AAAI 2024 Papers — Page 14
AAAI Conference on Artificial Intelligence · 2331 papers
MeDM: Mediating Image Diffusion Models for Video-to-Video Translation with Temporal Correspondence Guidance
Ernie Chu (Academia Sinica), Jun-Cheng Chen (Academia Sinica)
Image TranslationGenerationDiffusion modelOptical FlowVideoStochastic Differential Equation
🎯 What it does: Proposes the MeDM method, which utilizes a pre-trained image diffusion model to achieve high-quality and temporally consistent video editing and rendering in video-to-video conversion.
MedSegDiff-V2: Diffusion-Based Medical Image Segmentation with Transformer
Junde Wu (University of Oxford), Yanwu Xu (National University of Singapore)
SegmentationTransformerDiffusion modelImageMultimodalityMagnetic Resonance ImagingComputed TomographyPositron Emission TomographyUltrasound
🎯 What it does: A diffusion model MedSegDiff-V2 that integrates Transformer has been developed for multimodal medical image segmentation.
MELO: Enhancing Model Editing with Neuron-Indexed Dynamic LoRA
Lang Yu (East China Normal University), Liang He (East China Normal University)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: A model editing framework MELO based on neuron-indexed dynamic LoRA is proposed, which can achieve multi-attribute editing on LLMs and is easy to integrate.
Memory Asymmetry Creates Heteroclinic Orbits to Nash Equilibrium in Learning in Zero-Sum Games
Yuma Fujimoto (SOKENDAI), Kenshi Abe (University of Electro-Communications)
OptimizationReinforcement Learning from Human FeedbackReinforcement Learning
🎯 What it does: The study investigates the learning dynamics generated by the gradient ascent learning algorithm (MMGA) in a zero-sum game with memory asymmetry (for example, one party uses memory while the other does not) and proves the existence of heteroclinic orbits that lead learning to ultimately converge to the Nash equilibrium of the original memoryless game.
Memory-Efficient Prompt Tuning for Incremental Histopathology Classification
Yu Zhu (Chinese University of Hong Kong), Pheng Ann Heng
ClassificationDomain AdaptationComputational EfficiencyTransformerPrompt EngineeringBiomedical Data
🎯 What it does: A memory-efficient prompt tuning framework is proposed for incremental learning in different pathological domains as sequences arrive, avoiding model forgetting and enhancing generalization ability.
Memory-Efficient Reversible Spiking Neural Networks
Hong Zhang (Zhejiang University), Yu Zhang (Zhejiang University)
Spiking Neural NetworkTransformerImage
🎯 What it does: Proposes reversible spiking neural networks (RevSResNet and RevSFormer), which only save outputs during training and reconstruct intermediate activations and membrane potentials through a reverse process, significantly reducing memory usage.
MemoryBank: Enhancing Large Language Models with Long-Term Memory
Wanjun Zhong (Sun Yat-Sen University), Yanlin Wang (Sun Yat-Sen University)
RetrievalTransformerLarge Language ModelSupervised Fine-TuningTextMultimodalityRetrieval-Augmented Generation
🎯 What it does: A long-term memory mechanism called MemoryBank was designed and implemented for large language models (LLM) and integrated into the AI companion chatbot SiliconFriend to enable retrieval, updating of historical conversations, and user profile construction.
MEPSI: An MDL-Based Ensemble Pruning Approach with Structural Information
Xiao-Dong Bi (Nanjing University), Yuan Jiang (Nanjing University)
OptimizationImageTabular
🎯 What it does: This paper proposes a new ensemble pruning method called MEPSI, which models the pruning task as a bi-objective optimization problem that minimizes empirical error and the structural information of individual learners using Kolmogorov complexity and the MDL principle.
MERGE: Fast Private Text Generation
Zi Liang (Xi'an Jiaotong University), Ziyang Zhou (Xi'an Jiaotong University)
GenerationSafty and PrivacyComputational EfficiencyTransformerLarge Language ModelText
🎯 What it does: The MERGE framework is proposed to accelerate Transformer text generation in the MPC environment, addressing the slow issues of embedding queries and autoregressive generation.
MESED: A Multi-Modal Entity Set Expansion Dataset with Fine-Grained Semantic Classes and Hard Negative Entities
Yangning Li (Shenzhen International Graduate School Tsinghua University), Rui Zhang (Shenzhen International Graduate School Tsinghua University)
RetrievalRepresentation LearningTransformerContrastive LearningImageTextMultimodality
🎯 What it does: This paper proposes the Multi-modal Entity Set Expansion (MESE) task, which utilizes text and image representations of entities and achieves entity expansion through self-supervised pre-training and deep modal fusion.
Meta-Inverse Reinforcement Learning for Mean Field Games via Probabilistic Context Variables
Yang Chen (University of Auckland), Michael Witbrock (University of Auckland)
Meta LearningReinforcement LearningTabularTime SeriesFinance Related
🎯 What it does: In average field games, a meta-inverse reinforcement learning framework based on probabilistic context variables is proposed to simultaneously infer the reward function and task type from mixed-type demonstrations.
Meta-Learning-Based Adaptive Stability Certificates for Dynamical Systems
Amit Jena (Texas A&M University), Le Xie (Texas A&M University)
Meta LearningTime Series
🎯 What it does: This paper proposes a meta-learning based neural Lyapunov function (meta-NLF) that can quickly adapt to new dynamic models and provide stability region (ROA) estimates when system parameters change over time.
Meta-Reinforcement Learning via Exploratory Task Clustering
Zhendong Chu (University of Virginia), Hongning Wang (University of Virginia)
Meta LearningRecurrent Neural NetworkReinforcement LearningSequential
🎯 What it does: This paper proposes a method called MILET to enhance sample efficiency in meta reinforcement learning through task clustering, utilizing a structured heterogeneous task distribution to achieve cluster-level knowledge sharing.
MetaCARD: Meta-Reinforcement Learning with Task Uncertainty Feedback via Decoupled Context-Aware Reward and Dynamics Components
Min Wang (Beijing Institute of Technology), Mingzhong Wang (University of the Sunshine Coast)
Meta LearningReinforcement LearningSequential
🎯 What it does: The MetaCARD method is proposed, utilizing context-aware rewards and decoupling dynamics to achieve rapid adaptation in Meta-RL.
MetaDiff: Meta-Learning with Conditional Diffusion for Few-Shot Learning
Baoquan Zhang (Harbin Institute of Technology), Bowen Zhang (Shenzhen Technology University)
ClassificationMeta LearningDiffusion modelImage
🎯 What it does: A conditional diffusion model meta-learning framework named MetaDiff is designed for rapid learning of new classification tasks from a small number of samples.
MetaMix: Meta-State Precision Searcher for Mixed-Precision Activation Quantization
Han-Byul Kim (Seoul National University), Hong-Seok Kim (Google)
OptimizationMeta LearningConvolutional Neural NetworkImage
🎯 What it does: This paper proposes MetaMix—a two-stage mixed-precision activation quantization training framework that first uses meta-learning to obtain stable activation distributions, then determines the activation bit-width for each layer through fixed-weight bit-width search, and finally fine-tunes the weights.
MetaRLEC: Meta-Reinforcement Learning for Discovery of Brain Effective Connectivity
Zuozhen Zhang (Beijing University of Technology), Jinduo Liu (Beijing University of Technology)
Meta LearningRecurrent Neural NetworkTransformerReinforcement LearningTime SeriesBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A method for discovering effective brain connectivity (EC) based on meta reinforcement learning, called MetaRLEC, is proposed, which utilizes the actor-critic framework and a meta-critic to enhance learning efficiency on small sample high noise fMRI data.
MFABA: A More Faithful and Accelerated Boundary-Based Attribution Method for Deep Neural Networks
Zhiyu Zhu (University of Sydney), Kim-Kwang Raymond Choo (University of Texas at San Antonio)
ClassificationExplainability and InterpretabilityComputational EfficiencyAdversarial AttackConvolutional Neural NetworkImage
🎯 What it does: A more faithful and faster boundary-based attribution method, MFABA, is proposed, utilizing second-order Taylor expansion and adversarial attack gradients for efficient feature attribution.
MFOS: Model-Free & One-Shot Object Pose Estimation
JongMin Lee (Seoul National University), Jerome Revaud (Naver Labs Europe)
Pose EstimationTransformerImage
🎯 What it does: A MFOS method based on Vision Transformer is proposed, which can estimate the 6D pose of target objects that were never seen during the training phase, using only the query image and a small number of reference images with known poses, without the need for a 3D model.
MFTN: Multi-Level Feature Transfer Network Based on MRI-Transformer for MR Image Super-resolution
Shuying Huang (Tiangong University), Chenbin Liang (Tiangong University)
RestorationSuper ResolutionTransformerContrastive LearningImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A multi-layer feature transfer network (MFTN) based on MRI-Transformer is proposed to achieve super-resolution reconstruction of low-resolution MRI.
MGNet: Learning Correspondences via Multiple Graphs
Dai Luanyuan, Jinhui Tang (Nanyang Technological University)
Pose EstimationGraph Neural NetworkImage
🎯 What it does: This study focuses on outlier removal in sparse correspondences and proposes MGNet, which simultaneously constructs implicit and explicit local graphs and introduces Graph Soft Degree Attention to achieve high-quality correspondence filtering.
MGQFormer: Mask-Guided Query-Based Transformer for Image Manipulation Localization
Kunlun Zeng (Fudan University), Bo Yan (Fudan University)
Image TranslationObject DetectionTransformerImage
🎯 What it does: This paper proposes the Mask-Guided Query-based Transformer (MGQFormer) for efficiently and accurately locating image tampering regions.
MICA: Towards Explainable Skin Lesion Diagnosis via Multi-Level Image-Concept Alignment
Yequan Bie (Hong Kong University of Science and Technology), Hao Chen (Hong Kong University of Science and Technology)
ClassificationExplainability and InterpretabilityConvolutional Neural NetworkLarge Language ModelContrastive LearningImageMultimodality
🎯 What it does: A multi-modal interpretable skin disease diagnosis framework MICA is proposed, which achieves interpretable diagnosis through multi-layer image-concept alignment.
MID-FiLD: MIDI Dataset for Fine-Level Dynamics
Jesung Ryu (Pozalabs), Taehyun Kim (Duke University)
GenerationTransformerAudio
🎯 What it does: This paper presents the MID-FiLD dataset and a Transformer-based generative model to learn and generate fine-grained MIDI expressive dynamics.
Mimic: Speaking Style Disentanglement for Speech-Driven 3D Facial Animation
Hui Fu (South China University of Technology), Wenxiong Kang (South China University of Technology)
GenerationData SynthesisTransformerContrastive LearningVideoMesh
🎯 What it does: This paper presents the Mimic framework, which generates 3D facial animations that match the speaker and have high synchronization by separating speaking style from semantic content in 3D facial animation using short video references.
MIND: Multi-Task Incremental Network Distillation
Jacopo Bonato (Leonardo Labs), Alessandro Nicolosi (Leonardo Labs)
Knowledge DistillationImage
🎯 What it does: Designed and implemented MIND (Multi-Task Incremental Network Distillation), a replay-free, parameter-isolation-based incremental learning framework that achieves continuous learning by training a brand new model for each new task or using self-distillation, and then distilling its knowledge into the current sub-network.
MindMap: Constructing Evidence Chains for Multi-Step Reasoning in Large Language Models
Yangyu Wu (Capital Normal University), Fei Li (Wuhan University)
TransformerLarge Language ModelTextChain-of-Thought
🎯 What it does: The MindMap method is proposed, which organizes facts into theme-related evidence chains and utilizes LLM to generate summaries to support multi-step reasoning.
MINES: Message Intercommunication for Inductive Relation Reasoning over Neighbor-Enhanced Subgraphs
Ke Liang (National University of Defense Technology), Xinwang Liu (National University of Defense Technology)
Representation LearningGraph Neural NetworkGraph
🎯 What it does: A graph neural network framework named MINES has been designed and implemented for inductive relation reasoning in knowledge graphs. This framework improves the traditional GraIL model through neighbor-enhanced subgraphs and a message intercommunication mechanism.
Minibatch Stochastic Three Points Method for Unconstrained Smooth Minimization
Soumia Boucherouite (Mohammed VI Polytechnic University), El Houcine Bergou (Mohammed VI Polytechnic University)
OptimizationAdversarial AttackImageTabular
🎯 What it does: This paper proposes the Minibatch Stochastic Three Points (MiSTP) zero-order optimization method, which updates parameters using random search directions and sub-batch objective function approximations, and provides theoretical complexity and experimental validation in both non-convex and convex scenarios.
Minimal Macro-Based Rewritings of Formal Languages: Theory and Applications in Ontology Engineering (and Beyond)
Christian Kindermann (Stanford University), Uli Sattler (University of Manchester)
OptimizationComputational EfficiencyBiomedical Data
🎯 What it does: A finite form language rewriting framework based on nested grammar macros is proposed, providing polynomial time algorithms for various problems and applying it to the minimization rewriting of large biomedical OWL ontologies.
Minimum Coverage Sets for Training Robust Ad Hoc Teamwork Agents
Muhammad Rahman (University of Texas at Austin), Peter Stone (University of Texas at Austin)
OptimizationRobotic IntelligenceReinforcement LearningSequential
🎯 What it does: A new partner generation method L-BRDiv is studied to train adaptive team working (AHT) agents that are robust to unseen partners.
Mining Fine-Grained Image-Text Alignment for Zero-Shot Captioning via Text-Only Training
Longtian Qiu (ShanghaiTech University), Xuming He (ShanghaiTech University)
GenerationRetrievalTransformerLarge Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: A zero-shot image captioning framework based on CLIP, called MacCap, is proposed. It utilizes text data to train an adapter that only uses text, enabling the model to generate natural language descriptions directly from CLIP image embeddings, and extends this framework to zero-shot VQA.
Mining Gaze for Contrastive Learning toward Computer-Assisted Diagnosis
Zihao Zhao (ShanghaiTech University), Dinggang Shen (Shanghai United Imaging Intelligence Co., Ltd.)
ClassificationRepresentation LearningContrastive LearningImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A comparative learning pre-training method based on radiologists' eye movement data, McGIP, is proposed, which constructs positive sample pairs using eye movement similarity.
Mitigating Idiom Inconsistency: A Multi-Semantic Contrastive Learning Method for Chinese Idiom Reading Comprehension
Mingmin Wu (Huazhong Agricultural University), Ying Sha (Huazhong Agricultural University)
Contrastive LearningText
🎯 What it does: This paper proposes a Multi-Semantic Contrastive Learning Model (MSCLM) to address the issues of metaphor inconsistency and contextual inconsistency in the reading comprehension of Chinese idioms.
Mitigating Label Bias in Machine Learning: Fairness through Confident Learning
Yixuan Zhang (Hangzhou Dianzi University), Feng Zhou (Renmin University of China)
TabularFinance Related
🎯 What it does: This paper proposes a data filtering framework based on confident learning, utilizing truncated confidence thresholds and co-teaching to eliminate unfairness in data with biased labels.
Mitigating Label Noise through Data Ambiguation
Julian Lienen (Paderborn University), Eyke Hüllermeier (Ludwig Maximilian University of Munich)
ClassificationContrastive LearningImage
🎯 What it does: A method is designed to alleviate training set noise by 'fuzzifying' labels, using adaptive thresholds to generate reliable candidate labels and learning based on confidence sets.
Mitigating Large Language Model Hallucinations via Autonomous Knowledge Graph-Based Retrofitting
Xinyan Guan (Chinese Information Processing Laboratory, Institute of Software, Chinese Academy of Sciences), Le Sun (Chinese Information Processing Laboratory, Institute of Software, Chinese Academy of Sciences)
GenerationRetrievalTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: A post-repair framework based on knowledge graphs, KGR, has been designed and implemented, utilizing LLM to automatically extract, verify, and rewrite model-generated answers, thereby suppressing factual hallucinations.
Mitigating the Impact of False Negative in Dense Retrieval with Contrastive Confidence Regularization
Shiqi Wang (Nanjing University), Cam-Tu Nguyen (Nanjing University)
RetrievalContrastive LearningText
🎯 What it does: This paper addresses the issue of false negative samples caused by incomplete annotations in dense retrieval, proposing a robust NCE loss based on contrastive confidence regularization and a general passage sieve filtering algorithm to filter out noisy negative samples and improve retrieval quality.
Mixed Geometry Message and Trainable Convolutional Attention Network for Knowledge Graph Completion
Bin Shang (Xi'an Jiaotong University), Di Wang (Xidian University)
Graph Neural NetworkGraph
🎯 What it does: A knowledge graph completion model MGTCA is proposed, which combines hybrid geometric message passing with a trainable convolutional attention network.
Mixed-Effects Contextual Bandits
Kyungbok Lee (Seoul National University), Gi-Soo Kim (Ulsan National Institute of Science and Technology)
Recommendation SystemOptimizationReinforcement LearningTabular
🎯 What it does: A mixed-effects contextual multi-armed bandit model is proposed and analyzed, and two ME-CUCB algorithms are designed (for known and unknown covariance scenarios), providing an approximately optimal regret upper bound under this model.
Mixup-Induced Domain Extrapolation for Domain Generalization
Meng Cao (Nanjing University of Aeronautics and Astronautics), Songcan Chen (Nanjing University of Aeronautics and Astronautics)
Domain AdaptationContrastive LearningImage
🎯 What it does: This paper proposes an extrapolation strategy based on Mixup, called EDM, to expand the source domain space by generating extrapolated domains in domain generalization tasks, thereby enhancing the model's generalization ability.
MKG-FENN: A Multimodal Knowledge Graph Fused End-to-End Neural Network for Accurate Drug–Drug Interaction Prediction
Di Wu (Chongqing University of Posts and Telecommunications), Xin Luo (Southern Illinois University)
Drug DiscoveryGraph Neural NetworkMultimodalityGraph
🎯 What it does: A multi-modal knowledge graph fusion end-to-end neural network (MKG-FENN) is proposed for precise prediction of drug-drug interaction events.
MLNet: Mutual Learning Network with Neighborhood Invariance for Universal Domain Adaptation
Yanzuo Lu (Sun Yat-sen University), Jian-Huang Lai (Sun Yat-sen University)
Domain AdaptationKnowledge DistillationImage
🎯 What it does: MLNet is proposed, combining neighborhood invariance and cross-domain Manifold Mixup to achieve universal domain adaptation.
MM-Point: Multi-View Information-Enhanced Multi-Modal Self-Supervised 3D Point Cloud Understanding
Hai-Tao Yu (Southeast University), Mofei Song (Southeast University)
ClassificationSegmentationRepresentation LearningContrastive LearningMultimodalityPoint Cloud
🎯 What it does: This paper proposes a self-supervised point cloud representation learning framework called MM-Point based on multimodal contrastive learning, which jointly learns from multi-view 2D images and 3D point clouds of the same 3D object to better capture spatial structural information.
MM-TTS: Multi-Modal Prompt Based Style Transfer for Expressive Text-to-Speech Synthesis
Wenhao Guan (Xiamen University), Qingyang Hong (Xiamen University)
GenerationData SynthesisPrompt EngineeringRectified FlowMultimodalityAudio
🎯 What it does: MM-TTS is proposed, a multimodal speech synthesis framework that can use any modality (voice, image, text) as style prompts;
MmAP: Multi-Modal Alignment Prompt for Cross-Domain Multi-Task Learning
Yi Xin (Nanjing University), Shouhong Ding (Tencent)
ClassificationDomain AdaptationTransformerPrompt EngineeringContrastive LearningImageMultimodality
🎯 What it does: A multi-modal alignment prompt (MmAP) is designed, and a CLIP-based multi-task prompt learning framework is constructed for multi-task learning in cross-domain image classification.
MobileInst: Video Instance Segmentation on the Mobile
Renhong Zhang (Huazhong University of Science and Technology), Xinggang Wang (Huazhong University of Science and Technology)
Object TrackingSegmentationTransformerVideo
🎯 What it does: A lightweight mobile video instance segmentation framework called MobileInst is proposed, capable of achieving real-time instance segmentation and tracking on CPU.
MoDE: A Mixture-of-Experts Model with Mutual Distillation among the Experts
Zhitian Xie (Ant Group), Guannan Zhang (Ant Group)
ClassificationKnowledge DistillationConvolutional Neural NetworkTransformerMixture of ExpertsImageTextTabular
🎯 What it does: This paper proposes a training method that introduces mutual distillation in Mixture-of-Experts (MoE) called MoDE, to alleviate the 'narrow view' problem of experts.
Model Counting and Sampling via Semiring Extensions
Andreas Goral (Friedrich Schiller University Jena), Julien Klaus (Friedrich Schiller University Jena)
OptimizationGraph
🎯 What it does: This paper proposes a general framework based on selective semiring extension to transform MPF (Marginalize a Product Function) queries into model counting and model sampling MPF queries, thereby achieving a unified solution.
Model-Driven Deep Neural Network for Enhanced AoA Estimation Using 5G gNB
Shengheng Liu (Southeast University), Yongming Huang (Purple Mountain Laboratories)
Pose EstimationOptimizationConvolutional Neural NetworkTime Series
🎯 What it does: A model-driven deep neural network (MoD-DNN) framework is proposed for implementing angle of arrival (AoA) estimation and correcting hardware distortion on 5G gNB, significantly improving positioning accuracy.
Modeling Adaptive Inter-Task Feature Interactions via Sentiment-Aware Contrastive Learning for Joint Aspect-Sentiment Prediction
Wei Chen (Beihang University), Jiang Zhong (Chongqing University)
ClassificationRecommendation SystemTransformerMixture of ExpertsContrastive LearningText
🎯 What it does: In joint sentiment polarity analysis, the AIFI framework achieves feature alignment between aspect and sentiment prediction tasks through contrastive learning, enabling joint learning of the two tasks.
Modeling Continuous Motion for 3D Point Cloud Object Tracking
Zhipeng Luo (Nanyang Technological University), Shijian Lu (Nanyang Technological University)
Object TrackingAutonomous DrivingTransformerContrastive LearningPoint Cloud
🎯 What it does: The StreamTrack framework is proposed, which utilizes a memory bank to continuously store multi-frame features and achieves 3D single-object tracking by only inputting the current frame.
Modeling Knowledge Graphs with Composite Reasoning
Wanyun Cui (Shanghai University of Finance and Economics), Linqiu Zhang (Shanghai University of Finance and Economics)
Recommendation SystemGraph Neural NetworkGraphBenchmark
🎯 What it does: In the task of knowledge graph completion, a composite reasoning framework is introduced to unify the interpretation of various models, revealing that tensor decomposition-based models erroneously incorporate irrelevant entities into reasoning, leading to the failure of the collaborative filtering assumption. The 'combinatorial risk' is proposed and optimized to reduce erroneous reasoning, thereby enhancing model performance.
Modeling Stereo-Confidence out of the End-to-End Stereo-Matching Network via Disparity Plane Sweep
Jae Young Lee (Korea Advanced Institute of Science and Technology), Junmo Kim (Korea Advanced Institute of Science and Technology)
Depth EstimationAutonomous DrivingConvolutional Neural NetworkImage
🎯 What it does: This paper proposes an external stereo confidence measurement method that generates a disparity volume using disparity plane scanning and measures the predicted disparity deviation based on an ideal disparity line.
ModWaveMLP: MLP-Based Mode Decomposition and Wavelet Denoising Model to Defeat Complex Structures in Traffic Forecasting
Ke Sun (Central South University), Zhifang Liao
Time Series
🎯 What it does: Designed and implemented the ModWaveMLP model based on MLP, using pattern decomposition and wavelet denoising for traffic prediction.
Molecular Optimization Model with Patentability Constraint
Sally Turutov (Technion Israel Institute of Technology), Kira Radinsky (Technion Israel Institute of Technology)
OptimizationDrug DiscoveryAuto EncoderGraph
🎯 What it does: A molecular generation framework based on SMILES, called MOMP, is proposed, which jointly optimizes molecular properties and similarity to existing patented molecules (PL), achieving molecular design that improves drug properties while reducing the risk of patent infringement.
MolTailor: Tailoring Chemical Molecular Representation to Specific Tasks via Text Prompts
Haoqiang Guo (Harbin Institute of Technology), Bing Qin (Harbin Institute of Technology)
Drug DiscoveryTransformerLarge Language ModelPrompt EngineeringTextGraph
🎯 What it does: Combining language models with molecular pre-training models through a dual-tower architecture, utilizing text prompts to adaptively generate task-specific molecular representations.
Mono3DVG: 3D Visual Grounding in Monocular Images
Yang Zhan (Northwestern Polytechnical University), Zhitong Xiong (Technical University of Munich)
Object DetectionDepth EstimationAutonomous DrivingConvolutional Neural NetworkTransformerImageTextMultimodality
🎯 What it does: Achieve 3D visual localization based on natural language descriptions containing appearance and geometric information in monocular RGB images.
Monocular 3D Hand Mesh Recovery via Dual Noise Estimation
Hanhui Li (Shenzhen Campus of Sun Yat-sen University), Xiaodan Liang (Tencent)
RecognitionPose EstimationConvolutional Neural NetworkMesh
🎯 What it does: A monocular hand mesh recovery method based on dual noise estimation is proposed, which gradually refines the mesh and corrects camera parameters based on benchmark parametric fitting.
Monte Carlo Tree Search in the Presence of Transition Uncertainty
Farnaz Kohankhaki (University of Alberta), Martin Müller (University of Alberta)
Reinforcement Learning
🎯 What it does: The study investigates Monte Carlo Tree Search (MCTS) under model incompleteness and proposes UA-MCTS to leverage estimated uncertainty to guide the search.
Moral Uncertainty and the Problem of Fanaticism
Jazon Szabo (King's College London), Sanjay Modgil (King's College London)
🎯 What it does: This study formally defines 'fanaticism' within the framework of moral uncertainty and proposes improvements to the existing Maximizing Expected Choice (MEC) method, providing two non-fanatical aggregation methods—weighted k-trimmed mean (k-thm) and weighted maximum median (hm).
MorphVAE: Advancing Morphological Design of Voxel-Based Soft Robots with Variational Autoencoders
Junru Song (Renmin University of China), Wen Yao (Chinese Academy of Military Science)
OptimizationRobotic IntelligenceReinforcement LearningAuto EncoderTabular
🎯 What it does: This paper proposes MorphVAE, a soft robot morphology generation model based on variational autoencoders, which achieves the co-evolution of morphology and control through multi-task training and continuous natural selection sampling.
Motif-Aware Riemannian Graph Neural Network with Generative-Contrastive Learning
Li Sun (North China Electric Power University), Philip S. Yu
ClassificationRepresentation LearningGraph Neural NetworkGenerative Adversarial NetworkContrastive LearningGraph
🎯 What it does: This paper proposes MotifRGC, a framework for self-supervised graph representation learning on multi-curvature Riemannian manifolds, capable of capturing motif patterns in graphs.
Motion Deblurring via Spatial-Temporal Collaboration of Frames and Events
Wen Yang (Xidian University), Guangming Shi (Xidian University)
RestorationConvolutional Neural NetworkImageVideoMultimodality
🎯 What it does: This paper proposes STCNet, which utilizes the spatial-temporal collaborative fusion of event cameras and frame images to achieve motion deblurring.
MotionGPT: Finetuned LLMs Are General-Purpose Motion Generators
Yaqi Zhang (University of Science and Technology of China), Wanli Ouyang (Shanghai AI Laboratory)
GenerationPose EstimationRobotic IntelligenceTransformerLarge Language ModelSupervised Fine-TuningTextMultimodality
🎯 What it does: By fine-tuning a large language model, continuous human actions are generated using text and single-frame poses as multimodal control signals;
MotionMix: Weakly-Supervised Diffusion for Controllable Motion Generation
Nhat M. Hoang (Nanyang Technological University), Michael Bi Mi (Huawei Technologies)
GenerationData SynthesisDiffusion modelVideoTextMultimodality
🎯 What it does: This paper presents MotionMix, a weakly supervised diffusion model that trains on both noisy labeled and unlabeled action sequences simultaneously to achieve controllable human action generation.
MSGNet: Learning Multi-Scale Inter-series Correlations for Multivariate Time Series Forecasting
Wanlin Cai (Sichuan University), Yuankai Wu (Beijing Institute of Technology)
Graph Neural NetworkTransformerTime Series
🎯 What it does: A multivariate time series forecasting model MSGNet based on multi-scale frequency domain analysis and adaptive graph convolution is proposed, which can capture the interrelationships of multiple series at different time scales.
msLPCC: A Multimodal-Driven Scalable Framework for Deep LiDAR Point Cloud Compression
Miaohui Wang (Shenzhen University), Wuyuan Xie
SegmentationCompressionAutonomous DrivingTransformerContrastive LearningMultimodalityPoint Cloud
🎯 What it does: This paper proposes a multi-modal driven scalable LiDAR point cloud compression framework msLPCC, which achieves hierarchical encoding of large-scale, sparse point clouds and enhances compression quality through depth maps and semantic segmentation information.
Multi-Architecture Multi-Expert Diffusion Models
Yunsung Lee (Wrtn Technologies), Seungtaek Choi (Yanolja)
GenerationData SynthesisComputational EfficiencyConvolutional Neural NetworkTransformerMixture of ExpertsDiffusion modelImage
🎯 What it does: This paper studies a multi-architecture multi-expert diffusion model (MEME) that enhances model efficiency and generation quality by using different network architectures across different time step intervals.
Multi-Class Support Vector Machine with Maximizing Minimum Margin
Feiping Nie (Northwestern Polytechnical University), Rong Wang (Northwestern Polytechnical University)
ClassificationOptimizationTabular
🎯 What it does: A multi-class support vector machine (M3SVM) is proposed to enhance classification performance by maximizing the minimum margin between all classes.
Multi-Constellation-Inspired Single-Shot Global LiDAR Localization
Tongzhou Zhang (Jilin University), Jue Hu (Harbin Institute of Technology)
Pose EstimationAutonomous DrivingOptimizationSimultaneous Localization and MappingPoint Cloud
🎯 What it does: A multi-constellation heuristic single global LiDAR positioning method is proposed: first, global descriptors are used to roughly retrieve keyframes, then several nearby observation points are selected, and a lightweight LiDAR odometer is used to estimate the distance to the observation points. Finally, the position estimate is transformed into a multi-sphere equation solution to obtain the three-dimensional pose.
Multi-Cross Sampling and Frequency-Division Reconstruction for Image Compressed Sensing
Heping Song (Jiangsu University), Yuping Lai (Beijing University of Posts and Telecommunications)
RestorationCompressionConvolutional Neural NetworkImage
🎯 What it does: This paper proposes an end-to-end multi-level cross-sampling and frequency division reconstruction network (MCFDNet), which achieves efficient image sampling through the DMCS module and high-low frequency reconstruction based on discrete wavelet decomposition through the FDRM module, thereby improving the quality of image compressed sensing.
Multi-Dimensional Fair Federated Learning
Cong Su (Shandong University), Han Yu (Nanyang Technological University)
OptimizationFederated LearningTabular
🎯 What it does: This paper proposes a multidimensional fair federated learning method mFairFL, aimed at achieving both group fairness and client fairness in a decentralized data environment.
Multi-Domain Deep Learning from a Multi-View Perspective for Cross-Border E-commerce Search
Yiqian Zhang (Hangzhou Dianzi University), Jun Yu (Hangzhou Dianzi University)
Recommendation SystemTabular
🎯 What it does: A DAMO model based on low-rank multi-view space is proposed for CTR/CVR prediction in cross-border e-commerce search.
Multi-Domain Incremental Learning for Face Presentation Attack Detection
Keyao Wang (Baidu Inc), Jingdong Wang (Chinese Academy of Sciences)
ClassificationDomain AdaptationTransformerImage
🎯 What it does: A multi-domain incremental learning framework (MDIL) is proposed to learn new domains without accessing old domain data while maintaining detection performance on old domains.
Multi-Domain Multi-Scale Diffusion Model for Low-Light Image Enhancement
Kai Shang (China University of Petroleum), Shuigen Wang (University of Technology Sydney)
RestorationDiffusion modelImage
🎯 What it does: This paper proposes a Multi-Domain Multi-Scale Diffusion Model (MDMS) that simultaneously utilizes spatial and frequency domain information for low-light image enhancement tasks, significantly reducing checkerboard artifacts through multi-scale sampling techniques.
Multi-Domain Recommendation to Attract Users via Domain Preference Modeling
Hyunjun Ju (Pohang University of Science and Technology), Hwanjo Yu (Yonsei University)
Recommendation SystemTransformerTabular
🎯 What it does: A unified framework called DRIP is proposed for the multi-domain recommendation attracting users (MDRAU) task, which can recommend items in domains that users have not experienced.
Multi-Energy Guided Image Translation with Stochastic Differential Equations for Near-Infrared Facial Expression Recognition
Bingjun Luo (Tsinghua University), Yue Gao (Tsinghua University)
RecognitionImage TranslationDiffusion modelScore-based ModelImageStochastic Differential Equation
🎯 What it does: This paper proposes an energy-guided conditional SDE model, NFER-SDE, which converts visible light facial expression images into near-infrared images to enhance NIR FER performance.
Multi-Granularity Causal Structure Learning
Jiaxuan Liang (Shandong University), Guoyin Wang (Chongqing University of Posts and Telecommunications)
OptimizationExplainability and InterpretabilityComputational EfficiencyAuto EncoderGraphTime SeriesMagnetic Resonance Imaging
🎯 What it does: This study investigates multi-granularity causal structure learning and proposes the MgCSL method, which uses sparse autoencoders to automatically coarsen micro variables into macro variables, and then employs multilayer perceptrons and simplified acyclic constraints to learn the causal network between micro and macro variables.
Multi-Label Supervised Contrastive Learning
Pingyue Zhang (Shanghai Jiao Tong University), Mengyue Wu (Shanghai Jiao Tong University)
ClassificationRepresentation LearningConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: A multi-label supervised contrastive learning method called MulSupCon is proposed, which constructs a weighted positive sample set for each sample based on the label overlap ratio to improve the pre-training and fine-tuning processes of multi-label classification.
Multi-Level Cross-Modal Alignment for Image Clustering
Liping Qiu (Shenzhen University), Shaotian Cai (Shenzhen University)
ClassificationRepresentation LearningTransformerContrastive LearningImageMultimodality
🎯 What it does: Utilizing a cross-modal pre-training model (CLIP) to generate pseudo-labels, a multi-level cross-modal alignment (instance, prototype, semantic) and semantic space construction are proposed to enhance image clustering performance.
Multi-Modal Disordered Representation Learning Network for Description-Based Person Search
Fan Yang (Sichuan University), Jianwei Zhang (Sichuan University)
RetrievalRepresentation LearningTransformerVision Language ModelImageTextMultimodality
🎯 What it does: A multi-modal disorder representation learning network (MDRL) is proposed for description-based pedestrian retrieval.
Multi-Modal Latent Space Learning for Chain-of-Thought Reasoning in Language Models
Liqi He (Wuhan University), Ping Wang (Wuhan University)
Image TranslationGenerationData SynthesisTransformerLarge Language ModelVision Language ModelDiffusion modelAuto EncoderImageTextMultimodalityChain-of-Thought
🎯 What it does: A multimodal chain reasoning model DPMM-CoT based on diffusion latent space learning has been developed, which enhances the multimodal reasoning ability of large language models by utilizing the alignment features of text and images.
Multi-Modal Prompting for Open-Vocabulary Video Visual Relationship Detection
Shuo Yang (Shenzhen MSU-BIT University), Xinxiao Wu (Beijing Institute of Technology)
RecognitionObject DetectionTransformerPrompt EngineeringContrastive LearningVideoMultimodality
🎯 What it does: For the task of open vocabulary video visual relationship detection, a multimodal prompting method is proposed, combining spatio-temporal visual prompting with vision-guided language prompting, utilizing CLIP to achieve cross-modal knowledge transfer.
Multi-Modality Affinity Inference for Weakly Supervised 3D Semantic Segmentation
Xiawei Li (Beihang University), Dong Xu (Zhejiang University)
SegmentationTransformerMultimodalityPoint Cloud
🎯 What it does: This paper proposes a weakly supervised 3D semantic segmentation method enhanced by multimodal similarity (MMA), which alleviates the impact of long-tail distribution by combining geometric information with RGB information and normalizing classifier weights.
Multi-Objective Bayesian Optimization with Active Preference Learning
Ryota Ozaki (Nagoya Institute of Technology), Masayuki Karasuyama (Nagoya Institute of Technology)
OptimizationTabular
🎯 What it does: In multi-objective black-box optimization, a Bayesian optimization framework is proposed that actively estimates the decision maker's preferences through interactive preference learning (comparisons and improvement requests) and achieves optimal solution search by combining Chebyshev scalarization.
Multi-Prompts Learning with Cross-Modal Alignment for Attribute-Based Person Re-identification
Yajing Zhai (Hunan University), Da Cao (Hunan University)
RecognitionRetrievalTransformerVision Language ModelContrastive LearningImageMultimodality
🎯 What it does: A multi-prompt learning-based attribute-driven face re-identification framework MP-ReID is proposed.
Multi-Prototype Space Learning for Commonsense-Based Scene Graph Generation
Lianggangxu Chen (East China Normal University), Gaoqi He (East China Normal University)
Object DetectionGenerationContrastive LearningImageGraph
🎯 What it does: A multi-prototype learning framework (MPL) is proposed for commonsense-driven scene graph generation, addressing the issue of misclassification of diverse predicates caused by single prototypes.
Multi-Region Text-Driven Manipulation of Diffusion Imagery
Yiming Li (Shanghai Jiao Tong University), Yi Xu (China Mobile)
GenerationData SynthesisVision Language ModelDiffusion modelContrastive LearningImageText
🎯 What it does: This paper proposes a multi-region text-driven diffusion image editing framework (MRGD) that enables the addition, deletion, and attribute modification of multiple targets in an image through given region-level text prompts.
Multi-Scene Generalized Trajectory Global Graph Solver with Composite Nodes for Multiple Object Tracking
Yan Gao (Xidian University), Xinbo Gao (Chongqing University of Posts and Telecommunications)
Object TrackingGraph Neural NetworkTransformerVideo
🎯 What it does: A global multi-object tracking framework based on composite nodes, CoNo-Link, is proposed, which can efficiently construct sparse graphs and perform end-to-end association in long videos.
Multi-Source Collaborative Gradient Discrepancy Minimization for Federated Domain Generalization
Yikang Wei (Tianjin University), Yahong Han (Tianjin University)
Domain AdaptationFederated LearningConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a Multi-Source Collaborative Gradient Difference Minimization (MCGDM) framework for Federated Domain Generalization (FedDG) and Federated Domain Adaptation (FedDA), which learns domain-invariant models by performing intra-domain and cross-domain gradient matching at local clients.
Multi-Step Denoising Scheduled Sampling: Towards Alleviating Exposure Bias for Diffusion Models
Zhiyao Ren (University of Sydney), Dacheng Tao (University of Sydney)
RestorationGenerationDiffusion modelImage
🎯 What it does: A multi-step denoising scheduling sampling (MDSS) method is proposed and implemented to alleviate the exposure bias problem in diffusion models.
Multi-View Dynamic Reflection Prior for Video Glass Surface Detection
Fang Liu (City University of Hong Kong), Rynson W.H. Lau (City University of Hong Kong)
Object DetectionSegmentationTransformerVideo
🎯 What it does: The first video glass surface detection method, VGSD-Net, is proposed, which utilizes multi-view dynamic reflection information to locate the glass surface.
Multi-View People Detection in Large Scenes via Supervised View-Wise Contribution Weighting
Qi Zhang (Shenzhen University), Hui Huang (City University of Hong Kong)
Object DetectionDomain AdaptationConvolutional Neural NetworkSupervised Fine-TuningImageVideo
🎯 What it does: A multi-view pedestrian detection framework based on supervised perspective contribution weighting is proposed, achieving cross-scene generalization in large-scale scenarios.
Multi-View Randomized Kernel Classification via Nonconvex Optimization
Xiaojian Ding (Nanjing University of Finance and Economics), Fan Yang (Nanjing University of Finance and Economics)
ClassificationOptimizationTabular
🎯 What it does: A multi-kernel learning method based on randomized kernels, RMKL, is proposed, which selects a high-quality subset of kernels for classification by optimizing kernel diversity and generalization error.
MuLTI: Efficient Video-and-Language Understanding with Text-Guided MultiWay-Sampler and Multiple Choice Modeling
Jiaqi Xu (Alibaba Group), Xing Shi (Alibaba Group)
ClassificationRetrievalTransformerVision Language ModelVideoTextMultimodality
🎯 What it does: The MuLTI model is proposed to address the issues of long sequence processing in video-text understanding and the gap between pre-training and downstream tasks.
Multiagent Gumbel MuZero: Efficient Planning in Combinatorial Action Spaces
Xiaotian Hao (Tianjin University), Yan Zheng (Huawei)
OptimizationReinforcement LearningWorld ModelSequential
🎯 What it does: This paper proposes MA Gumbel MuZero and MA Gumbel AlphaZero, extending MuZero to multi-agent MMDP tasks with combinatorial action spaces, and improving policy update and sampling efficiency through non-replacement sampling and Gumbel-Top-k search.
Multichannel AV-wav2vec2: A Framework for Learning Multichannel Multi-Modal Speech Representation
Qiushi Zhu (University of Science and Technology of China), Lirong Dai (University of Science and Technology of China)
RecognitionRepresentation LearningTransformerContrastive LearningVideoMultimodalityAudio
🎯 What it does: Proposes the AV-wav2vec2 end-to-end multi-channel multi-modal self-supervised pre-training framework, utilizing video and multi-channel audio for contrastive learning;
Multilevel Attention Network with Semi-supervised Domain Adaptation for Drug-Target Prediction
Zhousan Xie (Shanghai Jiao Tong University), Lei Xu (Guangdong Institute of Intelligence Science and Technology)
Domain AdaptationDrug DiscoveryTransformerBiomedical Data
🎯 What it does: This paper proposes a semi-supervised domain adaptation multi-layer attention network MlanDTI for predicting drug-target interactions (DTI).
Multimodal Event Causality Reasoning with Scene Graph Enhanced Interaction Network
Jintao Liu (University of Chinese Academy of Sciences), Chenglong Liu (University of Chinese Academy of Sciences)
RecognitionObject DetectionGraph Neural NetworkImageTextMultimodality
🎯 What it does: This paper proposes a scene graph-based multimodal event causal reasoning framework called SEIN, which can simultaneously utilize graph convolutional networks to model the objects and their relationships within images, and achieve global alignment of objects in two sequential images through optimal transport, thereby capturing cross-image object interaction information. It then uses cross-modal attention to fuse textual and visual features, ultimately achieving event causal relationship prediction.