AAAI 2026 Papers — Page 2
AAAI Conference on Artificial Intelligence · 4149 papers
Adaptive Agent Selection and Interaction Network for Image-to-Point Cloud Registration
Zhixin Cheng (University Of Science And Technology Of China), Tianzhu Zhang (University Of Science And Technology Of China)
Pose EstimationTransformerReinforcement LearningAgentic AIContrastive LearningImageMultimodalityPoint Cloud
🎯 What it does: Proposed a cross-modal registration framework A2SI based on adaptive proxy selection and interaction, achieving efficient alignment between images and point clouds.
Adaptive and Asymptotic Mean-based Subclass Discriminant Analysis
Yuzhe Feng (Xiamen University), Feiping Nie (Northwest Polytechnical University)
ClassificationImageBenchmark
🎯 What it does: Designed and implemented an unconstrained proxy subclass discriminant analysis method called AASDA for learning subclass means and obtaining discriminative subspaces in multi-subclass scenarios.
Adaptive and Context-rich Generative Self-supervised Learning on Graphs
Yijun Tian (University of Connecticut), Nitesh V Chawla
Representation LearningGraph Neural NetworkReinforcement LearningAuto EncoderContrastive LearningGraph
🎯 What it does: Propose a novel graph self-supervised learning framework ACE-GSL, which comprehensively models node importance, global structure, semantic information, and reconstruction stability through four modules: adaptive masking, structural reconstruction, guided similarity, and consistency guarantee.
Adaptive Diffusion-based Augmentation for Recommendation
Na Li (Harbin Institute of Technology), Ying Ma (Harbin Institute of Technology)
Recommendation SystemDiffusion modelScore-based ModelGraphSequential
🎯 What it does: Proposed ADAR, a controllable negative sampling enhancement module based on diffusion models, which generates high-quality negative samples by progressively corroding positive samples during the diffusion process, thereby improving the performance of recommendation systems.
Adaptive Dynamic Dehazing via Instruction-Driven and Task-Feedback Closed-Loop Optimization for Diverse Downstream Task Adaptation
Yafei Zhang (Kunming University of Science and Technology), Yu Liu (Hefei University of Technology)
RestorationObject DetectionSegmentationDepth EstimationTransformerLarge Language ModelVision Language ModelContrastive LearningImageText
🎯 What it does: Proposed a closed-loop optimization based adaptive dynamic dehazing framework that can adjust dehazing results in real-time without retraining during inference through bidirectional guidance from task feedback and text instructions;
Adaptive Evidential Learning for Temporal-Semantic Robustness in Moment Retrieval
Haojian Huang (Hong Kong University of Science and Technology), Zhongjiang He (China Telecom)
RetrievalTransformerVision Language ModelVideoText
🎯 What it does: Propose the DEMR model, introducing deep evidence regression in temporal-semantic retrieval, and significantly improving uncertainty estimation and retrieval accuracy through cross-modal fusion, query reconstruction, and geometric regularization.
Adaptive Evolutionary Fusion for Multi-View Clustering
Yunxiao Zhao (Shanxi University), Xian Yang (University of Manchester)
OptimizationRepresentation LearningData-Centric LearningGraph Neural NetworkContrastive LearningMultimodality
🎯 What it does: Propose an adaptive hierarchical fusion framework AEF-MVC based on unsupervised evolutionary algorithms, which uses a tree structure to fuse multi-view features to enhance clustering performance.
Adaptive Fidelity Estimation for Quantum Programs with Graph-Guided Noise Awareness
Tingting Li (Shanghai Qi Zhi Institute), Jianwei Yin (Zhejiang University)
GraphPhysics Related
🎯 What it does: QuFid proposes an adaptive quantum program fidelity estimation framework that dynamically allocates measurement budgets based on circuit structure and runtime noise.
Adaptive Frequency Pathways for Spatiotemporal Forecasting
Yanjun Qin (Xinjiang University), Xiaoming Tao (Tsinghua University)
Graph Neural NetworkTime Series
🎯 What it does: Proposed the AdaFre model, which utilizes multi-frequency decomposition and frequency-specific spatial embeddings to achieve adaptive frequency paths for spatiotemporal prediction.
Adaptive Graph Attention Based Discrete Hashing for Incomplete Cross-modal Retrieval
Shuang Zhang (Hebei Normal University), Pengtao Lv (Henan University of Technology)
RetrievalGraph Neural NetworkMultimodality
🎯 What it does: Designed the Adaptive Graph Attention-Based Discrete Hashing (AGADH) framework, leveraging mask completion strategy, graph attention network (GAT) encoding-decoding, and adaptive weight fusion to generate high-quality binary hash codes for incomplete cross-modal data, improving retrieval efficiency and semantic consistency.
Adaptive Hallucination Alleviation in Multimodal Large Language Models: From Strategic Data Selection to Severity-Guided Training
Yuanyi Xu, Wei Wang (Fudan University)
OptimizationData-Centric LearningReinforcement Learning from Human FeedbackLarge Language ModelReinforcement LearningContrastive LearningImageTextMultimodality
🎯 What it does: By screening image samples prone to severe hallucinations and calculating hallucination severity, HD-DPO performs fine-grained preference optimization on multimodal large language models, significantly reducing hallucination rates while enhancing multimodal reasoning performance.
Adaptive Hyperbolic Kernels: Modulated Embedding in de Branges-Rovnyak Spaces
Leping Si (Southeast University), Pengfei Fang (Southeast University)
ClassificationRetrievalRepresentation LearningImageText
🎯 What it does: This paper proposes an adaptive hyperbolic kernel based on curvature-adjustable de Branges-Rovnyak spaces, utilizing isometric mapping to embed hyperbolic spaces with arbitrary curvature into RKHS, and dynamically modulating hyperbolic features through learnable multipliers, further constructing adaptive hyperbolic kernels applicable to linear, polynomial, RBF, Laplacian, and the novel AHRad kernel;
Adaptive Initial Residual Connections for GNNs with Theoretical Guarantees
Mohammad Shirzadi (Australian National University), Ahad N. Zehmakan (Australian National University)
ClassificationRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: Propose Adaptive Initial Residual Connections (Adaptive IRC) to alleviate the over-smoothing problem in graph neural networks and enhance their expressiveness.
Adaptive LiDAR Scanning: Harnessing Temporal Cues for Efficient 3D Object Detection via Multi-Modal Fusion
Sara Shoouri (University of Michigan), Hun-Seok Kim (University of Michigan)
Object DetectionAutonomous DrivingComputational EfficiencyTransformerMultimodalityPoint Cloud
🎯 What it does: Propose a multi-modal perception framework that predicts based on historical information and adaptively controls LiDAR scan density;
Adaptive Momentum and EMA-weighted Modeling for Imbalanced Label Distribution Learning
Yongbiao Gao (Qilu University of Technology), Guohua Lv (Qilu University of Technology)
ClassificationOptimizationImageBenchmark
🎯 What it does: Proposed the AMEMA framework, which addresses the issues of gradient vanishing and optimization imbalance in imbalanced label distribution learning by splitting the label distribution into dominant and non-dominant branches, and applying exponential moving average (EMA) dynamic reweighting and adaptive momentum allocation on each branch.
Adaptive Morph-Patch Transformer for Aortic Vessel Segmentation
Zhenxi Zhang (Institute of Scientific Instrumentation, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences), Shoujun Zhou (Institute of Scientific Instrumentation, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences)
SegmentationTransformerBiomedical Data
🎯 What it does: Proposed an adaptive Morph-Patch Transformer (MPT) for aortic vessel segmentation, addressing the issues of vascular shape distortion and insufficient semantic hierarchy caused by the fixed-size rectangular patches in traditional Transformers.
Adaptive Piecewise Distillation for Efficient LiDAR Data Generation
Ruibo Li (Nanyang Technological University), Guosheng Lin (Nanyang Technological University)
Data SynthesisAutonomous DrivingKnowledge DistillationFlow-based ModelRectified FlowAuto EncoderMultimodalityPoint CloudOrdinary Differential Equation
🎯 What it does: Propose an adaptive segmented distillation method to efficiently generate LiDAR point clouds within four-step sampling, compatible with unconditional and multi-modal conditional generation.
Adaptive Riemannian Graph Neural Networks
Xudong Wang (Chinese University of Hong Kong), Jicong Fan (Chinese University of Hong Kong)
ClassificationRepresentation LearningGraph Neural NetworkGraphBenchmark
🎯 What it does: This paper proposes an Adaptive Riemannian Graph Neural Network (ARGNN), which achieves continuous variable geometric embeddings for graph data by learning node-level differentiable diagonal metric tensors.
Adaptive Theory of Mind for LLM-based Multi-Agent Coordination
Chunjiang Mu (Northwestern Polytechnical University), Shuyue Hu (Shanghai Artificial Intelligence Laboratory)
TransformerLarge Language ModelAgentic AITabular
🎯 What it does: Designed and implemented an adaptive Theory of Mind (A-ToM) agent that dynamically estimates and matches the ToM levels of partners using large language models, thereby enhancing zero-shot multi-agent coordination performance.
Adaptive-Learngene: Continual Expansion and Task-Aware Selection of Learngenes for Dynamic Environments
Shuxia Lin (Southeast University), Xin Geng (Southeast University)
ClassificationRecognitionSegmentationTransformerReinforcement LearningImage
🎯 What it does: Propose the Adaptive-Learngene framework, which supports continuous learning through an expandable ViT without accessing old data, and uses a Task-Adaptive Learngene Selector (TALS) to sparsely select the most relevant learngenes from a global learngene pool to initialize sub-models for each downstream task.
Adaptive-Smooth LiDAR-Camera Knowledge Distillation with Heterogeneous Fusion for Multi-View 3D Object Detection
Rui Zhao (Shenzhen University), Shijian Gao (Hong Kong University of Science and Technology)
Object DetectionAutonomous DrivingKnowledge DistillationContrastive LearningImageMultimodalityPoint CloudBenchmark
🎯 What it does: Propose a multi-view 3D object detection framework based on LiDAR-camera knowledge distillation, primarily introducing adaptive smooth distillation, heterogeneous fusion, and soft weighted response distillation to enhance the student model's spatial reasoning and semantic fusion capabilities.
AdaReason: Progressive Training of Multi-LoRA Adapters for Budget-Adaptive Language Reasoning Models
Jiacheng Wang (Xi'an Jiao Tong University), Jiacheng Liu (Shanghai Jiao Tong University)
Computational EfficiencyRepresentation LearningTransformerLarge Language ModelReinforcement LearningTextBenchmark
🎯 What it does: Under limited computational budgets, a language reasoning model with dynamically adjustable reasoning length was achieved by training a base model and a set of multi-objective low-rank adapters (LoRA). This method employs a progressively converging training strategy and enables zero-training budget switching during inference through adaptive merging of adapters.
AdaSpec: Adaptive Multilingual Speculative Decoding with Self-Synthesized Language-Aware Training and Vocabulary Simplification
Dinh-Truong Do (Japan Advanced Institute of Science and Technology), Le-Minh Nguyen (Japan Advanced Institute of Science and Technology)
Data SynthesisComputational EfficiencyTransformerLarge Language ModelTextBenchmark
🎯 What it does: Implemented ADASPEC—an adaptive speculative decoding framework for multilingual settings, dynamically adjusting the generator and vocabulary. It trains the generator for low-resource languages using self-generated instruction data and performs language-adaptive vocabulary simplification;
Addressing Polarization and Unfairness in Performative Prediction
Kun Jin (University of Michigan), Xueru Zhang (University of California, Santa Cruz)
OptimizationData-Centric LearningImageTabular
🎯 What it does: This paper studies the impact of data distribution shifts caused by model deployment under the Performative Prediction framework on fairness, and proposes a Fair-PS solution.
Advanced Black-Box Tuning of Large Language Models with Limited API Calls
Zhikang Xie (Fudan University), Cheng Jin (Fudan University)
Computational EfficiencyKnowledge DistillationLarge Language ModelText
🎯 What it does: This paper proposes a method that uses a Gaussian Process (GP) surrogate model to approximate the outputs of large language models under the premise of limited API calls, and uses this approximation to guide the training of the surrogate model, thereby achieving efficient black-box fine-tuning.
Advancing Multimodal Teacher Sentiment Analysis: The Large-Scale T-MED Dataset & the Effective AAM-TSA Model
Zhiyi Duan (Inner Mongolia University), QianLi Xing
ClassificationRecognitionTransformerLarge Language ModelVideoTextMultimodalityBenchmarkAudio
🎯 What it does: Constructed a large-scale teacher multimodal emotion analysis dataset named T-MED, and proposed the AAM-TSA model.
Advancing Protein Design via Multi-Agent Reinforcement Learning with Pareto-Based Collaborative Optimization
Mingming Zhu (Sun Yat-sen University), Yuedong Yang (Sun Yat-sen University)
OptimizationDrug DiscoveryTransformerLarge Language ModelReinforcement LearningAgentic AIBiomedical DataBenchmark
🎯 What it does: Propose the MAProt multi-agent framework, integrating structural models and protein language models for protein design;
Adversarial Attack on Black-Box Multi-Agent by Adaptive Perturbation
Jianming Chen (Institute of Software Chinese Academy of Sciences), Fanjiang Xu (Institute of Software Chinese Academy of Sciences)
Adversarial AttackReinforcement LearningGenerative Adversarial NetworkBenchmark
🎯 What it does: Proposes the AdapAM framework, enabling adaptive attacks in strict black-box multi-agent systems that can both select victim agents and induce them to perform malicious actions;
Adversarial Fair Incomplete Multi-View Clustering
Qianqian Wang, Quanxue Gao (Northwest A F University)
Representation LearningGenerative Adversarial NetworkContrastive LearningTabularFinance Related
🎯 What it does: Proposed an adversarial fair incomplete multi-view clustering framework named AFIMVC, which fills missing views by leveraging contextual information from complete views through cross-sample attention mechanisms, and achieves independence between clustering results and sensitive attributes via an adaptive fair decoupling module.
Adversarial Perturbation Shield: Preventing Concept Bleed-through in Continual Learning of Personalized Generative Models
Ziwen Lan, Miki Haseyama (Hokkaido University)
GenerationData SynthesisRepresentation LearningDiffusion modelImage
🎯 What it does: Propose a training strategy based on adversarial perturbations, which injects subtle perturbations into training images in continual learning to widen the semantic representations of different concepts in the latent space, thereby preventing concept leakage in personalized diffusion models;
AEDR: Training-Free AI-Generated Image Attribution via Autoencoder Double-Reconstruction
Chao Wang (University of Science and Technology of China), Kejiang Chen (University of Science and Technology of China)
ClassificationRecognitionAnomaly DetectionDiffusion modelAuto EncoderImage
🎯 What it does: This paper proposes a training-agnostic, non-intrusive AI image source attribution method called AEDR, which uses dual autoencoder reconstruction and employs the ratio of reconstruction loss for attribution judgment.
AerialFusion: Co-Motion-Driven Unified Registration and Fusion on Multi-modal Data Streams from Aerial View
Junhui Qiu (Huazhong University of Science and Technology), Jiaqi Gui (Huazhong University of Science and Technology)
Convolutional Neural NetworkGenerative Adversarial NetworkSimultaneous Localization and MappingMultimodality
🎯 What it does: Proposed a unified registration and fusion framework for multi-modal aerial visual streams based on collaborative motion, named AerialFusion, and constructed the EUM3D dataset.
AerialMind: Towards Referring Multi-Object Tracking in UAV Scenarios
Chenglizhao Chen (China University of Petroleum (East China)), Qing-Long Han (Shanghai Jiao Tong University)
Object TrackingTransformerLarge Language ModelVideoTextBenchmark
🎯 What it does: Proposed the AerialMind large-scale UAV scenario referential multi-object tracking (RMOT) dataset, and developed a semi-automated collaborative annotation framework COALA and a novel tracking method HETrack
AerialVLA: A Vision-Language-Action Model for Aerial Navigation with Online Dialogue
Jinyu Chen (Beihang University), Si Liu (Beihang University)
Autonomous DrivingRobotic IntelligenceTransformerLarge Language ModelVision-Language-Action ModelImageVideoTextMultimodalityBenchmark
🎯 What it does: Propose AerialVLA, a UAV vision-language navigation framework integrating active question answering and spatiotemporal history memory;
Affordance-Guided Coarse-to-Fine Exploration for Base Placement in Open-Vocabulary Mobile Manipulation
Tzu-Jung Lin (National Taiwan University), Winston H. Hsu (National Yang Ming Chiao Tung University)
OptimizationRobotic IntelligenceTransformerVision Language ModelImageMultimodality
🎯 What it does: Studied a zero-shot base placement method for open-vocabulary mobile manipulation (OVMM), achieving semantic and geometric balanced base selection through cross-modal projection and iterative optimization.
Affordance-R1: Reinforcement Learning for Generalizable Affordance Reasoning in Multimodal Large Language Models
Hanqing Wang (Hong Kong University of Science and Technology), Yuexin Ma (ShanghaiTech University)
SegmentationExplainability and InterpretabilityLarge Language ModelReinforcement LearningVision Language ModelMultimodalityChain-of-Thought
🎯 What it does: Developed a multimodal large language model framework named Affordance-R1 based on reinforcement learning, capable of reasoning and locating operable regions in images, and generating interpretable thought chains during inference.
Agent Journey Beyond RGB: Hierarchical Semantic-Spatial Representation Enrichment for Vision-and-Language Navigation
Xuesong Zhang (Hefei University of Technology), Zhenzhen Hu (Shanghai Jiao Tong University)
Representation LearningGraph Neural NetworkTransformerReinforcement LearningVision Language ModelVision-Language-Action ModelContrastive LearningImageTextPoint Cloud
🎯 What it does: Propose a hierarchical semantic-spatial representation enhancement framework (SUSA), which improves environmental perception and instruction alignment in vision-language navigation through a text semantic view and a deep exploration graph.
Agent-SAMA: State-Aware Mobile Assistant
Linqiang Guo (Concordia University), Yang Wang (Concordia University)
TransformerLarge Language ModelAgentic AITextSequential
🎯 What it does: Designed Agent-SAMA, a multi-agent mobile GUI agent based on finite state machine (FSM) modeling, capable of real-time construction and utilization of FSM for planning, execution, verification, and recovery during task execution, and storing the FSM and action sequences as long-term memory upon completion.
AgentCDM: Enhancing Multi-Agent Collaborative Decision-Making via ACH-Inspired Structured Reasoning
Xuyang Zhao (Nankai University), Qicheng Li (Nankai University)
TransformerLarge Language ModelReinforcement LearningTextBenchmarkChain-of-Thought
🎯 What it does: Propose the AgentCDM framework, leveraging two-stage training to enhance collaborative decision-making capabilities in LLM-based multi-agent systems.
Agentic Design Review System
Sayan Nag (Adobe Research), Balaji Vasan Srinivasan (Adobe Research)
Large Language ModelAgentic AIPrompt EngineeringVision Language ModelImageMultimodalityBenchmarkRetrieval-Augmented Generation
🎯 What it does: Proposed a multi-agent collaborative Agentic Design Review System (Agentic-DRS), which leverages a meta-agent to schedule static and dynamic review agents, utilizing graph matching and structured descriptions to achieve multi-dimensional evaluation and actionable feedback for flat design.
AgentMental: An Interactive Multi-Agent Framework for Explainable and Adaptive Mental Health Assessment
Jinpeng Hu (Hefei University of Technology), Dan Guo (Hefei University of Technology)
Explainability and InterpretabilityLarge Language ModelAgentic AIText
🎯 What it does: Designed and implemented a multi-agent framework called AgentMental, using interactive doctor-patient dialogues to simulate mental health assessments.
AgentODRL: A Large Language Model-based Multi-agent System for ODRL Generation
Wanle Zhong (Renmin University of China), Xiaoyong Du (Renmin University of China)
GenerationData SynthesisTransformerLarge Language ModelSupervised Fine-TuningAgentic AIText
🎯 What it does: Proposed AgentODRL, a multi-agent system that automatically converts natural language data into ODRL format using rules.
AgentSense: Virtual Sensor Data Generation Using LLM Agents in Simulated Home Environments
Zikang Leng (Georgia Institute of Technology), Thomas Plötz
Data SynthesisLarge Language ModelSupervised Fine-TuningAgentic AIPrompt EngineeringTime SeriesRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: This paper proposes AgentSense, which generates structured, privacy-preserving environmental sensor data by having LLM-driven virtual agents perform diversified daily activities in an extended VirtualHome simulation environment, used for training and pretraining human activity recognition (HAR) models.
AgentSwift: Efficient LLM Agent Design via Value-Guided Hierarchical Search
Yu Li (Tsinghua University), Fengli Xu (Tsinghua University)
OptimizationNeural Architecture SearchLarge Language ModelAgentic AITextBenchmark
🎯 What it does: Propose the AgentSwift framework to achieve automated design of LLM agents, jointly optimizing workflows and pluggable functional modules (memory, tool usage, planning), and employing a value model and uncertainty-guided hierarchical MCTS for efficient search.
Aggregate-Combine-Readout GNNs Can Express Logical Classifiers Beyond the Logic C2
Stan P Hauke (King's College London), Przemysław Andrzej Wałęga (Queen Mary University of London)
Representation LearningGraph Neural Network
🎯 What it does: This paper theoretically proves that Aggregation-Composition-Reading (ACR-GNN) can express a first-order node classifier stronger than two-variable counting logic C^2, addressing the open question proposed by Barceló et al. regarding the expressive power between ACR-GNN and C^2.
Aggregating Diverse Cue Experts for AI-Generated Image Detection
Lei Tan (National University of Singapore), Robby T. Tan (National University of Singapore)
Anomaly DetectionTransformerMixture of ExpertsVision Language ModelImageBenchmark
🎯 What it does: Propose a multi-clue aggregation network (MCAN) that integrates three signals: original images, high-frequency components, and chromatic inconsistency (CI). The network employs a Mixture-of-Encoder Adapter to dynamically aggregate multimodal features for detecting AI-generated images.
AgriEval: A Comprehensive Chinese Agricultural Benchmark for Large Language Models
Lian Yan (Harbin Institute of Technology), Jingchi Jiang (Harbin Institute of Technology)
TransformerLarge Language ModelTextBenchmarkAgriculture RelatedRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose AgriEval, a large-scale Chinese agricultural domain evaluation benchmark covering 6 major categories and 29 subcategories, with a total of 14,697 multiple-choice questions and 2,167 open-ended questions.
AHAMask: Reliable Task Specification for Large Audio Language Models Without Instructions
Yiwei Guo (Shanghai Jiao Tong University), Kai Yu (Shanghai Jiao Tong University)
RecognitionTransformerLarge Language ModelAudio
🎯 What it does: This paper proposes using 'AHAMask' to specify task functions without instructions by masking partial attention heads in the Transformer decoder of large audio language models;
AHAN: Asymmetric Hierarchical Attention Network for Identical Twin Face Verification
Hoang-Nhat Nguyen (Hanoi University of Science and Technology)
RecognitionTransformerContrastive LearningImage
🎯 What it does: Proposes a novel network architecture named AHAN for facial verification of monozygotic twins with identical genes, achieving excellent performance in ultra-fine-grained identification tasks involving monozygotic twins.
AI-Salesman: Towards Reliable Large Language Model Driven Telemarketing
Qingyu Zhang (Chinese Information Processing Laboratory, Institute of Software, Chinese Academy of Sciences), Xingxing Wang (Independent Researcher)
Recommendation SystemTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningAgentic AITextFinance Related
🎯 What it does: This paper proposes an end-to-end dialogue framework called AI-Salesman tailored for telemarketing scenarios, addressing three major challenges in persuasive dialogues: user satisfaction, authenticity, and personalization.
AIM: Manifold-based Data Filtering for Representation Finetuning
Qing Li (Shanghai Jiao Tong University), Xingchun Diao (Advanced Institute of Big Data)
Representation LearningData-Centric LearningTransformerSupervised Fine-TuningText
🎯 What it does: Proposed a geometry consistency data filtering method called AIM based on ReFT, which filters training noise samples by leveraging the alignment degree between intervention vectors and low-dimensional subspaces
AIR-DR: Adaptive Image Retargeting with Instance Relocation and Dual-guidance Repainting
Zhitong Dong (Southeast University), Hao Chen (Alibaba Group)
GenerationTransformerDiffusion modelImage
🎯 What it does: Proposes the AIR-DR framework, treating image aspect ratio adaptation as instance-level relayout to avoid pixel-level deletion, and combines adaptive decision-making with bidirectional guided background repainting to achieve high-quality, detail-preserving retargeting.
AirCopBench: A Benchmark for Multi-drone Collaborative Embodied Perception and Reasoning
Jirong Zha (Tsinghua University), Xinlei Chen (Tsinghua University)
Large Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Constructed the AirCopBench benchmark, containing 2.9k multi-perspective UAV images and 14.6k semantic VQA questions, to evaluate multi-UAV collaborative perception and reasoning capabilities
AirDDE: Multifactor Neural Delay Differential Equations for Air Quality Forecasting
Binqing Wu (Zhejiang University), Ling Chen (Zhejiang University)
Recurrent Neural NetworkTabularTime SeriesPhysics RelatedOrdinary Differential Equation
🎯 What it does: This paper proposes AirDDE, which is based on a neural delay differential equation framework, integrating multi-factor memory-enhanced attention and physics-guided delay evolution functions to achieve continuous-time air quality prediction.
AirWino: Optimized Winograd Convolution for Accelerating CNN Inference on ARMv8 Processors
Haoyuan Gui (University of Chinese Academy of Sciences), Huiyuan Li (Chinese Academy of Sciences)
Computational EfficiencyConvolutional Neural NetworkBenchmark
🎯 What it does: This paper proposes AirWino, a high-performance Winograd convolution implementation targeting ARMv8-A processors, covering 2D and 3D convolutions, FP32/FP16 precision, 3×3 and 5×5 kernels, and achieving significant acceleration through multiple microkernels, double buffering, and custom data layouts.
ALEX:A Light Editing-knowledge Extractor
Minghu Wang (Hebei Normal University), Hongxia Xu (Hebei Normal University)
RetrievalComputational EfficiencyTransformerLarge Language ModelTextChain-of-Thought
🎯 What it does: Propose a lightweight knowledge editing framework ALEX, which improves knowledge retrieval and reasoning in multi-hop QA using hierarchical memory structure, inferential query synthesis, and dynamic evidence adjudication.
Alfa: Attentive Low-Rank Filter Adaptation for Structure-Aware Cross-Domain Personalized Gaze Estimation
He-Yen Hsieh (Harvard University), H. T. Kung
Pose EstimationDomain AdaptationTransformerLarge Language ModelImageText
🎯 What it does: This paper proposes an attention-based low-rank filter adaptation method (Alfa), which achieves test-time personalization fine-tuning with a few unlabeled samples by performing SVD decomposition on pre-trained model weights and reweighting the spatial structure using multi-head attention, thereby improving the accuracy of cross-domain gaze estimation.
Algorithms for Structured Elections Under Thiele Voting Rules
Alexandra Lassota (Eindhoven University of Technology), Krzysztof Sornat (AGH University of Science and Technology)
Optimization
🎯 What it does: Studied the computational complexity of determining winners in approval voting committees based on Thiele's rule, proposed structural characterizations, and provided an FPT algorithm on the VI domain. It solved the polynomial-time algorithm for ∆C=2 and presented an FPT algorithm parameterized by the total score.
Align When They Want, Complement When They Need! Human-Centered Ensembles for Adaptive Human-AI Collaboration
Hasan Amin (Purdue University), Rajiv Khanna (Purdue University)
Convolutional Neural NetworkSupervised Fine-TuningImageTabularBenchmark
🎯 What it does: Propose a human-centric adaptive AI integration framework that can switch between alignment models and complementary models based on context in human-AI collaboration to achieve a balance between alignment and complementarity;
Align³GR: Unified Multi-Level Alignment for LLM-based Generative Recommendation
Wencai Ye (Kuaishou Technology), Peng Jiang (Kuaishou Technology)
Recommendation SystemTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextSequential
🎯 What it does: Propose a unified multi-level alignment framework, Align GR 3, converting LLM into a generative recommendation system by integrating alignment at the token-level, behavior modeling layer, and preference layer;
AlignCVC: Aligning Cross-View Consistency for Single-Image-to-3D Generation
Xinyue Liang (Hong Kong Polytechnic University), Lei Zhang (Hong Kong Polytechnic University)
GenerationData SynthesisSupervised Fine-TuningDiffusion modelScore-based ModelImageMesh
🎯 What it does: Propose the AlignCVC framework, which jointly post-trains multi-view generation and reconstruction models, enhances cross-view consistency through distribution alignment, and compresses the 3D-aware sampling steps to 4 steps.
Aligning Cross-View Visual Geometries in LVLMs Through Human-Like Reasoning Learning
Yuming Qiao (OPPO Research Institute), Xudong Zhang (OPPO Research Institute)
Object DetectionObject TrackingSegmentationData SynthesisPose EstimationDepth EstimationTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningVision Language ModelVideoTextMultimodalityBenchmarkChain-of-Thought
🎯 What it does: Proposes CVVG-Reasoner, a LVLM capable of cross-view spatial reasoning, achieving unified cross-view spatial understanding through human-like two-tier reasoning (low-level single-view spatial perception and high-level multi-view geometric alignment); simultaneously constructs an expandable MV3DSR data generation pipeline and MV3DSR-Bench evaluation benchmark; and significantly enhances the model's cross-view reasoning ability through three-stage training (SFT+human-like reasoning data→RL).
Aligning the True Semantics: Constrained Decoupling and Distribution Sampling for Cross-Modal Alignment
Xiang Ma (Shandong University), Caiming Zhang (Shandong University)
RetrievalRepresentation LearningConvolutional Neural NetworkTransformerContrastive LearningImageTextMultimodality
🎯 What it does: Proposed a cross-modal alignment method (CDDS) based on constraint decoupling and distribution sampling, which splits visual and textual embeddings into semantic and modal components, aligning only the semantic part;
AlignTrack: Top-Down Spatiotemporal Resolution Alignment for RGB-Event Visual Tracking
Chuanyu Sun (Dalian University Of Technology), Xin Yang (Beijing University Of Technology)
Object TrackingTransformerContrastive LearningMultimodality
🎯 What it does: Proposed AlignTrack, an RGB-Event visual tracking framework based on Top-Down Alignment, which can effectively align features from two modalities under spatiotemporal misalignment conditions and achieve high-precision tracking.
ALTER: Asymmetric LoRA for Token-Entropy-Guided Unlearning of LLMs
Xunlei Chen (University of Electronic Science and Technology of China), Wenhong Tian (University of Electronic Science and Technology of China)
Safty and PrivacyExplainability and InterpretabilityTransformerLarge Language ModelText
🎯 What it does: Propose ALTER, a lightweight LLM unlearning framework based on heterogeneous LoRA and token-entropy guided selective forgetting, achieving knowledge forgetting without modifying the base model's weights.
Ambiguity-aware Truncated Flow Matching for Ambiguous Medical Image Segmentation
Fanding Li, Shuo Li (Harbin Institute Of Technology)
SegmentationDiffusion modelFlow-based ModelBiomedical DataComputed Tomography
🎯 What it does: Proposed the Ambiguity-aware Truncated Flow Matching (ATFM) framework for simultaneously improving prediction accuracy and diversity in medical image segmentation.
Ambiguity-Tolerant Cross-Modal Hashing with Partial Labels
Chao Su (Sichuan University), Yuan Sun (Sichuan University)
RetrievalRepresentation LearningContrastive LearningMultimodality
🎯 What it does: This paper studies the partially labeled scenario in cross-modal retrieval, proposing a fuzzy-tolerant cross-modal hashing method called ATCH to address label ambiguity and modality alignment issues.
Amplifying Discrepancies: Exploiting Macro and Micro Inconsistencies for Image Manipulation Localization
Shenghao Chen, Zan Gao (Harbin Institute Of Technology)
Anomaly DetectionTransformerImage
🎯 What it does: Proposed FRD-Net, a network that achieves precise localization of image tampered regions by amplifying macro semantic differences and micro pixel differences.
AMS-IO-Bench and AMS-IO-Agent: Benchmarking and Structured Reasoning for Analog and Mixed-Signal Integrated Circuit Input/Output Design
Zhishuai Zhang (Tsinghua University), Nan Sun (Princeton University)
TransformerLarge Language ModelAgentic AITextBenchmark
🎯 What it does: This paper proposes AMS-IO-Agent, a domain-specific agent based on large language models (LLMs), designed to convert natural language design intent into structured I/O subsystem generation, while simultaneously providing the AMS-IO-Bench evaluation benchmark.
AMS-KV: Adaptive KV Caching in Multi-Scale Visual Autoregressive Transformers
Boxun Xu (University of California Santa Barbara), Peng Li (University of California Santa Barbara)
GenerationComputational EfficiencyTransformerAuto EncoderImage
🎯 What it does: This paper addresses the KV cache explosion problem in the multi-scale generation process of visual autoregressive models (VAR), proposing an adaptive multi-scale KV cache strategy AMS-KV, significantly reducing memory usage.
An Adaptive Configuration-Aware Simulated Annealing for the Maximally Diverse Grouping Problem
Baiyu Chen (Huazhong University of Science and Technology), Zhipeng Lü (Huazhong University of Science and Technology)
OptimizationGraphBenchmark
🎯 What it does: Propose an Adaptive Configuration-Aware Simulated Annealing (ACSA) algorithm to solve the Maximum Diversity Grouping Problem (MDGP), achieving search through relaxed insertion, memory-based exchange, and vertex-level sequence coordination.
An Adaptive Sampling Framework for Diffusion-based Dataset Distillation with High Fidelity and Diversity
Sunbeom Jeong (Seoul National University), Jungwoo Lee (Seoul National University)
Data SynthesisKnowledge DistillationDiffusion modelImage
🎯 What it does: This study proposes a dataset distillation framework based on diffusion models, which generates both faithful and diverse synthetic datasets by leveraging adaptive sampling (Bayesian optimization to select CFG scale and denoising strength) and repulsion regularization, without requiring fine-tuning of the diffusion model.
An Efficient and Harmonized Framework for Balanced Cross-Domain Feature Integration
Shaoxu Li (Shanghai Jiao Tong University), Ye Pan (Shanghai Jiao Tong University)
Image TranslationSupervised Fine-TuningDiffusion modelImage
🎯 What it does: Fine-tune Stable Diffusion using LoRA and achieve single-image style transfer through cross-model feature and attention injection, supporting mask-based local transfer and multi-style combinations
An Epistemic Perspective on Agent Awareness
Pavel Naumov (University of Southampton), Alexandra Pavlova (TU Wien)
Agentic AI
🎯 What it does: This paper views an agent's consciousness as knowledge, proposes distinguishing two forms of consciousness, de re and de dicto, and formalizes their expression by introducing new R and D modalities;
An Improved Privacy and Utility Analysis of Differentially Private SGD with Bounded Domain and Smooth Losses
Hao Liang (Hong Kong University of Science and Technology), Hong Xing (Hong Kong University of Science and Technology)
OptimizationSafty and Privacy
🎯 What it does: This paper presents an improved analysis of privacy and utility for differential privacy stochastic gradient descent (DPSGD), particularly under bounded domains and smooth loss functions, revealing the convergence of privacy loss over iterations.
An Information Theoretic Evaluation Metric for Strong Unlearning
Dongjae Jeon (Yonsei University), Jonghyun Choi (Seoul National University)
Safty and PrivacyExplainability and InterpretabilityConvolutional Neural NetworkTransformerContrastive LearningImage
🎯 What it does: This paper proposes the Information Difference Index (Information Difference Index, IDI) as a white-box evaluation metric for measuring the strength of machine unlearning (MU), and based on this, introduces the COLA method to further eliminate residual information left in intermediate layers by forgotten data.
An Invariant Latent Space Perspective on Language Model Inversion
Wentao Ye (Zhejiang University), Junbo Zhao (Southeast University)
Representation LearningTransformerLarge Language ModelSupervised Fine-TuningContrastive LearningText
🎯 What it does: Studies the language model inversion (LMI) problem, proposes the Invariant Latent Space Hypothesis (ILSH), and designs the Inv 2 A model: using LLM as an invariant decoder, learning a lightweight inverse encoder to map outputs to the LLM latent space for recovering hidden prompts.
An LLM-based Quantitative Framework for Evaluating High-Stealthy Backdoor Risks in OSS Supply Chains
Zihe Yan (Shanghai Jiao Tong University), Guancheng Li (Tencent)
Anomaly DetectionExplainability and InterpretabilityTransformerLarge Language ModelText
🎯 What it does: This paper proposes a fine-grained, attacker-perspective backdoor risk assessment framework with high stealthiness, combining large language models for semantic analysis to automatically evaluate backdoor risks in open-source software supply chains, and experimentally validates it on 66 high-priority packages in the Debian ecosystem.
An LLM-based Simulation Framework for Embodied Conversational Agents in Psychological Counseling
Lixiu Wu (Tsinghua University), Jiangtao Gong (Tsinghua University)
GenerationData SynthesisTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: Developed an 'ECAs' framework based on a large language model to simulate embodied dialogue agents in psychological counseling and generate high-fidelity, context-rich dialogue data.
Analyze–Compose–Execute: A Dynamic Dialogue Framework for Multi-Agent Debate
Wenyuan Gu (Beijing University of Posts and Telecommunications), Bo Cheng (Beijing University of Posts and Telecommunications)
Reinforcement Learning from Human FeedbackTransformerLarge Language ModelAgentic AITextBenchmark
🎯 What it does: Propose the ACE framework to enable multi-agent dynamic debates, allowing LLMs to freely choose strategies in discussions through an analysis-combine-execute loop.
Analyzing and Mitigating Object Hallucination: A Training Bias Perspective
Yifan Li (Renmin University of China), Jirong Wen
Explainability and InterpretabilityLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: This paper systematically evaluates the object hallucination phenomenon caused by training data bias in large vision-language models by constructing the POPEv2 benchmark.
Anatomical Region-Guided Contrastive Decoding: A Plug-and-Play Strategy for Mitigating Hallucinations in Medical VLMs
Xiao Liang (Xidian University), Yuanyuan Shi (Chinese PLA General Hospital)
TransformerSupervised Fine-TuningPrompt EngineeringVision Language ModelContrastive LearningBiomedical Data
🎯 What it does: Proposed an anatomy-region guided three-layer contrastive decoding (ARCD) framework to suppress hallucinations in medical vision-language models;
Anchor Watermark: Robust Attribution for Diffusion-based Text-to-Audio Model
Xianjin Rong (Hefei University of Technology), Donghui Hu (Hefei University of Technology)
GenerationData SynthesisDiffusion modelAudio
🎯 What it does: In text-to-audio diffusion models, watermarks are embedded into the initial latent vector and extracted in real-time through optimization during reverse diffusion of the generated audio, achieving audio attribution without additional training.
Anchor-Driven Nyström for Deep Graph-Level Clustering
Jiaxin Wang (Hainan University), Yue Yang (Hainan University)
Computational EfficiencyRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Designed an end-to-end graph-level clustering framework named ANGC, combining GNN encoding with learnable Nyström anchors to compute kernel similarity.
Anchor-Guided Discriminative Subspace Alignment and Clustering for Cross-Scene Hyperspectral Imagery
Yongshan Zhang (China University of Geosciences), Zhihua Cai (Wuhan University)
Domain AdaptationImage
🎯 What it does: Proposed a cross-scenario hyperspectral image clustering framework (ADSAC), which obtains source scenario clustering labels through anchor-enhanced graph learning (APGL), then eliminates distribution shift by discriminative cross-scenario subspace alignment (DCSA), and infers target scenario labels using KNN in the aligned subspace.
AnchorDS: Anchoring Dynamic Sources for Semantically Consistent Text-to-3D Generation
Jiayin Zhu (National University of Singapore), Angela Yao (National University of Singapore)
GenerationData SynthesisSupervised Fine-TuningDiffusion modelScore-based ModelImageTextBenchmark
🎯 What it does: Propose the AnchorDS method, which improves text-to-3D generation through dynamic source anchoring.
AnchorHOI: Zero-shot Generation of 4D Human-Object Interaction via Anchor-based Prior Distillation
Sisi Dai (National University of Defense Technology), Kai Xu (Institute of AI for Industries, Chinese Academy of Sciences)
GenerationPose EstimationVision-Language-Action ModelDiffusion modelScore-based ModelNeural Radiance FieldVideoText
🎯 What it does: Achieved zero-shot text-driven 4D human-robot interaction (HOI) generation by introducing a two-stage prior with AnchorNeRF and anchor keypoint.
AncientBench: Towards Comprehensive Evaluation on Excavated and Transmitted Chinese Corpora
Zhihan Zhou (Jilin University), Hao Xu (Jilin University)
Large Language ModelSupervised Fine-TuningImageTextBenchmark
🎯 What it does: Created the AncientBench evaluation framework, covering four capabilities of ancient characters: shape, sound, meaning, and context, and designed 10 multiple-choice tasks, achieving a three-stage method for digitization and encoding of ancient characters.
Angular Gradient Sign Method: Uncovering Vulnerabilities in Hyperbolic Networks
Minsoo Jo, Taesup Kim (Seoul National University)
ClassificationRetrievalAdversarial AttackConvolutional Neural NetworkVision Language ModelImageTextMultimodality
🎯 What it does: Proposed an adversarial attack method specifically for hyperbolic space—Angular Gradient Sign Method (AGSM) and its multi-step extension PAGD, leveraging radial-angular gradient decomposition in hyperbolic geometry to generate more effective adversarial examples.
Annealed Relaxation of Speculative Decoding for Faster Autoregressive Image Generation
Xingyao Li (National University Of Singapore), Hui Ji (National University Of Singapore)
GenerationComputational EfficiencyTransformerImage
🎯 What it does: Proposes COOL-SD, an acceleration method for autoregressive image generation based on theoretical analysis, improving the relaxation acceptance strategy and resampling distribution of Speculative Decoding;
Anomagic: Crossmodal Prompt-driven Zero-shot Anomaly Generation
Yuxin Jiang (Huazhong University of Science and Technology), Yunkang Cao (Hunan University)
GenerationAnomaly DetectionPrompt EngineeringVision Language ModelDiffusion modelContrastive LearningImageTextMultimodality
🎯 What it does: The study proposes a zero-shot anomaly generation framework that utilizes cross-modal prompts (visual + text) to drive Stable Diffusion for generating anomalous images in painting;
AnomalyMoE: Towards a Language-free Generalist Model for Unified Visual Anomaly Detection
Zhaopeng Gu (Foundation Model Research Center, Institute of Automation, Chinese Academy of Sciences), Jinqiao Wang (Foundation Model Research Center, Institute of Automation, Chinese Academy of Sciences)
Anomaly DetectionMixture of ExpertsImagePoint CloudBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: Proposed a language-agnostic unified visual anomaly detection framework named AnomalyMoE.
AnomalyPainter: Vision-Language-Diffusion Synergy for Realistic and Diverse Unseen Industrial Anomaly Synthesis
Zhangyu Lai (Xiamen University), Liujuan Cao (Xiamen University)
Data SynthesisAnomaly DetectionLarge Language ModelVision Language ModelDiffusion modelImage
🎯 What it does: Propose the AnomalyPainter framework, which leverages VLLM to generate anomaly descriptions, combines the Tex-9K texture library with LDM+ControlNet, and generates diverse and realistic industrial anomaly samples.
AnoStyler: Text-Driven Localized Anomaly Generation via Lightweight Style Transfer
Yulim So (Sungkyunkwan University), Seokho Kang (Sungkyunkwan University)
GenerationAnomaly DetectionConvolutional Neural NetworkImageText
🎯 What it does: Propose a lightweight zero-shot anomaly generation method, AnoStyler, which generates high-quality and diverse anomaly images on a single normal image by utilizing text-guided local style transfer;
Answering the Unanswerable Is to Err Knowingly: Analyzing and Mitigating Abstention Failures in Large Reasoning Models
Yi Liu (Nanjing University), Wei Hu (Nanjing University)
Anomaly DetectionExplainability and InterpretabilityComputational EfficiencyTransformerPrompt EngineeringTextChain-of-Thought
🎯 What it does: This paper analyzes the failure of large reasoning models to deny unsolvable problems and proposes a two-phase cognitive monitoring and intervention method during reasoning.
Anti-adversarial Learning: Desensitizing Prompts for Large Language Model
Xuan Li (Shanghai Jiao Tong University), Beijun Shen (Shanghai Jiao Tong University)
Safty and PrivacyTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Propose PromptObfus, a method for desensitizing LLM prompts through 'oppositional adversarial learning,' replacing sensitive words to make them unrecognizable to humans while preserving the original task output.
Anti-Avatar: Protect Against Unauthorized 3D Head Avatar Generation via Dual-Space Divergence
Lingzhuang Meng, Jie Zhang (Shandong Key Laboratory Of Intelligent Oil Gas Industrial Software)
GenerationAdversarial AttackConvolutional Neural NetworkTransformerImage
🎯 What it does: Propose the Anti-Avatar scheme, which utilizes adversarial perturbations to induce dual deviations in both the geometric and perceptual spaces of 3D avatar reconstruction, thereby preventing unauthorized 3D avatar generation from single images.
AntiDote: Bi-level Adversarial Training for Tamper-Resistant LLMs
Debdeep Sanyal (KIIT Bhubaneswar), Murari Mandal (KIIT Bhubaneswar)
OptimizationSafty and PrivacyAdversarial AttackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningText
🎯 What it does: Propose the AntiDote method, which trains LLMs through bi-level optimization to make them resistant to malicious fine-tuning.
Any-Optical-Model: A Universal Foundation Model for Optical Remote Sensing
Xuyang Li (Aerospace Information Research Institute, Chinese Academy of Sciences), Danfeng Hong (Southeast University)
ClassificationSegmentationTransformerAuto EncoderContrastive LearningImage
🎯 What it does: Proposed a universal optical remote sensing foundation model (AOM) capable of handling arbitrary spectral bands, resolutions, and sensor types.
Any2Critical: Safety-Critical Scenario Generation from Arbitrary Real-World Driving Contexts
Yao Huang (Beihang University), Xingxing Wei (Beihang University)
Data SynthesisAutonomous DrivingTransformerLarge Language ModelVision Language ModelImageTextMultimodalityRetrieval-Augmented Generation
🎯 What it does: Generate safety-critical (collision) test scenarios from any real driving scene, ensuring both diversity and behavioral plausibility.