of Android GUI agents in real-world anomaly environments.
D2 Prune: Sparsifying Large Language Models via Dual Taylor Expansion and Attention Distribution Awareness
Lang Xiong (Chongqing University), Duo Liu (Chongqing University)
Computational EfficiencyKnowledge DistillationTransformerLarge Language ModelImageText
🎯 What it does: For post-training sparsification of large language models, D Prune 2 is proposed to achieve precise pruning and weight updates through dual Taylor expansion and attention distribution awareness.
D²-VPR: A Parameter-efficient Visual-foundation-model-based Visual Place Recognition Method via Knowledge Distillation and Deformable Aggregation
Zheyuan Zhang (Beijing University of Posts and Telecommunications), Hong Chen (Beijing University of Posts and Telecommunications)
RetrievalKnowledge DistillationRepresentation LearningTransformerImageBenchmark
🎯 What it does: Designed a lightweight visual localization method D2-VPR based on a vision foundation model, achieving efficient feature learning through knowledge distillation and deformable aggregation.
D2MoRA: Diversity-Regulated Asymmetric MoE-LoRA Decomposition for Efficient Multi-Task Adaptation
Jianhui Zuo (Shandong University), Liqiang Nie (Harbin Institute of Technology)
Computational EfficiencyRepresentation LearningMixture of ExpertsText
🎯 What it does: Propose D MoRA, a multi-task adaptation multi-expert LoRA framework that achieves flexible knowledge sharing through asynchronous expert splitting.
D²PPO: Diffusion Policy Policy Optimization with Dispersive Loss
Guowei Zou (Sun Yat-sen University), Haitao Wang (Sun Yat-sen University)
OptimizationRobotic IntelligenceTransformerReinforcement LearningDiffusion modelContrastive LearningImageBenchmark
🎯 What it does: This paper proposes a two-stage training framework called D²PPO, which enhances the representation diversity of diffusion strategies using dispersion loss during the pre-training phase, followed by fine-tuning with PPO to achieve efficient and precise robotic arm control.
D²Pruner: Debiased Importance and Structural Diversity for MLLM Token Pruning
Evelyn Zhang (Tencent YouTu Lab), Linfeng Zhang (Shanghai Jiao Tong University)
Computational EfficiencyTransformerLarge Language ModelVision Language ModelMultimodality
🎯 What it does: Propose D²Pruner, a visual token pruning framework that combines debiased importance and structural diversity
D3-RSMDE: 40× Faster and High-Fidelity Remote Sensing Monocular Depth Estimation
Ruizhi Wang (Zhejiang University), Li Sun
Depth EstimationTransformerDiffusion modelAuto EncoderImageBenchmark
🎯 What it does: Proposes the D3-RSMDE framework, which first generates a rough depth map as a structural prior using ViT, and then employs a lightweight diffusion model within the VAE latent space to perform Progressive Linear Blending Refinement, achieving high-fidelity remote sensing monocular depth estimation.
D3ToM: Decider-Guided Dynamic Token Merging for Accelerating Diffusion MLLMs
Shuochen Chang (Shanghai Jiao Tong University), Li Niu (Shanghai Jiao Tong University)
Computational EfficiencyTransformerVision Language ModelDiffusion modelMultimodalityBenchmark
🎯 What it does: Proposed D³ToM, a dynamic visual token merging method based on decoder decisions, to accelerate inference in diffusion multimodal large language models (Diffusion MLLMs).
DA-DFGAS:Differentiable Federated Graph Neural Architecture Search with Distribution-Aware Attentive Aggregation
Zhaowei Liu (Yantai University), Dong Yang (Georgia State University)
Federated LearningNeural Architecture SearchGraph Neural NetworkImageGraph
🎯 What it does: Proposed a federated graph neural network differentiable architecture search framework DA-DFGAS, combining distribution-aware self-attention aggregation and dual objectives to achieve unified training of client-specific personalized models and global consistency.
DANCE: Density-agnostic and Class-aware Network for Point Cloud Completion
Da-Yeong Kim (Chonnam National University), Yeong-Jun Cho (Chonnam National University)
RestorationGenerationTransformerNeural Radiance FieldPoint Cloud
🎯 What it does: Proposed a density-agnostic, class-aware point cloud completion framework called DANCE, which can complete missing regions under any sparse input while preserving existing geometry;
DanceHA: A Multi-Agent Framework for Document-Level Aspect-Based Sentiment Analysis
Lei Wang (Temple University), Eduard Dragut (Independent Researcher)
ClassificationKnowledge DistillationTransformerLarge Language ModelAgentic AITextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: This paper proposes a multi-agent framework called DanceHA for handling the document-level aspect-based sentiment intensity analysis (ABSIA) task, with a particular focus on informal writing styles;
DANS-KGC: Diffusion Based Adaptive Negative Sampling for Knowledge Graph Completion
Haoning Li (Northwestern Polytechnical University), Qinghua Huang (Northwestern Polytechnical University)
Representation LearningDiffusion modelContrastive LearningGraph
🎯 What it does: This paper proposes an adaptive negative sampling method called DANS-KGC based on diffusion models, aiming to improve the performance of knowledge graph completion.
DAPE: Harmonizing Content-Position Encoding for Versatile Dense Visual Prediction
Xiuquan Hou (Xi'an Jiaotong University), Shaoyi Du (Zhejiang University)
Object DetectionSegmentationTransformerImageBenchmark
🎯 What it does: Propose a unified DETR framework DAPE, which achieves co-optimization of content and position encoding through Shifted Query Sampler and Low-Rank Position Encoder, applicable to object detection, instance segmentation, and few-shot detection.
DAPointMamba: Domain Adaptive Point Mamba for Point Cloud Completion
Yinghui Li (Deakin University), Xuequan Lu (University of Western Australia)
RestorationDomain AdaptationPoint Cloud
🎯 What it does: Proposed the DAPointMamba framework to achieve unsupervised domain adaptation for cross-domain point cloud completion;
DAPrompt: Dual Alignment Prompt of Structure and Semantics for Few-shot Graph Learning
Lifan Jiang, Shenglin Ben (Zhejiang University)
ClassificationRepresentation LearningMeta LearningGraph Neural NetworkPrompt EngineeringContrastive LearningGraph
🎯 What it does: Propose a dual-alignment prompt framework called DAPrompt for few-shot graph learning in heterophilic graph environments, addressing issues of graph structural noise and semantic inconsistency.
DarkFarseer: Robust Spatio-Temporal Kriging Under Graph Sparsity and Noise
Zhuoxuan Liang, Moustafa Youssef (American University in Cairo)
Representation LearningGraph Neural NetworkContrastive LearningGraphTime Series
🎯 What it does: This paper proposes the DarkFarseer framework to improve the inductive spatiotemporal kriging (ISK) method based on graph neural networks, addressing issues such as insufficient spatiotemporal feature capture, graph sparsity, and noise in virtual node inference.
DARLING: Dual Hypergraph-Enhanced Curriculum-Guided Graph Structure Learning for Node Classification
Guangkai Wu, Zhongying Zhao (Shandong University of Science and Technology)
ClassificationGraph Neural NetworkGraph
🎯 What it does: Proposed a graph structure learning framework named DARLING, aiming to enhance node classification effectiveness through adaptive curriculum learning, dual hypergraph similarity learning, and multi-view graph structure enhancement.
Data Complexity of Querying Description Logic Knowledge Bases Under Cost-Based Semantics
Meghyn Bienvenu (Université de Bordeaux), Quentin Manière (Leipzig University)
🎯 What it does: The study investigates the data complexity of querying inconsistent weighted description logic knowledge bases under cost-based semantics, providing an upper bound of Δp₂ for optimal cost semantics, improving and enhancing the NP/coNP lower bounds under fixed costs, and proving that AC⁰ complexity can be achieved via first-order rewritings in DL-Lite H bool.
Data Heterogeneity and Forgotten Labels in Split Federated Learning
Joana Tirana (University College Dublin), Nicolas Kourtellis (Telefónica Scientific Research)
Federated LearningConvolutional Neural NetworkMixture of ExpertsImage
🎯 What it does: Studied the catastrophic forgetting (CF) phenomenon caused by data heterogeneity in split federated learning (SFL), and proposed the Hydra method to alleviate this issue.
Data-Centric Sequential Recommendation with Relation-Augmented Generation
Yichen Li (Huazhong University of Science and Technology), Ruixuan Li (Huazhong University of Science and Technology)
Recommendation SystemData-Centric LearningTransformerLarge Language ModelGraphSequential
🎯 What it does: Propose the RaSR framework, which utilizes multi-relational graphs (co-occurrence, temporal, semantic) to augment and regenerate sequential recommendation data, thereby improving data quality and enhancing the performance of subsequent models;
DAVID: Dual-stage Adaptive Vision-text Integrated Decoupling for Multimodal KV Cache Eviction
Yifeng Gu (South China University of Technology), Xiangmin Xu (Foshan University)
OptimizationComputational EfficiencyTransformerLarge Language ModelVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Proposed DAVID, a two-stage adaptive KV cache eviction strategy for multimodal large language models, dynamically distinguishing the separation and fusion phases of visual and text modalities, significantly reducing KV cache occupancy and improving inference speed.
DAWN: Distributed LLM Multi-Agent Workflow Synthesis
Guancheng Wan (Wuhan University), Wenke Huang (Princeton University)
Federated LearningTransformerLarge Language ModelReinforcement LearningAgentic AITextBenchmark
🎯 What it does: Designed and implemented a distributed LLM multi-agent workflow synthesis framework called DAWN, capable of generating collaborative communication topologies under strict privacy and distribution difference environments.
DBGroup: Dual-Branch Point Grouping for Weakly Supervised 3D Semantic Instance Segmentation
Xuexun Liu (Shenzhen University), Xu Wang (Meituan)
SegmentationConvolutional Neural NetworkVision Language ModelPoint Cloud
🎯 What it does: Propose DBGroup, a weakly supervised 3D instance segmentation framework based on scene-level labels, which generates pseudo labels through dual-branch point grouping and achieves segmentation via multi-round self-training.
DC-SPAN: A Dual Contrastive Attention Network for Multi-View Clustering
Jingyi Chen (National University of Defense Technology), Yibo Han (National University of Defense Technology)
Representation LearningAuto EncoderContrastive LearningMultimodality
🎯 What it does: Propose a multi-view clustering method based on the dual contrastive attention network (DC-SPAN), which eliminates feature entanglement by explicitly separating shared and private latent spaces, and generates more discriminative clustering embeddings through attention fusion.
DCA-LUT: Deep Chromatic Alignment with 5D LUT for Purple Fringing Removal
Jialang Lu (Hubei University), Zhuoran Zheng (Shandong Normal University)
RestorationData SynthesisConvolutional Neural NetworkImageVideo
🎯 What it does: Proposed a deep learning-based purple edge suppression framework called DCA-LUT, which first maps RGB images to a chromatic aberration space via a color-aware coordinate transformation, then performs local brightness channel restoration using a direction-aware 5D LUT, and finally applies a 1D LUT for global color correction, completely eliminating purple halos caused by camera long-range chromatic aberration.
DCAC: Dynamic Class-Aware Cache Creates Stronger Out-of-Distribution Detectors
Yanqi Wu (Sun Yat-sen University), Ruixuan Wang (Sun Yat-sen University)
Anomaly DetectionImageVideoBenchmark
🎯 What it does: Propose a training-free, dynamic class-aware cache (DCAC) during testing, which enhances out-of-distribution (OOD) detection by collecting high-entropy samples and leveraging their visual features to calibrate model outputs.
DCHO: A Decomposition–Composition Framework for Predicting Higher-Order Brain Connectivity to Enhance Diverse Downstream Applications
Weibin Li (Southern University of Science and Technology), Quanying Liu (Southern University of Science and Technology)
Graph Neural NetworkTransformerBiomedical DataMagnetic Resonance Imaging
🎯 What it does: Proposed the DCHO framework for modeling and predicting the temporal evolution of higher-order brain connectivity, decomposing the prediction task into two subtasks: higher-order connectivity inference and latent trajectory prediction.
DcMatch: Unsupervised Multi-Shape Matching with Dual-Level Consistency
Tianwei Ye (Wuhan University), Xiaoguang Mei (Wuhan University)
Representation LearningGraph Neural NetworkDiffusion modelPoint CloudMesh
🎯 What it does: Propose an unsupervised multi-shape matching framework DcMatch, which achieves more accurate correspondences by leveraging dual-layer consistency.
DCMM-Transformer: Degree-Corrected Mixed-Membership Attention for Medical Imaging
Huimin Cheng (Boston University), Ping Ma (University of Texas at Arlington)
ClassificationTransformerImageBiomedical Data
🎯 What it does: Propose DCMM-Transformer, which integrates the Degree-Corrected Mixed-Membership model into the self-attention mechanism of ViT, using additive bias to enhance medical image classification.
DcSplat: Dual-Constraint Human Gaussian Splatting with Latent Multi-View Consistency
Tengfei Xiao (Xidian University), Mingyang Zhang (Xidian University)
GenerationData SynthesisPose EstimationConvolutional Neural NetworkTransformerNeural Radiance FieldGaussian SplattingImage
🎯 What it does: Propose a single-view dual-constrained human Gaussian splatting framework called DcSplat, which generates high-quality new perspective portraits from a single image.
DCTR: Dual-Constraint Subgraph Optimization for Knowledge Graph-based Retrieval-Augmented Generation
Yukun Cao (Shanghai University of Electric Power), Lisheng Wang (University of Technology Sydney)
RetrievalOptimizationLarge Language ModelGraphRetrieval-Augmented Generation
🎯 What it does: This paper proposes a dual-constrained subgraph optimization framework called DCTR to improve the quality of subgraphs in knowledge graph retrieval-augmented generation.
De Novo Molecular Generation from Mass Spectra via Many-Body Enhanced Diffusion
Xichen Sun (Sun Yat-sen University), Yuedong Yang (Sun Yat-sen University)
GenerationDrug DiscoveryGraph Neural NetworkSupervised Fine-TuningDiffusion modelBiomedical DataBenchmark
🎯 What it does: Propose the MBGen framework to achieve de novo molecular structure generation from mass spectrometry data.
De-biased Natural Language Egocentric Task Verification via Prototypical Evidence Learning
Chong Liu (University of Electronic Science and Technology of China), Xing Xu (Tongji University)
ClassificationDomain AdaptationVision Language ModelContrastive LearningMultimodalityBenchmark
🎯 What it does: This paper proposes a framework named Prototypical Evidential Learning (PEL) to address domain differences between synthetic and real videos, as well as biases in binary classification decisions in natural language-oriented first-person task verification (NLETV).
De-collapsing User Intent: Adaptive Diffusion Augmentation with Mixture-of-Experts for Sequential Recommendation
Xiaoxi Cui (Beijing Institute of Technology), Xiangmin Zhou (RMIT University)
Recommendation SystemRecurrent Neural NetworkTransformerMixture of ExpertsDiffusion modelContrastive LearningSequential
🎯 What it does: This paper proposes the ADARec framework, which adaptively generates user intent hierarchies from extremely sparse user behavior sequences by leveraging the denoising trajectory of diffusion models, and enhances the accuracy of sequence recommendation through a hybrid expert network for decoding the hierarchy.
DEALT: LLM-driven Diversity-Enhanced Data Augmentation for Long-Tail Text Classification
Wayne Lu (Independent Researcher), Xiaoxi Cui (Takway.AI)
ClassificationTransformerLarge Language ModelTextChain-of-Thought
🎯 What it does: Proposed the DEALT framework, which leverages large language models (LLMs) for data augmentation in long-tail text classification, aiming to enhance the performance of underrepresented classes through diverse augmentation.
Debate over Mixed-knowledge: A Robust Multi-Agent Reasoning Framework for Incomplete Knowledge Graph Question Answering
Jilong Liu (Hefei University of Technology), Richang Hong (Hefei University of Technology)
Computational EfficiencyData-Centric LearningTransformerLarge Language ModelAgentic AITextGraphBenchmarkRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose the DoM (Debate over Mixed-knowledge) framework, which dynamically integrates structured knowledge graphs and unstructured text through multi-agent debate (KG Agent, RAG Agent, and Judge Agent), enabling question-answering reasoning on incomplete knowledge graphs; and constructs a more realistic IKGWQ dataset.
Debiased Cognitive Diagnosis: A Contrastive Counterfactual Modeling Method via Variational Autoencoder
Shangshang Yang (Anhui University), Xingyi Zhang (Anhui University)
Auto EncoderContrastive LearningTabular
🎯 What it does: Propose a debiasing framework DBCD in cognitive diagnosis, eliminating bias caused by students' selective responses by aligning the predictive distributions of factual and counterfactual data.
Debiased Dual-Invariant Defense for Adversarially Robust Person Re-Identification
Yuhang Zhou (Harbin Institute of Technology), Leo Yu Zhang (Griffith University)
RecognitionAdversarial AttackMeta LearningConvolutional Neural NetworkDiffusion modelImage
🎯 What it does: Proposing a bias-removing double-invariant defense framework for adversarial attacks in person re-identification.
Debiased Multiplex Tokenizer for Efficient Map-Free Visual Relocalization
Wenshuai Wang (Peking University), Runwei Ding (Pengcheng Laboratory)
Pose EstimationComputational EfficiencyTransformerImage
🎯 What it does: Propose Debiased Multiplex Tokenizer (DeMT) to achieve efficient map-free visual localization.
Debiasing Diffusion Priors via 3D Attention for Consistent Gaussian Splatting
Shilong Jin (Nanjing University of Information Science and Technology), Yuan Zhou (Nanjing University of Information Science and Technology)
GenerationData SynthesisLarge Language ModelDiffusion modelGaussian SplattingTextMultimodalityPoint Cloud
🎯 What it does: This paper proposes a framework called TD-Attn, which addresses the multi-view inconsistency (Janus) problem in text-driven 3D generation and editing by introducing 3D attention guidance and hierarchical attention modulation into 3D Gaussian Splatting (3DGS).
Deblur4DGS: 4D Gaussian Splatting from Blurry Monocular Video
Renlong Wu (Harbin Institute of Technology), Wangmeng Zuo (Harbin Institute of Technology)
RestorationGaussian SplattingOptical FlowVideo
🎯 What it does: This paper proposes a framework called Deblur4DGS based on 4D Gaussian Splatting, which is used to recover high-quality 4D (three-dimensional space + time) models from monocular videos with motion blur.
Decentralized Non-convex Stochastic Optimization with Heterogeneous Variance
Hongxu Chen (Fudan University), Luo Luo (Fudan University)
OptimizationFederated LearningImageTabular
🎯 What it does: Study the impact of node heteroscedasticity on sample complexity in distributed non-convex stochastic optimization, proposing node-specific sampling-based D-NSS and its variant D-NSS-VR, achieving optimal sample complexity and improved variance reduction schemes respectively.
Decentralized Online Convex Optimization with Unknown Feedback Delays
Hao Qiu (Universitá degli Studi di Milano), Juliette Achddou (Université de Lille)
OptimizationFederated LearningTabular
🎯 What it does: Proposes a decentralized online convex optimization algorithm applicable to unknown, time-varying, and heterogeneous agent feedback delays, capable of adaptively adjusting the learning rate and achieving near-optimal regret upper bounds.
Decidable Multi-agent Epistemic Planning: A Situation Calculus Approach
Qihui Feng (RWTH Aachen University), Gerhard Lakemeyer (RWTH Aachen University)
Optimization
🎯 What it does: This paper proposes a modal logic framework based on situation calculus to formulate and solve multi-agent knowledge/belief planning (MEP) problems, and presents regression methods and optimal planning algorithms;
Decision-Driven Orthogonal Learning with Complementary Feature Mining for Robust Synthetic Image Detection
Kai Li (Sun Yat-sen University), Wenqi Ren (Sun Yat-sen University)
Anomaly DetectionConvolutional Neural NetworkTransformerImage
🎯 What it does: Investigated the robustness of synthetic image detection under compression inconsistencies in social networks, and proposed a decision-driven orthogonal constraint and low/high-frequency interaction framework to enhance detection performance under compression.
Decoding with Structured Awareness: Integrating Directional, Frequency-Spatial, and Structural Attention for Medical Image Segmentation
Fan Zhang (Shandong Technology and Business University), Hua Wang (Ludong University)
SegmentationConvolutional Neural NetworkTransformerBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: Proposed a novel medical image segmentation decoder framework comprising three major modules: ACFA, TFFA, and SMMM;
Decompose and Attribute: Boosting Generalizable Open-Set Object Detection via Objectness Score
Yuxuan Yuan (Xiamen University), Xinghao Ding (Xiamen University)
Object DetectionDomain AdaptationImage
🎯 What it does: Propose a unified framework DOAT that utilizes wavelet frequency decomposition to separate style and semantics, and achieves generalization detection for unknown categories and domain drift through low-frequency style expansion and high-frequency object attribution.
Decompose and Conquer: Compositional Reasoning for Zero-Shot Temporal Action Localization
Haoyu Tang (Shandong University), Yupeng Hu (Shandong University)
RecognitionTransformerLarge Language ModelVision Language ModelVideoMultimodality
🎯 What it does: Proposed a training-free stagewise decomposition and synthesis framework named CASCADE for zero-shot temporal action localization.
Decomposing and Composing: Towards Efficient Vision-Language Continual Learning via Rank-1 Expert Pool in a Single LoRA
Zhan Fa (Nanjing University), Yinghuan Shi (Nanjing University)
Computational EfficiencyRepresentation LearningTransformerSupervised Fine-TuningMixture of ExpertsContrastive LearningMultimodality
🎯 What it does: In the continual learning scenario, we propose decomposing a single LoRA module into a dynamically combinable Rank-1 expert pool, using CLS semantics to guide expert selection, and designing an Activation-Guided Orthogonal (AGO) loss to reduce task interference.
Decomposing Direct and Indirect Biases in Linear Models Under Demographic Parity Constraint
Bertille Tierny (Milliman France), Francois Hu (Université du Québec à Montréal)
OptimizationExplainability and InterpretabilityTabular
🎯 What it does: The study decomposes direct and indirect bias in linear models under demographic parity constraints and proposes a post-processing analytical fair linear regression method.