AAAI 2026 Papers — Page 25
AAAI Conference on Artificial Intelligence · 4149 papers
MoEG-HOI: Mixture of Expert Groups for One-Stage Hand-Object Interaction Motion Generation with Hand-Finger-Joint Semantic Guidance
Hang Xu (Huazhong University of Science and Technology), Ran Wang (ByteDance Inc.)
GenerationTransformerLarge Language ModelMixture of ExpertsDiffusion modelTextPoint CloudMesh
🎯 What it does: This paper proposes a one-stage hand-object interaction motion generation framework called MoEG-HOI, which combines hierarchical semantic-guided expert groups (hand, fingers, joints) with an action and noise-aware router through Mixture-of-Experts (MoE) technology, achieving end-to-end training for hand motion generation.
MoEGaze: A Mixture of Experts Approach for Generalizable Gaze Estimation
Zheng Liu (Beihang University), Feng Lu (Beihang University)
Pose EstimationTransformerMixture of ExpertsImage
🎯 What it does: Propose MoEGaze—a gaze estimation framework based on Mixture-of-Experts, which dynamically routes experts using facial appearance features and predicts gaze direction by adaptively aggregating intermediate features generated by each expert.
MoETTA: Test-Time Adaptation Under Mixed Distribution Shifts with MoE-LayerNorm
Xiao Fan (Tongji University), Zhi Wang (Tsinghua University)
Domain AdaptationConvolutional Neural NetworkTransformerMixture of ExpertsImageBenchmark
🎯 What it does: This paper proposes MoETTA, a LayerNorm structure based on Mixture-of-Experts (MoE), to achieve test-time adaptation (TTA) with adaptive updates under mixed distribution shifts, and introduces two more realistic benchmarks, potpourri and potpourri+, to evaluate model robustness under multi-source, mixed distribution conditions.
MoFu: Scale-Aware Modulation and Fourier Fusion for Multi-Subject Video Generation
Run Ling (JD.com, Inc), Xingwei Wang (JD.com, Inc)
GenerationTransformerLarge Language ModelDiffusion modelVideoTextBenchmark
🎯 What it does: Proposed a unified multi-agent video generation framework called MoFu, aiming to simultaneously address the problems of inconsistent reference image scales and input order sensitivity.
MOGO: Residual Quantized Hierarchical Causal Transformer for Real-Time and Infinite-Length 3D Human Motion Generation
Dongjie Fu (MogoAI), Hansung Kim (University of Southampton)
GenerationTransformerLarge Language ModelTextSequentialChain-of-Thought
🎯 What it does: Proposes the MOGO framework, which realizes real-time, one-time generation from text to 3D human actions through hierarchical encoding with residual quantization (MoSA-VQ) and a single-pass forward residual quantization hierarchical causal transformer (RQHC-Transformer), supporting infinite-length generation.
MoHoBench: Assessing Honesty of Multimodal Large Language Models via Unanswerable Visual Questions
Yanxu Zhu (Beijing Jiaotong University), Xing Xie (Microsoft Research Asia)
Explainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningPrompt EngineeringVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Constructed the MoHoBench large-scale multimodal model honesty evaluation benchmark and systematically evaluated the honesty performance of 28 multimodal large language models on unanswerable visual questions.
MoLe-VLA: Dynamic Layer-skipping Vision Language Action Model via Mixture-of-Layers for Efficient Robot Manipulation
Rongyu Zhang (Hong Kong Polytechnic University), Shanghang Zhang (Peking University)
Robotic IntelligenceTransformerLarge Language ModelSupervised Fine-TuningMixture of ExpertsVision-Language-Action ModelMultimodality
🎯 What it does: This study proposes MoLe-VLA, a dynamic layer skipping framework based on a hybrid layer, aiming to significantly enhance the inference efficiency and performance of vision-language-action (VLA) models in robotic grasping and manipulation tasks.
MoLoRA: Boosting LLM-based End-to-end Speech Translation with Mixture of Low-rank Experts
Hao Zhang (Information Engineering University), Dan Qu (Information Engineering University)
TransformerLarge Language ModelSupervised Fine-TuningMixture of ExpertsTextAudio
🎯 What it does: Propose the MoLoRA framework, integrating LoRA as low-rank experts with the Mixture-of-Experts (MoE) mechanism, complemented by a multi-granularity representation fusion (MGRF) module, to achieve multilingual end-to-end speech translation;
MolSight: Optical Chemical Structure Recognition with SMILES Pretraining, Multi-Granularity Learning and Reinforcement Learning
Wenrui Zhang (Huazhong University of Science and Technology), Wenyu Liu (Huazhong University of Science and Technology)
RecognitionDrug DiscoveryTransformerSupervised Fine-TuningReinforcement LearningVision Language ModelImageText
🎯 What it does: Achieved optical chemical structure recognition (OCSR) through a three-stage training process (pre-training, hierarchical fine-tuning, reinforcement learning), with significant improvements in stereochemistry recognition.
MoMoREC: A Multi-agent Motivation Generation Framework for Residual Semantic ID-Aware Recommendation
Yige Wang (Taobao & Tmall Group of Alibaba), Xueying LI
Recommendation SystemTransformerLarge Language ModelAgentic AIContrastive LearningSequential
🎯 What it does: Propose the MoMoREC framework, which leverages multi-agent LLMs to generate user purchase motivations, and converts high-dimensional semantic embeddings into low-dimensional discrete IDs through motivation propagation and residual semantic ID methods, ultimately integrating them into the traditional sequence recommendation model (SASRec).
Mono3DVG-EnSD: Enhanced Spatial-aware and Dimension-decoupled Text Encoding for Monocular 3D Visual Grounding
Yuzhen Li, Yaonan Wang (Hunan University)
Object DetectionAutonomous DrivingTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Proposed a new framework called Mono3DVG-EnSD for locating 3D objects in monocular RGB images based on text descriptions.
MonoCloth: Reconstruction and Animation of Cloth-Decoupled Human Avatars from Monocular Videos
Daisheng Jin (Nanyang Technological University), Ying He (Nanyang Technological University)
GenerationData SynthesisPose EstimationRecurrent Neural NetworkGraph Neural NetworkNeural Radiance FieldGaussian SplattingVideo
🎯 What it does: Reconstruct and animate clothed human avatars from monocular video.
MonoCLUE: Object-Aware Clustering Enhances Monocular 3D Object Detection
Sunghun Yang (Yonsei University), Sangyoun Lee (Yonsei University)
Object DetectionAutonomous DrivingTransformerImage
🎯 What it does: Propose the MonoCLUE framework, which enhances the accuracy and robustness of 3D detection through object-level clustering and scene memory on monocular images.
Monocular Vehicle Pose and Shape Reconstruction via Dynamic Context Adaptation and Progressive Geometry Refinement
Wei Li (Southwest Jiaotong University), Penglin Dai (Southwest Jiaotong University)
Pose EstimationDepth EstimationConvolutional Neural NetworkTransformerImageMesh
🎯 What it does: Propose the MonoVPR framework to achieve joint reconstruction of vehicle 3D pose and shape from monocular images.
MonoDream: Monocular Vision-Language Navigation with Panoramic Dreaming
Shuo Wang (Renmin University of China), Deying Li (Horizon Robotics)
Autonomous DrivingRobotic IntelligenceLarge Language ModelSupervised Fine-TuningVision-Language-Action ModelImageTextSequential
🎯 What it does: In the monocular vision-language navigation task, MonoDream, a lightweight VLA framework, is proposed, which utilizes Unified Navigation Representation (UNR) and Latent Panoramic Dreaming (LPD) tasks to enable agents with monocular cameras to infer panoramic views, depth, and future states from limited perspectives, thereby achieving more reliable navigation decisions.
Monte Carlo Diffusion for Generalizable Learning-Based RANSAC
Jiale Wang (Xi'an Jiaotong University), Tong Zhang (EPFL)
Data SynthesisOptimizationGraph Neural NetworkDiffusion modelPoint Cloud
🎯 What it does: By introducing a diffusion-based Monte Carlo sampling mechanism, noise is incrementally injected into ground truth matches to generate diverse synthetic matches, which are used to train learning-based RANSAC models, significantly enhancing their generalization capability for discrete distribution matches.
More than Irrational: Modeling Belief-Biased Agents
Yifan Zhu (ELLIS Institute), Samuel Kaski (ELLIS Institute)
Explainability and InterpretabilityReinforcement LearningSequential
🎯 What it does: Proposed a user model based on the computational rationality framework, utilizing a parameterizable memory decay function to explain irrational behaviors, and developed an online inference algorithm with nested particle filters to simultaneously estimate users' memory boundary parameters and internal biased beliefs from passively observed behavioral sequences;
MoReMouse: Monocular Reconstruction of Laboratory Mouse
Yuan Zhong (Tsinghua University), Yebin Liu (Tsinghua University)
GenerationData SynthesisTransformerNeural Radiance FieldGaussian SplattingVideoMeshBiomedical Data
🎯 What it does: Developed MoReMouse, a monocular dense 3D reconstruction network specifically designed for C57BL/6 experimental mice, capable of generating high-quality surface geometry and appearance from a single image.
MORGAN: To Bridge Mixture of Experts and Spectral Graph Neural Network
Lihui Liu (Wayne State University), Yuchen Yan (University of Illinois at Urbana Champaign)
ClassificationGraph Neural NetworkMixture of ExpertsGraphBenchmark
🎯 What it does: Propose MORGAN, a Mixture-of-Experts spectral GNN that treats eigengraphs as experts to address edge heterogeneity.
Mosaic Pruning: A Hierarchical Framework for Generalizable Pruning of Mixture-of-Experts Models
Wentao Hu (Xi'an Jiaotong University), Jiayin Wang (Xi'an Jiaotong University)
Computational EfficiencyMixture of ExpertsTextBenchmark
🎯 What it does: Propose a hierarchical 'clustering-reselection' framework called Mosaic Pruning (MoP) for structured pruning in large-scale sparse expert models (MoE), enabling cross-domain deployment after one-time pruning while maintaining functional diversity.
MosaicDoc: A Large-Scale Bilingual Benchmark for Visually Rich Document Understanding
Ketong Chen (South China University of Technology), Yang Xue (South China University of Technology)
TransformerLarge Language ModelVision Language ModelTextMultimodalityBenchmark
🎯 What it does: Propose the DocWeaver multi-agent automated generation framework and construct the first visually rich document understanding benchmark MosaicDoc for Chinese and English newspapers and magazines.
MoSE: Hierarchical Self-Distillation Enhances Early Layer Embeddings
Andrea Gurioli (University of Bologna), Maurizio Gabbrielli (University of Bologna)
ClassificationRetrievalKnowledge DistillationTransformerText
🎯 What it does: Designed and implemented a 1B parameter multi-output encoder MOSE based on STARCODER-2, enhancing early layer representations through self-distillation for code retrieval and classification tasks.
MoSs: Mixture of Scales for Efficient High-Resolution Autoregressive Image Generation
Yaoxiu Lian, Ningyi Xu (Shanghai Jiao Tong University)
GenerationTransformerImage
🎯 What it does: Proposed a hybrid scale (MoSs) framework that requires no training and can be directly embedded into existing high-resolution visual autoregressive (VAR) models, achieving parallel refinement through frequency-aware token allocators and cross-scale self-attention.
MOTIF: Multi-strategy Optimization via Turn-based Interactive Framework
Nguyen Viet Tuan Kiet (Hanoi University of Science and Technology), Huynh Thi Thanh Binh (Hanoi University of Science and Technology)
OptimizationTransformerLarge Language ModelAgentic AIPrompt EngineeringBenchmark
🎯 What it does: Through a two-phase interactive framework based on MCTS, two LLM agents take turns improving multiple solver strategies to achieve collaborative optimization.
Motion-Aware Object Tracking via Motion and Geometry-Aware Cues
Hongtao Yang (Guangxi Normal University), Shuxiang Song (Guangxi Normal University)
Object TrackingDepth EstimationVideo
🎯 What it does: Proposed a motion-aware spatiotemporal target tracking framework that enhances motion perception through explicit matching of motion patterns and modeling of inter-frame motion relationships.
MotionCharacter: Fine-Grained Motion Controllable Human Video Generation
Haopeng Fang (Huazhong University of Science and Technology), He Tang (Meituan)
GenerationTransformerVision Language ModelDiffusion modelOptical FlowVideoText
🎯 What it does: Propose the MotionCharacter framework to achieve personalized human video generation with fine-grained control over action types and motion intensity
MotionFlow: Attention-Driven Motion Transfer in Video Diffusion Models
Tuna Han Salih Meral (Virginia Tech), Pinar Yanardag (Virginia Tech)
GenerationDiffusion modelVideoTextMultimodality
🎯 What it does: Propose MotionFlow, a training-agnostic test-time latent optimization framework that transfers motion from a source video to generate a new video under target text prompts.
MotionPhysics: Learnable Motion Distillation for Text-Guided Simulation
Miaowei Wang (University of Edinburgh), Amir Vaxman (University of Edinburgh)
GenerationData SynthesisKnowledge DistillationTransformerLarge Language ModelDiffusion modelGaussian SplattingVideoTextMeshPhysics Related
🎯 What it does: MotionPhysics automatically infers physical parameters through natural language prompts and achieves 3D dynamic simulation with distinguishable materials by combining a learnable motion distillation loss.
MotivDance: Fine-Grained Text-Guided Motivation Choreography with Music Synchronization
Chenguang Li (State Key Laboratory of Advanced Rail Autonomous Operation), Liping Jing (State Key Laboratory of Advanced Rail Autonomous Operation)
GenerationData SynthesisTransformerVision-Language-Action ModelDiffusion modelContrastive LearningTextMultimodalityAudio
🎯 What it does: This paper proposes a framework called MotivDance, which can generate high-quality 3D dance motions synchronized with music based on fine-grained text descriptions. The core idea is to first generate key poses from text (Motivation), then align these poses to musical beats using an adaptive keyframe locator, and finally use a Transformer-U-Net diffusion model to interpolate motions, achieving dance generation under dual constraints of text and music.
MoToRec: Sparse-Regularized Multimodal Tokenization for Cold-Start Recommender
Jialin Liu, Ray C.C. Cheung (City University Of Hong Kong)
Recommendation SystemGraph Neural NetworkAuto EncoderContrastive LearningImageTextGraph
🎯 What it does: Propose the MoToRec framework, which converts the multimodal recommendation task into a sparse regularized residual quantized variational autoencoder (RQ-VAE) to generate interpretable discrete semantic tokens, and achieves cold start recommendations through adaptive sparsity amplification and hierarchical multi-source graph encoding.
MovieGraph-ToM: Evaluating Long-Range Theory of Mind in Large Language Models via Implicit Social-Causal Graphs
Tingjiang Wei (East China Normal University), Liang He (East China Normal University)
Explainability and InterpretabilityTransformerLarge Language ModelTextMultimodalityGraphBenchmark
🎯 What it does: Constructed a large-scale multimodal benchmark, MovieGraph-ToM, to evaluate LLMs' long-range theory of mind (ToM) reasoning capabilities in full movies.
MovSemCL: Movement-Semantics Contrastive Learning for Trajectory Similarity
Zhichen Lai, Christian S. Jensen (Roskilde University)
RetrievalRepresentation LearningGraph Neural NetworkTransformerContrastive LearningGraphSequential
🎯 What it does: Designed and implemented a motion semantic contrastive learning framework (MovSemCL) to extract motion semantic features from raw GPS trajectories, employing hierarchical patch encoding and curvature-guided enhancement, ultimately generating efficient and robust trajectory embeddings.
MP-ISMoE: Mixed-Precision Interactive Side Mixture-of-Experts for Efficient Transfer Learning
Yutong Zhang (Beihang University), Jiaxin Chen (United Arab Emirates University)
Computational EfficiencySupervised Fine-TuningMixture of ExpertsVision Language ModelTextMultimodality
🎯 What it does: Propose a Mixed Precision Interactive Side Expert Network (MP-ISMoE), which significantly improves the performance of memory-efficient transfer learning by efficiently quantizing the frozen backbone and introducing sparse MoE in the side network.
MP1: MeanFlow Tames Policy Learning in 1-step for Robotic Manipulation
Juyi Sheng (Peking University), Mengyuan Liu (Peking University)
Robotic IntelligenceFlow-based ModelPoint Cloud
🎯 What it does: Propose the MP1 framework, which utilizes MeanFlow to generate robot control trajectories in a single step.
MPA: Multimodal Prototype Augmentation for Few-Shot Learning
Liwen Wu (Yunnan University), Bin Pu (Yunnan University)
ClassificationMeta LearningLarge Language ModelVision Language ModelMultimodality
🎯 What it does: This paper proposes a multi-modal prototype enhancement framework called MPA, which improves few-shot learning performance through LLM-generated diverse semantics, hierarchical multi-perspective enhancement, and adaptive absorption of uncertain categories.
MPAS: Breaking Sequential Constraints of Multi-Agent Communication Topologies via Individual-Epistemic Message Propagation
Jingxuan Yu (Southeast University), Guang Cheng (Southeast University)
OptimizationComputational EfficiencyGraph Neural NetworkTransformerLarge Language ModelAgentic AIPrompt EngineeringTextGraph
🎯 What it does: Propose a node-level multi-agent system MPAS based on graph neural network message propagation mechanisms, breaking the traditional sequential topology constraints to enable parallel information exchange under arbitrary topologies.
MPD-SGR: Robust Spiking Neural Networks with Membrane Potential Distribution-Driven Surrogate Gradient Regularization
Runhao Jiang (Zhejiang University), Huajin Tang (Zhejiang University of Technology)
Adversarial AttackSpiking Neural NetworkImage
🎯 What it does: This paper proposes an alternative gradient regularization (MPD-SGR) method based on membrane potential distribution (MPD) to enhance the adversarial robustness of deep spiking neural networks (SNNs).
MPI-Mamba: Latent Feature Fusion Mamba for Anisotropic Image Calibration and Deblurring in Magnetic Particle Imaging
Liwen Zhang (Key Laboratory of Cognition and Decision Intelligence for Complex Systems Institute of Automation Chinese Academy of Sciences), Jie Tian (Southeast University)
RestorationDiffusion modelBiomedical Data
🎯 What it does: Propose MPI-Mamba, an end-to-end framework integrating Mamba-based LFF-SSM with conditional latent diffusion models (CL-DM), for anisotropic image correction and deblurring in magnetic particle imaging (MPI) caused by unidirectional scanning.
MPJudge: Towards Perceptual Assessment of Music-Induced Paintings
Shiqi Jiang (East China Normal University), Chenhui Li (East China Normal University)
Convolutional Neural NetworkTransformerVision Language ModelImageMultimodalityAudio
🎯 What it does: Proposed a perceptual evaluation task for music-inspired paintings, constructed a large-scale manually annotated dataset named MPD, and designed the MPJudge model to achieve cross-modal fusion and preference optimization.
MPMA: Preference Manipulation Attack Against Model Context Protocol
Zihan Wang (University of Electronic Science and Technology of China), Guowen Xu (University of Electronic Science and Technology of China)
Adversarial AttackLarge Language ModelPrompt EngineeringText
🎯 What it does: Proposed a novel security threat targeting the Model Context Protocol (MCP) called MPMA, and designed two attack strategies: Direct Preference Manipulation Attack (DPMA) and Genetic Algorithm-based Preference Manipulation Attack using Advertising Strategies (GAPMA); conducted systematic experiments on attack effectiveness and stealthiness.
MR-COSMO: Visual-Text Memory Recall and Direct CrOSs-MOdal Alignment Method for Query-Driven 3D Segmentation
Chade Li (Chinese Academy of Sciences), Yihong Wu (Chinese Academy of Sciences)
SegmentationTransformerLarge Language ModelVision Language ModelContrastive LearningImageTextMultimodalityPoint Cloud
🎯 What it does: Proposed the MR-COSMO model, using direct cross-modal alignment and visual-textual memory recall to achieve query-driven 3D segmentation;
MrCoM: A Meta-Regularized World-Model Generalizing Across Multi-Scenarios
Xuantang Xiong (Chinese Academy of Sciences), Bo Xu (Chinese Academy of Sciences)
Meta LearningTransformerReinforcement LearningAuto EncoderWorld Model
🎯 What it does: Proposes the Meta-Regularized Contextual World-Model (MrCoM), achieving the learning and transfer of a unified world model across multiple scenarios.
MRGeo: Robust Cross-View Geo-Localization of Corrupted Images via Spatial and Channel Feature Enhancement
Le Wu (Shenzhen University), Yingying Zhu (Shenzhen University)
RetrievalTransformerContrastive LearningImageBenchmark
🎯 What it does: Propose MRGeo, a systematic approach for robust cross-view geolocation (CVGL) under image distortion environments, primarily through multi-layer defense strategies to enhance feature quality and enforce geometric consistency.
MrM: Black-Box Membership Inference Attacks Against Multimodal RAG Systems
Peiru Yang (Tsinghua University), Tao Qi (Beijing University of Posts and Telecommunications)
RetrievalSafty and PrivacyAdversarial AttackVision Language ModelImageRetrieval-Augmented Generation
🎯 What it does: This work proposes MrM, a black-box membership inference attack (MIA) framework targeting multi-modal retrieval-augmented generation (RAG) systems, which leverages visual object masking and adversarial mask selection to induce the system to leak information about whether data exists in the knowledge base during retrieval and generation stages.
MRT: Learning Compact Representations with Mixed RWKV-Transformer for Extreme Image Compression
Han Liu, Debin Zhao (Harbin Institute of Technology)
CompressionRecurrent Neural NetworkTransformerGenerative Adversarial NetworkImage
🎯 What it does: Propose a hybrid RWKV-Transformer (MRT) architecture that achieves ultra-low bitrate image compression using 1-D sparse representations;
MS-PPO: Mean Standard Deviation Proximal Policy Optimization for Reliable Parking Space Search in Structured Environments
Haoming Chen (Sichuan University), Hongliang Guo (Sichuan University)
Reinforcement LearningGraphTabular
🎯 What it does: Proposed a reliable parking spot search algorithm called MS-PPO for structured parking environments, aiming to minimize a linear combination of the mean and standard deviation of parking search time.
MSAnchor: De Novo Molecular Generation from Mass Spectrometry Data with Anchor-Extended Molecular Scaffolds
Xiaohan Qin (University of Science and Technology of China), Yang Wang (University of Science and Technology of China)
Drug DiscoveryGraph Neural NetworkTransformerDiffusion modelBiomedical Data
🎯 What it does: Developed a two-stage molecular structure generation framework called MSAnchor, which first uses a Transformer to predict the Anchor-Extended Molecular Scaffold (AEMS) with retained side chain anchor points, and then employs a conditional information bottleneck (CIB) to refine mass spectrometry features and reconstruct molecular structures through a diffusion model.
MSAT-LDM: Toward Transferable High-Fidelity Watermarking for Latent Diffusion Model via Modular Self-Augmented Training
Lu Zhang (Huazhong University of Science and Technology), Liang Zeng (Huazhong University of Science and Technology)
GenerationData SynthesisAdversarial AttackDiffusion modelImage
🎯 What it does: This paper proposes MSAT-LDM, a high-fidelity and transferable watermark embedding framework implemented on a latent diffusion model (LDM).
MSCFL: Model Structure-Aware Clustered Federated Learning for System Heterogeneity and Data Drift
Yang Xu (Hunan University), Yaoxue Zhang (Tsinghua University)
Domain AdaptationFederated LearningImage
🎯 What it does: Proposes the MSCFL framework, integrating model pruning with clustering federated learning to address system heterogeneity, data heterogeneity, and data drift issues.
MSME: A Multi-Stage Multi-Expert Framework for Zero-Shot Stance Detection
Yuanshuo Zhang (Minzu University of China), Xiaobing Zhao (Minzu University of China)
ClassificationTransformerLarge Language ModelPrompt EngineeringMixture of ExpertsTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Proposed the MSME (Multi-Stage Multi-Expert) framework, using three-stage decomposition (knowledge preparation, expert reasoning, decision aggregation) to address issues such as missing background knowledge, unclear target-label mapping, and rhetorical complexity in zero-shot stance detection.
MSPCaps: A Multi-Scale Patchify Capsule Network with Cross-Agreement Routing for Visual Recognition
Yudong Hu (Chinese University of Hong Kong (Shenzhen)), Jinke Ren (Chinese University of Hong Kong (Shenzhen))
RecognitionConvolutional Neural NetworkImage
🎯 What it does: Proposed a multi-scale Patchify Capsule network (MSPCaps), achieving efficient fusion and voting of multi-scale features through a multi-scale ResNet Backbone, PatchifyCaps, and Cross-Agreement Routing (CAR).
MSR-Rec: Multi-Step Reasoning-Enhanced LLM for Sequential Recommendation
Tuo Wang, Yashen Wang (Beijing University Of Technology)
Recommendation SystemTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextSequentialChain-of-Thought
🎯 What it does: This paper proposes a multi-step reasoning enhanced large language model, MSR-Rec, for sequential recommendation tasks.
MSTDiff: Multiscale-Aware Transformer Diffusion Network for Video Object Detection
Qiang Qi (Qingdao University of Science and Technology), Shuyuan Lin (Peng Cheng Laboratory)
Object DetectionTransformerDiffusion modelVideo
🎯 What it does: Proposes the MSTDiff framework for end-to-end video object detection, incorporating diffusion-driven adaptive queries and multi-scale perception Transformer encoders.
MTAttack: Multi-Target Backdoor Attacks Against Large Vision-Language Models
Zihan Wang (Beihang University), Xiao Bai (Beihang University)
Adversarial AttackSupervised Fine-TuningVision Language ModelImageText
🎯 What it does: The study addresses multi-target backdoor attacks in large-scale vision-language models (LVLMs), proposing the MTAttack framework to successfully implant backdoors.
MTP: Exploring Multimodal Urban Traffic Profiling with Modality Augmentation and Spectrum Fusion
Haolong Xiang (Nanjing University of Information Science and Technology), Wei Fan (University of Auckland)
ClassificationRepresentation LearningTransformerLarge Language ModelContrastive LearningMultimodalityTime Series
🎯 What it does: Proposed a multi-modal traffic flow portrait framework called MTP, which learns numerical, visual, and textual multi-modal features in the frequency domain and achieves fine-grained urban traffic state portraits through hierarchical contrastive learning.
MTRL-CG: Multi-Task Reinforcement Learning Method with Spectral Clustering-Based Task Grouping
Wenjia Meng (Shandong University), Yilong Yin (Shandong University)
OptimizationReinforcement LearningBenchmark
🎯 What it does: Propose a multi-task reinforcement learning method, MTRL-CG, which groups related tasks using spectral clustering and reduces negative interference by employing group-specific policies.
MUG: Meta-path-aware Universal Heterogeneous Graph Pre-Training
Lianze Shan (Tianjin University), Weixiong Zhang (Hong Kong Polytechnic University)
ClassificationRepresentation LearningGraph Neural NetworkAuto EncoderContrastive LearningGraph
🎯 What it does: Proposed a general pre-training method called MUG for heterogeneous graphs, which can be pre-trained on different heterogeneous graph datasets and directly transferred to downstream tasks;
Multi-Agent Corridor Reasoning for Multi-Agent Path Finding
Yiran Ni (Zhejiang University), Deshi Ye (Zhejiang University)
OptimizationBenchmark
🎯 What it does: Propose a multi-agent corridor reasoning technique for symmetric conflicts in multi-agent corridors, improving the CBS algorithm to resolve multiple conflicts in a single step, significantly reducing high-level node splitting.
Multi-agent In-context Coordination via Decentralized Memory Retrieval
Tao Jiang (Nanjing University), Deheng Ye (Tencent)
TransformerReinforcement LearningSequentialRetrieval-Augmented Generation
🎯 What it does: This paper proposes a multi-agent parameter-free update context learning coordination framework MAICC, which achieves rapid collaborative adaptation through decentralized memory retrieval.
Multi-Agent Pointer Transformer: Seq-to-Seq Reinforcement Learning for Multi-Vehicle Dynamic Pickup-Delivery Problems
Zengyu Zou (Beihang University), Junjie Wu (Beihang University)
OptimizationTransformerReinforcement LearningSequential
🎯 What it does: This paper proposes a multi-agent pointer Transformer (MAPT) framework based on Transformer, designed to solve the multi-vehicle dynamic random request delivery problem, generating joint action sequences through self-regressive decoding at each step;
Multi-Agent Undercover Gaming: Hallucination Removal Through Counterfactual Test for Multimodal Reasoning
Dayong Liang (South China University of Technology), Changmeng Zheng (Hong Kong Polytechnic University)
Anomaly DetectionExplainability and InterpretabilityTransformerLarge Language ModelAgentic AIVision Language ModelMultimodalityBenchmarkChain-of-Thought
🎯 What it does: Propose a Multi-Agent Covert Game (MUG) protocol that detects and eliminates hallucinations in LLMs by generating counterfactual images for counterfactual testing.
Multi-Aspect Cross-modal Quantization for Generative Recommendation
Fuwei Zhang (Beihang University), Zhao Zhang (Beihang University)
Recommendation SystemTransformerLarge Language ModelAuto EncoderContrastive LearningMultimodality
🎯 What it does: This paper proposes the MACRec framework, which utilizes cross-modal quantization to generate hierarchical semantic IDs and introduces multi-dimensional cross-modal alignment in the generative recommendation stage, thereby significantly enhancing recommendation performance.
Multi-dimensional Adaptive Mix-hop Contextual Learning Framework for Universal Graph Anomaly Detection
Zhaowei Liu (Yantai University), Haitao Yang (Yantai University)
Anomaly DetectionGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposes the SAARCS general graph anomaly detection framework, which can detect abnormal nodes on different graph datasets without requiring fine-tuning.
Multi-dimensional Neural Decoding with Orthogonal Representations for Brain-Computer Interfaces
Kaixi Tian (Institute of Automation, Chinese Academy of Sciences), Shan Yu (Institute of Automation, Chinese Academy of Sciences)
Convolutional Neural NetworkRecurrent Neural NetworkBiomedical Data
🎯 What it does: Proposes a multi-dimensional neural decoding (MND) task that can simultaneously extract multiple related motor variables, including direction, position, velocity, and acceleration, from a single neural population.
Multi-District School Choice: Playing on Several Fields
Yannai A. Gonczarowski (Harvard University), Shirley Zhang (Harvard University)
🎯 What it does: Analyzed the interactions between honest and strategic students under a multi-zone school choice mechanism and their impact on matching outcomes, revealing phenomena opposite to those predicted by single-zone models.
Multi-Granular Graph Learning with Fine-Grained Behavioral Pattern Awareness for Session-Based Recommendation
Ming Li (Zhejiang Normal University), Zhao Li (Zhejiang Lab)
Recommendation SystemGraph Neural NetworkGraphSequential
🎯 What it does: Propose a multi-grained graph learning framework called GraphFine, which models short-term, segmented, and long-term behavior patterns in sessions through graph and hypergraph neural networks to enhance session recommendation performance.
Multi-granularity Intent Modeling with Adversarial Robustness for Sequential Recommendation
Yangyi Fang (Tsinghua University), Haolin Shi (Tsinghua University)
Recommendation SystemRecurrent Neural NetworkTransformerLarge Language ModelContrastive LearningSequential
🎯 What it does: Proposes a sequence recommendation framework named MIMAR-SRec that integrates multi-granularity intent modeling with adversarial robustness;
Multi-granularity Interactive Attention Framework for Residual Hierarchical Pronunciation Assessment
Hong Han (Shandong University), Xin-Shun Xu (Shandong University)
RecognitionConvolutional Neural NetworkTransformerSequentialAudio
🎯 What it does: Propose a residual hierarchical interactive attention framework (HIA) that achieves bidirectional interaction at three granularities—phoneme, word, and sentence—and utilizes a residual structure to alleviate feature forgetting in hierarchical modeling.
Multi-granularity Temporal Knowledge Editing over Large Language Models
Simiao Zhao (National University of Defense Technology), Xiang Zhao (National University of Defense Technology)
TransformerLarge Language ModelTime SeriesSequentialBenchmark
🎯 What it does: This paper proposes a Multi-Granularity Temporal Knowledge Editing (MTKE) framework and constructs a corresponding benchmark dataset, MTKE.
Multi-graph Fusion Cross-model Contrastive Learning for Recommendation
Shengjun Ma (Northeastern University), Wen Shan (Singapore Institute of Technology)
Recommendation SystemGraph Neural NetworkContrastive LearningGraphBenchmark
🎯 What it does: Designed and implemented an end-to-end recommendation framework called MFCCL, which leverages information from user interest graph (UiG), item association graph (IaG), and knowledge graph (KG), integrating multi-graph attention mechanisms with traditional collaborative filtering to achieve multi-perspective representation learning for users and items;
Multi-Horizon Time Series Forecasting of Non-Parametric CDFs with Deep Lattice Networks
Niklas Erdmann (University of Oslo), Paal E. Engelstad (University of Oslo)
Recurrent Neural NetworkTime Series
🎯 What it does: Achieved same-quantile quantile regression (SQR) prediction for multi-time domain non-parametric cumulative distribution functions (CDF) by combining deep lattice networks (DLN) with LSTM, and first applied DLN to multi-step forecasting of time series.
Multi-knowledge Enhanced Graph Neural Network for Multi-trait Essay Scoring
Shiman Zhao (Nanyang Technological University), Zhiqi Shen (Nanyang Technological University)
ClassificationRepresentation LearningGraph Neural NetworkTransformerLarge Language ModelText
🎯 What it does: Propose a multi-knowledge enhanced graph neural network for multi-attribute essay scoring (MES), enhancing scoring effectiveness through three layers: sentence syntax graphs, feature graphs, and ordinal knowledge of scores.
Multi-Label Classification with Incremental and Decremental Features
Mingdie Jiang (National University of Defense Technology), Chenping Hou (National University of Defense Technology)
Classification
🎯 What it does: Proposes MLID, a two-stage, one-shot learning framework for multi-label classification in dynamic environments with feature increment and decay;
Multi-Level Blur-Aware Stable Diffusion for Region-Adaptive Defocus Deblurring
Xiaopan Li (Hubei University of Economics), Sos Agaian (A*STAR)
RestorationConvolutional Neural NetworkMixture of ExpertsDiffusion modelImage
🎯 What it does: Propose the Multi-Level Blur-Aware Stable Diffusion (MBSD) framework, combining patch-level blur detection with Stable Diffusion to achieve region-aware defocus deblurring.
Multi-Level Domain Adaptation and Contrastive Domain Isolation with Bilinear Fusion for Patient Drug Response Prediction
Yuting Bai (Hunan University), Jiawei Luo (Hunan University)
Domain AdaptationDrug DiscoveryContrastive LearningBiomedical Data
🎯 What it does: Designed and implemented a three-stage hierarchical domain adaptation framework, MACB-DRP, to transfer knowledge of cell line drug sensitivity to patient drug response prediction.
Multi-level Style Preference Optimization: An Adaptive Detection Framework for Human-Machine Hybrid Text
Zehao Wang (Northwestern Polytechnical University), Yaxiong Wang (Northwestern Polytechnical University)
ClassificationData-Centric LearningText
🎯 What it does: Propose the Multi-level Style Preference Optimization (MSPO) framework, which detects human-machine mixed text through multi-level (sequence, phrase, lexical) style preference optimization.
Multi-Metric Preference Alignment for Generative Speech Restoration
Junan Zhang (Chinese University of Hong Kong), Zhizheng Wu (Chinese University of Hong Kong)
RestorationReinforcement LearningBenchmarkAudio
🎯 What it does: A multi-criteria preference alignment method was studied, aligning the training objective of the speech restoration model with human auditory preferences to improve restoration quality.
Multi-Modal Assistance for Unsupervised Domain Adaptation on Point Cloud 3D Object Detection
Shenao Zhao (Zhengzhou University), Zhoufan Yang (Zhengzhou University)
Object DetectionDomain AdaptationAutonomous DrivingTransformerLarge Language ModelVision Language ModelMultimodalityPoint Cloud
🎯 What it does: Proposed the MMAssist framework, which achieves multi-modal assisted unsupervised domain adaptation for point cloud 3D object detection by introducing image and text features as bridges between source and target domain 3D detection models.
Multi-modal Dynamic Proxy Learning for Personalized Multiple Clustering
Jinfeng Xu (University of Hong Kong), Edith C. H. Ngai (University of Nottingham)
Recommendation SystemLarge Language ModelVision Language ModelContrastive LearningMultimodality
🎯 What it does: Proposed Multi-DProxy, a personalized multi-clustering method based on multi-modal dynamic proxy learning.
Multi-Modal Fact Knowledge Generation for Imbalanced Cross-Source Entity Alignment
Qian Li (Beijing University Of Posts And Telecommunications), Siyuan Liang (Beijing University Of Posts And Telecommunications)
Representation LearningGraph Neural NetworkTransformerLarge Language ModelMultimodalityGraphBenchmark
🎯 What it does: Propose a multimodal factual knowledge generation framework based on large language models (LLMEA), addressing information gaps in cross-source entity alignment of multimodal knowledge graphs through neighbor entity completion, attribute completion, and fact evaluation.
Multi-Modal Style Transfer-based Prompt Tuning for Efficient Federated Domain Generalization
Yuliang Chen (Shanghai Jiao Tong University), Xiu Su (Central South University)
Domain AdaptationFederated LearningRepresentation LearningPrompt EngineeringVision Language ModelContrastive LearningMultimodality
🎯 What it does: Under the federated learning framework, the FaST-PT method is proposed, which employs CLIP for multi-modal style transfer (MST) to locally enhance image embeddings, and achieves efficient out-of-domain generalization through dual prompts (global prompt + domain prompt) and domain-aware prompt generation (DPG);
Multi-Objective Bilevel Learning
Zhiyao Zhang (Ohio State University), Jia Liu (Ohio State University)
OptimizationMeta LearningImage
🎯 What it does: This paper proposes the Weighted-Chebyshev Multi-Hypergradient Descent (WC-MHGD) framework for multi-objective bilevel learning, which can solve ε-Pareto steady-state solutions in both deterministic and stochastic environments.
Multi-Semantic Modeling for Glass Surface Detection in the Wild
Qianyu Cheng (City University of Hong Kong), Rynson W. H. Lau (City University of Hong Kong)
Object DetectionVision Language ModelImage
🎯 What it does: Proposed a multi-semantic modeling framework MSNet for detecting glass surfaces in complex scenes
Multi-Step Deformable Gaussian Splatting for Dynamic Scene Rendering
Jiaheng Hu (East China Normal University), Yuan Xie (East China Normal University)
GenerationGaussian SplattingImage
🎯 What it does: Propose a multi-step deformable Gaussian splatting framework for rendering dynamic scenes
Multi-Task Test-time Adaptation via Gradient Consensus and Plasticity Constraint
Zhong Ye (Guangdong University of Technology), Zhenguo Yang (Guangdong University of Technology)
Domain AdaptationImage
🎯 What it does: This paper proposes a multi-task test-time adaptation method called CoCo-MT-TTA to address gradient conflicts and catastrophic forgetting in multi-task scenarios.
Multi-Value Alignment for LLMs via Value Decorrelation and Extrapolation
Hefei Xu (Hefei University of Technology), Hao Liu (Hefei University of Technology)
OptimizationReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningText
🎯 What it does: Propose a multivalued alignment framework called MVA to address parameter interference issues in large language models (LLMs) when dealing with conflicting human values. The framework utilizes information-theoretic regularization and parameter extrapolation to generate diverse models with multiple values.
Multi-View Clustering with Granularity-Aware Pseudo Supervision
Jie Yang (University of Sydney), Bingbing Jiang (Nanjing University)
Computational EfficiencyRepresentation LearningMultimodality
🎯 What it does: Designed a multi-view clustering framework GAPS, leveraging adjustable granularity pseudo-labels and reliable view selection to achieve adaptive clustering
Multi-View Differential Mixing and Graph-Guided Structural Region Selection for Cross-Modal Alignment
Linlin Ji (Shandong Normal University), Li Liu (Shandong Normal University)
RetrievalTransformerMixture of ExpertsVision Language ModelMultimodality
🎯 What it does: This paper proposes the MG-Net framework to address the global representation information bottleneck and the lack of spatial structure in local features in image-text matching;
Multi-view Invariance Learning for 3D Scene Graph Pre-training via Collaborative Cross-Modal Regularization
Yucheng Huang (University of Electronic Science and Technology of China), Jiayuan Sun (University of Electronic Science and Technology of China)
Representation LearningTransformerVision Language ModelPoint Cloud
🎯 What it does: This paper proposes a multi-view invariant learning framework for 3D scene graph pre-training, combining cross-modal regularization and knowledge filtering gate adapters to achieve self-supervised learning of object and predicate features.
Multi-view Learning via Trusted Pairwise Entity Energy
Yalan Qin (Shanghai University), Xinpeng Zhang (Shanghai University)
ClassificationConvolutional Neural NetworkMultimodality
🎯 What it does: Proposed a pairwise trusted problem for long-tailed multi-view classification, along with a general framework and an implementation based on ENIG (Enhanced Normal-Inverse Gamma distribution), achieving the fusion of multi-view data at the evidence level to generate reliable classification results.
Multi-Window Gabor Transform Network for Ground Penetrating Radar B-Scan Image Reconstruction
Huabin Wang (Anhui University), Zilong Ling (Anhui University)
RestorationGenerative Adversarial NetworkImagePhysics Related
🎯 What it does: Propose an MGT-Net based on multi-window Gabor transform and defect guidance, capable of automatically enhancing gain and defects in original GPR B-Scan images;
MULTIBENCH++: A Unified and Comprehensive Multimodal Fusion Benchmarking Across Specialized Domains
Leyan Xue (Tianjin University), Zongbo Han (Beijing University of Posts and Telecommunications)
Hyperparameter SearchTransformerMultimodalityBenchmark
🎯 What it does: Built MULTIBENCH++, a unified large-scale multimodal fusion benchmark covering 30+ datasets, 15+ modalities, and 20+ tasks, along with an automated evaluation pipeline and standard implementations.
Multifaceted Scenario-Aware Hypergraph Learning for Next POI Recommendation
Yuxi Lin (Jilin University), Zipei Fan (Jilin University)
Recommendation SystemGraph Neural NetworkContrastive LearningGraphSequential
🎯 What it does: Proposed a multi-context-aware hypergraph learning framework named MSAHG for next POI recommendation.
Multigranular Evaluation for Brain Visual Decoding
Weihao Xia (University of Cambridge), Cengiz Oztireli (University of Cambridge)
Explainability and InterpretabilityGraph Neural NetworkLarge Language ModelVision Language ModelVideoMultimodalityBiomedical DataMagnetic Resonance Imaging
🎯 What it does: Propose the BASIC framework to perform multi-grained structural, reasoning, and contextual three-dimensional evaluation of brain vision decoding results.
MultiKD: Backdoor Defense in Federated Graph Learning via Attention-Guided Multi-Teacher Distillation
Jiale Zhang (Yangzhou University), Yu Li (Yangzhou University)
Federated LearningSafty and PrivacyKnowledge DistillationGraph Neural NetworkGraph
🎯 What it does: Proposes a server-side defense framework called MULTIKD based on multi-teacher attention-guided distillation to eliminate backdoor attacks in federated graph learning.
MultiMedBench: A Scenario-Aware Benchmark for Evaluating Knowledge Editing in Medical VQA
Shengtao Wen (Nanjing University of Aeronautics and Astronautics), Sheng-Jun Huang (Tsinghua University)
Prompt EngineeringMultimodalityBiomedical DataBenchmark
🎯 What it does: Proposes MultiMedBench, a VQA benchmark specifically designed for medical multimodal knowledge editing, covering two tasks: understanding and reasoning.
Multimodal DeepResearcher: Generating Text-Chart Interleaved Reports from Scratch with Agentic Framework
Zhaorui Yang (Zhejiang University), Wei Chen (Zhejiang University)
GenerationLarge Language ModelAgentic AIMultimodalityBenchmark
🎯 What it does: Proposes the Multimodal DeepResearcher framework, capable of generating interactive reports containing text and multiple visualizations from scratch.
Multimodal Gaussian Mixture Variational Autoencoder with Consistency Regularizations
Yarui Chen (Tianjin University of Science and Technology), Yancui Shi (Tianjin University of Science and Technology)
ClassificationImage TranslationGenerationRetrievalMixture of ExpertsAuto EncoderContrastive LearningImageMultimodality
🎯 What it does: Propose a Multimodal Gaussian Mixture Variational Autoencoder (MGMVAE), which separates shared and private latent spaces, employs a Gaussian Mixture Prior, and combines cluster-guided and self-supervised contrast regularization to achieve cross-modal consistency.
Multimodal Graph Representation Learning with Dynamic Information Pathways
Xiaobin Hong (Nanjing University), Wenzhong Li (Nanjing Forest University)
Representation LearningGraph Neural NetworkVision Language ModelMultimodalityGraph
🎯 What it does: Propose a multimodal graph representation learning framework based on dynamic information paths (DiP), which utilizes modality-specific pseudo nodes to achieve adaptive information passing within and across modalities.
Multimodal Mixture-of-Experts with Retrieval Augmentation for Protein Active Site Identification
Jiayang Wu (Westlake University), Yefeng Zheng (Jiangnan University)
Protein Structure PredictionTransformerMixture of ExpertsTextMultimodalitySequentialBiomedical DataRetrieval-Augmented Generation
🎯 What it does: Proposes a retrieval-enhanced multimodal expert mixture framework called MERA for residue-level identification of protein active sites.