AAAI 2025 Papers — Page 20
AAAI Conference on Artificial Intelligence · 3028 papers
Multi-Scale Contrastive Learning for Video Temporal Grounding
Thong Thanh Nguyen (National University of Singapore), Anh Tuan Luu (Nanyang Technological University)
RecognitionRetrievalContrastive LearningVideo
🎯 What it does: A multi-scale contrastive learning framework is proposed, utilizing multi-layer features from a video encoder for video temporal localization;
Multi-Shape Matching with Cycle Consistency Basis via Functional Maps
Yifan Xia (Wuhan University), Jiayi Ma (Wuhan University)
OptimizationGraph Neural NetworkMeshBenchmark
🎯 What it does: A multi-shape matching method based on functional mapping is proposed, achieving a balance between matching accuracy and consistency through a two-stage optimization of graph structures.
Multi-StyleGS: Stylized Gaussian Splatting with Multiple Styles
Yangkai Lin (South China University of Technology), Kui Jia (Chinese University of Hong Kong)
Image TranslationSegmentationGenerationGaussian SplattingImage
🎯 What it does: This paper proposes Multi-StyleGS, which utilizes Gaussian Splatting combined with segmentation and bidirectional matching to achieve multi-style local style transfer in 3D scene rendering.
Multi-Subspace Matrix Recovery from Permuted Data
Liangqi Xie (Chinese University of Hong Kong), Jicong Fan (Chinese University of Hong Kong)
RestorationAnomaly DetectionAuto EncoderTabularElectronic Health Records
🎯 What it does: A recovery framework for multi-subspace matrices under partial or complete permutation contamination (PMSDR) is proposed, and a four-stage processing pipeline is implemented.
Multi-task Visual Grounding with Coarse-to-Fine Consistency Constraints
Ming Dai (Southeast University), Wankou Yang (Southeast University)
Object DetectionSegmentationTransformerVision Language ModelImageMultimodality
🎯 What it does: This paper proposes a multi-task visual alignment framework named C³VG, which jointly performs localization (REC) and segmentation (RIS), achieving multi-task consistency and refined predictions through a coarse perception stage and a refined consistency interaction stage.
Multi-Teacher Knowledge Distillation with Reinforcement Learning for Visual Recognition
Chuanguang Yang (Institute of Computing Technology, Chinese Academy of Sciences), Yongjun Xu (Institute of Computing Technology, Chinese Academy of Sciences)
ClassificationRecognitionObject DetectionSegmentationKnowledge DistillationTransformerReinforcement LearningImage
🎯 What it does: A multi-teacher knowledge distillation framework based on reinforcement learning, MTKD-RL, is proposed, which dynamically generates teacher weights through agents to achieve better balance and integration of multi-teacher information.
Multi-to-Single: Reducing Multimodal Dependency in Emotion Recognition Through Contrastive Learning
Yan-Kai Liu (Shanghai Jiao Tong University), Wei-Long Zheng (Shanghai Jiao Tong University)
RecognitionTransformerContrastive LearningMultimodalityTime Series
🎯 What it does: A multi-modal to single-modal emotion recognition framework (M2S) is proposed, which achieves high performance in cross-modal and single-modal tasks by pre-training on unlabeled multi-modal data and then fine-tuning on a single modality.
Multi-Turn Jailbreaking Large Language Models via Attention Shifting
Xiaohu Du (Huazhong University of Science and Technology), Jie Shi (Huawei International)
OptimizationAdversarial AttackTransformerLarge Language ModelText
🎯 What it does: This paper proposes a multi-round jailbreak method ASJA that creates multi-round dialogue history through attention transfer, utilizing genetic algorithms to iteratively optimize dialogue content, enticing LLMs to shift attention from harmful questions to historical responses, thereby breaking through safety alignment.
Multi-type MOOCs Recommendation: Leveraging Deep Multi-Relational Representation and Hierarchical Reasoning
Ye Zhang (Northeast Normal University), Minghao Yin (Northeast Normal University)
Recommendation SystemGraph Neural NetworkReinforcement LearningMultimodality
🎯 What it does: A multi-type MOOCs recommendation framework TOME is proposed, which integrates multi-relation representation and hierarchical reasoning to achieve linked recommendations for courses, knowledge points, and videos.
Multi-View 3D Human Pose Estimation with Weakly Synchronized Images
Ling Li (Tsinghua University), Xiao-Ping Zhang (Tsinghua University)
Pose EstimationDiffusion modelImageVideo
🎯 What it does: A multi-view 3D human pose estimation method based on diffusion models, SyncDiffPose, has been developed to directly output poses for weakly synchronized image scenes.
Multi-View Collaborative Learning Network for Speech Deepfake Detection
Kuiyuan Zhang (Harbin Institute of Technology), Guoai Xu
ClassificationAnomaly DetectionConvolutional Neural NetworkTransformerContrastive LearningAudio
🎯 What it does: A multi-view collaborative learning network is proposed to learn representations of deepfake speech from both raw waveforms and spectrograms for detection.
Multi-view Consistent 3D Panoptic Scene Understanding
Xianzhu Liu (Harbin Institute of Technology), Shengping Zhang (Harbin Institute of Technology)
Object DetectionSegmentationDepth EstimationNeural Radiance FieldPoint Cloud
🎯 What it does: This paper proposes the MVC-PSU method, achieving multi-view consistent 3D panoramic scene understanding.
Multi-View Empowered Structural Graph Wordification for Language Models
Zipeng Liu (Tianjin University), Nan Feng (Tianjin University)
ClassificationGraph Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningGraph
🎯 What it does: An end-to-end framework called Dr.E is proposed, which maps graph-structured data to the vocabulary of LLM through multi-view structural enhancement and code quantization, achieving token-level alignment between graphs and LLM, thus directly outputting natural language predictions in graph node classification tasks without textual information.
Multi-view Evidential Learning-based Medical Image Segmentation
Chao Huang (Sun Yat-sen University), Jie Wen (Harbin Institute of Technology)
SegmentationBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A medical image segmentation framework based on multi-view evidence learning is designed, combining traditional models and visual foundation models to extract domain-specific and general knowledge.
Multi-view Granular-ball Contrastive Clustering
Peng Su (Sichuan University), Jiancheng Lv (Sichuan University)
Auto EncoderContrastive LearningImage
🎯 What it does: This paper proposes a multi-view particle ball contrastive clustering method called MGBCC, which captures the local structure of samples using particle ball partitioning and performs contrastive learning on cross-view particle balls in the latent space to achieve unsupervised multi-view clustering.
Multi-View Incremental Learning with Structured Hebbian Plasticity for Enhanced Fusion Efficiency
Yuhong Chen (Fuzhou University), Mingkun Xu (Guangdong Institute of Intelligence Science and Technology)
Graph Neural NetworkGraph
🎯 What it does: A multi-view incremental learning framework (MVIL) based on biological neural plasticity mechanisms is proposed, which can gradually integrate new views while retaining old knowledge.
Multi-View Multi-Label Classification via View-Label Matching Selection
Hao Wei (Beijing University of Technology), Gengyu Lyu (Beijing University of Technology)
ClassificationGraph Neural NetworkMultimodality
🎯 What it does: A multi-view multi-label classification method VAMS based on view-label matching selection is proposed.
Multi-View Pedestrian Occupancy Prediction with a Novel Synthetic Dataset
Sithu Aung (Korea Institute of Science and Technology), Junghyun Cho (Korea Institute of Science and Technology)
Object DetectionSegmentationDomain AdaptationAutonomous DrivingConvolutional Neural NetworkImage
🎯 What it does: This study proposes a new multi-view pedestrian occupancy prediction task, constructs a large-scale synthetic dataset MVP-Occ, and designs a baseline model OmniOcc;
MultiBooth: Towards Generating All Your Concepts in an Image from Text
Chenyang Zhu (Tsinghua University), Xiu Li (Hong Kong University of Science and Technology)
GenerationData SynthesisTransformerDiffusion modelImageText
🎯 What it does: This paper proposes MultiBooth, a two-stage multi-concept customization framework that first learns single-concept modules for each concept and then stitches these modules together by location during generation using a Region Customization Module (RCM) to produce multi-concept images.
Multifaceted User Modeling in Recommendation: A Federated Foundation Models Approach
Chunxu Zhang (Jilin University), Bo Yang (Tsinghua University)
Recommendation SystemFederated LearningTransformerTabular
🎯 What it does: In the federated learning framework, the MRFF system is proposed, utilizing a lightweight Transformer as the base model and incorporating group gating networks at each layer to group users, achieving user-specific and group-level personalized joint learning, ultimately used for CTR prediction tasks.
Multilingual LLMs Inherently Reward In-Language Time-Sensitive Semantic Alignment for Low-Resource Languages
Ashutosh Bajpai (Indian Institute of Technology Delhi), Tanmoy Chakraborty (Indian Institute of Technology Delhi)
RetrievalTransformerLarge Language ModelContrastive LearningTextBenchmark
🎯 What it does: Improving temporal reasoning under low-resource languages, proposing a multilingual temporal reasoning dataset mTEMPREASON, and developing a cross-lingual time-sensitive semantic alignment retriever CLiTSSA;
Multilingual Mathematical Reasoning: Advancing Open-Source LLMs in Hindi and English
Avinash Anand (Indraprastha Institute of Information Technology), Rajiv Ratn Shah (Indraprastha Institute of Information Technology)
TransformerLarge Language ModelSupervised Fine-TuningTextChain-of-Thought
🎯 What it does: This paper develops a bilingual (English and Hindi) mathematical problem-solving system that enhances the mathematical reasoning ability of open-source LLMs by integrating curriculum learning, decomposition strategies, and structured solving frameworks.
Multimodal Class-aware Semantic Enhancement Network for Audio-Visual Video Parsing
Pengcheng Zhao (Hefei University of Technology), Yanxiang Chen (Hefei University of Technology)
ClassificationRecognitionSegmentationVideoMultimodalityAudio
🎯 What it does: This study addresses the audio-video video parsing (AVVP) task and proposes a multi-modal class-aware semantic enhancement network (MM-CSE). By decoupling mixed features into event-specific and background class features, and achieving event co-occurrence modeling and global semantic fusion at a fine-grained level, it significantly improves event localization performance.
Multimodal Fine-Grained Apparent Personality Trait Recognition: Joint Modeling of Big Five and Questionnaire Item-level Scores
Ryo Masumura (NTT Corporation), Nobukatsu Hojo (NTT Corporation)
RecognitionTransformerVideoTextMultimodalityAudio
🎯 What it does: This paper proposes a multimodal fine-grained level explicit personality trait recognition method, jointly estimating Big Five scores and questionnaire item-level scores, implemented through a multimodal Transformer architecture.
Multimodal Fusion Using Multi-View Domains for Data Heterogeneity in Federated Learning
Min Gao (Fuzhou University), Ran Tao (Beijing Institute of Technology)
Federated LearningSafty and PrivacyMultimodality
🎯 What it does: Proposes the FedMVD framework, which combines Global Alignment (GLA) and Local Angle Margin (LAM) in multi-view domains to address data heterogeneity and long-tail distribution issues in multimodal federated learning.
Multimodal Hypothetical Summary for Retrieval-based Multi-image Question Answering
Peize Li (Jilin University), Yan Wang (Jilin University)
RetrievalTransformerLarge Language ModelContrastive LearningImageTextMultimodalityRetrieval-Augmented Generation
🎯 What it does: This paper proposes a retrieval-based multi-image question answering framework that closely integrates information retrieval with the answering process, achieving text-to-text retrieval by generating multimodal hypothesis summaries (MHyS) to replace real images, thereby reducing two-stage errors.
Multimodal Promptable Token Merging for Diffusion Models
Cheng-Yao Hong (Academia Sinica), Tyng-Luh Liu (Academia Sinica)
Object DetectionGenerationCompressionTransformerPrompt EngineeringDiffusion modelImageMultimodalityAudio
🎯 What it does: A multi-modal promptable token merging (MPTM) method is proposed to compress tokens based on multi-modal prompt information in diffusion models, reducing attention computation costs.
Multimodal Variational Autoencoder: A Barycentric View
Peijie Qiu (Washington University in St. Louis), Aristeidis Sotiras (Clemson University)
GenerationData SynthesisMixture of ExpertsAuto EncoderMultimodality
🎯 What it does: This study investigates the aggregation methods of multimodal variational autoencoders, proposing a general framework based on a Bayesian perspective, and aggregates unimodal inference distributions through Wasserstein barycenter, presenting two models: WB-VAE and MWB-VAE.
Multiple Feature Refining Network for Visual Emotion Distribution Learning
Qinfu Xu (China University of Petroleum), Chunlei Wu (China University of Petroleum)
Convolutional Neural NetworkTransformerImage
🎯 What it does: This paper proposes a Multi-Feature Refinement Network (MFRN) that achieves visual emotion distribution learning through spectral mixing and semantic map prompt learning.
Multiple Mean-Payoff Optimization Under Local Stability Constraints
David Klaška (Masaryk University), Vojtěch Řehák (Masaryk University)
OptimizationReinforcement LearningGraph
🎯 What it does: Designed and implemented a scalable policy synthesis algorithm for multi-window mean reward optimization, based on differentiable dynamic programming and gradient descent, supporting finite memory stochastic policies.
Multiple Purchase Chains with Negative Transfer Elimination for Multi-Behavior Recommendation
Shuwei Gong (Northeastern University), Xingwei Wang (Northeastern University)
Recommendation SystemGraph Neural NetworkTabular
🎯 What it does: This study focuses on multi-behavior recommendation and proposes the MPC model, which enhances recommendation effectiveness by utilizing multiple purchase chains and a negative transfer elimination mechanism.
Multiple Trade-offs: An Improved Approach for Lexicographic Linear Bandits
Bo Xue (City University of Hong Kong), Qingfu Zhang (City University of Hong Kong)
OptimizationReinforcement LearningTabular
🎯 What it does: This paper studies dictionary order online learning under the multi-objective linear Bandit (MOSLB) framework and proposes two algorithms, STLO and MTLO.
Multiplex Graph Representation Learning with Homophily and Consistency
Yudi Huang (Guangxi Normal University), Xiaofeng Zhu (Guangxi Normal University)
Representation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: This paper proposes an unsupervised multi-graph representation learning framework called MGHC, which can simultaneously enhance homophily and reduce the interference of heterophily on structural information in multi-graphs, achieving consistency at both node and category levels through a self-expression matrix.
MultiSFL: Towards Accurate Split Federated Learning via Multi-Model Aggregation and Knowledge Replay
Zeke Xia (East China Normal University), Mingsong Chen (East China Normal University)
Federated LearningConvolutional Neural NetworkImage
🎯 What it does: A Split Federated Learning framework named MultiSFL is proposed, which combines multi-model aggregation and knowledge replay to address the issues of data heterogeneity and catastrophic forgetting.
Multispectral Pedestrian Detection with Sparsely Annotated Label
Chan Lee (Kyung Hee University), Jung Uk Kim (Kyung Hee University)
RecognitionObject DetectionConvolutional Neural NetworkImageMultimodality
🎯 What it does: A new framework for sparse annotated multispectral pedestrian detection, SAMPD, is proposed to address the issues of pseudo-label quality and multimodal learning.
MuMA-ToM: Multi-modal Multi-Agent Theory of Mind
Haojun Shi (Johns Hopkins University), Tianmin Shu (Johns Hopkins University)
Large Language ModelVision Language ModelVideoTextMultimodalityBenchmark
🎯 What it does: The MuMA-ToM benchmark is proposed to evaluate Theory of Mind (ToM) reasoning in multi-modal multi-agent settings; based on this benchmark, a language model-based inverse multi-agent planning method called LIMP is designed and implemented; its effectiveness is validated through human experiments and comparisons with various large models.
MUN: Image Forgery Localization Based on M³ Encoder and UN Decoder
Yaqi Liu (Beijing Electronic Science and Technology Institute), Qiang Cai (Beijing Technology and Business University)
SegmentationAnomaly DetectionConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: This paper proposes an image forgery localization network called MUN, based on the M3 encoder and UN decoder, for pixel-level localization of various types of forgeries (splicing, copy-move, deletion, AI-generated) images.
MUSE: Mamba Is Efficient Multi-scale Learner for Text-video Retrieval
Haoran Tang (Peking University), Xiaodan Liang (Sun Yat-sen University)
RetrievalTransformerContrastive LearningVideoText
🎯 What it does: A feature pyramid is constructed based on CLIP to obtain multi-scale video features, and the Mamba linear state space model is used to efficiently model across resolutions, thereby improving text-video retrieval performance.
Muses: 3D-Controllable Image Generation via Multi-Modal Agent Collaboration
Yanbo Ding (Shenzhen Institute of Advanced Technology), Yali Wang (Shenzhen Institute of Advanced Technology)
GenerationData SynthesisLarge Language ModelSupervised Fine-TuningAgentic AIImageTextMultimodality
🎯 What it does: MUSES is proposed, a multimodal agent collaboration system that implements a complete process from user text to controllable 3D images.
MV-VTON: Multi-View Virtual Try-On with Diffusion Models
Haoyu Wang (Peking University), Wangmeng Zuo (Harbin Institute of Technology)
GenerationData SynthesisDiffusion modelImage
🎯 What it does: Achieved multi-view virtual try-on (MV-VTON)
MVREC: A General Few-shot Defect Classification Model Using Multi-View Region-Context
Shuai Lyu (Hong Kong Polytechnic University), Waikeung Wong (Hong Kong Polytechnic University)
ClassificationAnomaly DetectionVision Language ModelContrastive LearningImageBenchmark
🎯 What it does: A general few-shot defect classification framework based on multi-view region-context, MVREC, is proposed to address the issues of context information dependency and insufficient data generalization in defect classification.
MVReward: Better Aligning and Evaluating Multi-View Diffusion Models with Human Preferences
Weitao Wang (Tsinghua University), Haoqian Wang (Tsinghua University)
GenerationReinforcement Learning from Human FeedbackReinforcement LearningDiffusion modelImage
🎯 What it does: This paper proposes the MVReward framework for better aligning and evaluating multi-view diffusion models with human preferences.
MYOPIA: Protecting Face Privacy from Malicious Personalized Text-to-Image Synthesis via Unlearnable Examples
Zhihao Wu (Zhejiang University), Wenyuan Xu (Zhejiang University)
GenerationSafty and PrivacyDiffusion modelImage
🎯 What it does: A method called MYOPIA was developed, which is a no-learning sample approach that prevents personalized text-to-image models from learning real facial features by adding imperceptible perturbations to facial images, thereby protecting facial privacy.
nach0-pc: Multi-task Language Model with Molecular Point Cloud Encoder
Maksim Kuznetsov (Insilico Medicine Canada Inc), Zulfat Miftahutdinov (Insilico Medicine Canada Inc)
GenerationDrug DiscoveryTransformerLarge Language ModelTextPoint Cloud
🎯 What it does: A multi-task language model named nach0-pc has been developed, capable of receiving molecular point cloud and text input, and outputting three-dimensional molecular structures that include SMILES and XYZ coordinates.
NaviFormer: A Spatio-Temporal Context-Aware Transformer for Object Navigation
Wei Xie (Nanjing University of Science and Technology), Jin Xie (Nanjing University)
Robotic IntelligenceTransformerReinforcement LearningPoint Cloud
🎯 What it does: This paper proposes NaviFormer, an encoder-decoder Transformer designed to aggregate spatial layout, temporal pose, and traversable frontier information for object navigation.
Navigating Label Ambiguity for Facial Expression Recognition in the Wild
JunGyu Lee (Korea Institute of Science and Technology), Gi Pyo Nam (Korea Institute of Science and Technology)
ClassificationRecognitionConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a facial expression recognition framework called NLA that simultaneously addresses the issues of label ambiguity and class imbalance.
Navigating Towards Fairness with Data Selection
Yixuan Zhang (Southeast University), Feng Zhou (Renmin University of China)
ClassificationData-Centric LearningTransformerContrastive LearningImage
🎯 What it does: To address the decline in fairness caused by label bias, a data selection method based on zero-shot predictors and a peer prediction mechanism is proposed.
NBA3D: Neighbor-Based Confidence Adjustment for 3D Rare Object Detection Using LiDAR
Jooyoung Lee (Chung Ang University), Jongwon Choi (Chung Ang University)
Object DetectionAutonomous DrivingGraph Neural NetworkContrastive LearningPoint Cloud
🎯 What it does: The NBA3D method is proposed, which constructs a neighborhood graph using LiDAR detection results and reallocates confidence through a graph neural network, specifically enhancing the detection performance of rare categories (such as strollers).
Nearly Tight Bounds for Exploration in Streaming Multi-Armed Bandits with Known Optimality Gap
Nikolai Karpov (Indiana University), Chen Wang (Texas A&M University)
OptimizationReinforcement Learning from Human FeedbackTabular
🎯 What it does: Under the condition of the known optimal gap Δ[2], this study conducts pure exploration for multi-channel streaming multi-armed bandits (MAB) and provides approximate optimal trade-off bounds between samples, memory, and the number of channels.
Nearly Tight Bounds on Approximate Equilibria in Spatial Competition on the Line
Umang Bhaskar (Tata Institute of Fundamental Research), Soumyajit Pyne (Tata Institute of Fundamental Research)
Optimization
🎯 What it does: This paper studies the approximate Nash equilibrium of candidate positions in the Hotelling spatial competition model, providing upper and lower bounds for ϵ-equilibrium, and proving that the error decreases with the number of candidates in the case of multiple candidates.
Neighbor Does Matter: Density-Aware Contrastive Learning for Medical Semi-supervised Segmentation
Feilong Tang (Monash University), Zongyuan Ge (Monash University)
SegmentationContrastive LearningBiomedical DataComputed Tomography
🎯 What it does: This paper proposes a semi-supervised multi-organ segmentation method based on density-aware contrastive learning.
Neighborhood-Aware Negative Sampling for Student Knowledge and Behavior Modeling
Siqian Zhao (University at Albany), Sherry Sahebi (University at Albany)
Recommendation SystemRecurrent Neural NetworkGraph Neural NetworkTabular
🎯 What it does: A neighborhood-based negative sampling method (NANS) and a multi-objective multi-task student knowledge and behavior modeling framework (KoBeM) are proposed to simultaneously predict students' next topic and answer scores.
NEST: A Neuromodulated Small-world Hypergraph Trajectory Prediction Model for Autonomous Driving
Chengyue Wang (University of Macau), Zhenning Li (University of Macau)
Autonomous DrivingGraph Neural NetworkTransformerMultimodality
🎯 What it does: Proposed and implemented the NEST model, which realizes a trajectory prediction framework based on small-world networks and hypergraphs.
NeSyCoCo: A Neuro-Symbolic Concept Composer for Compositional Generalization
Danial Kamali (Michigan State University), Parisa Kordjamshidi (Michigan State University)
TransformerLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: A neural symbolic framework called NeSyCoCo is proposed, which generates symbolic programs using large language models and achieves combinatorial generalization for image-language reasoning tasks through soft combination and distributed word embeddings.
Neural Assembler: Learning to Generate Fine-Grained Robotic Assembly Instructions from Multi-View Images
Hongyu Yan (Peking University), Yadong Mu (Peking University)
GenerationPose EstimationRobotic IntelligenceGraph Neural NetworkTransformerImagePoint Cloud
🎯 What it does: An end-to-end Neural Assembler is proposed, which automatically generates fine-grained assembly instructions for robots from multi-view images;
Neural Block Compression: Variable Bitrates Feature Blocks for Texture Representation
Rui Shi (Shanghai Jiao Tong University), Bingbing Ni (Shanghai Jiao Tong University)
CompressionImage
🎯 What it does: This paper proposes a Neural Block Compression (NBC) method that achieves low bitrate texture compression through multi-resolution neural feature blocks and variable bitrate quantization.
Neural Collapse Inspired Knowledge Distillation
Shuoxi Zhang (Huazhong University of Science and Technology), Kun He (Huazhong University of Science and Technology)
Object DetectionKnowledge DistillationConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: This paper introduces the Neural Collapse (NC) structure into the knowledge distillation framework and designs the NCKD method to encourage the student network to learn the teacher's NC structure.
Neural Combinatorial Clustered Bandits for Recommendation Systems
Baran Atalar (Carnegie Mellon University), Carlee Joe-Wong (Carnegie Mellon University)
Recommendation SystemReinforcement LearningTabular
🎯 What it does: A new neural combinatorial clustering gambling algorithm, NeUClust, is proposed for the contextual combinatorial gambling problem in recommendation systems, aiming to simultaneously learn the unknown reward function and select the arm with the highest reward.
Neural Combinatorial Optimization for Stochastic Flexible Job Shop Scheduling Problems
Igor G. Smit (Eindhoven University of Technology), Wim P.M. Nuijten (Eindhoven University of Technology)
OptimizationGraph Neural NetworkTransformerReinforcement LearningTabular
🎯 What it does: This paper proposes an attention-based Scene Processing Module (SPM) that integrates random processing time information into the Neural Combinatorial Optimization (NCO) framework to address the Flexible Job Shop Scheduling Problem (FJSP);
Neural Conformal Control for Time Series Forecasting
Ruipu Li (University of Michigan), Alexander Rodríguez (University of Michigan)
Recurrent Neural NetworkGraph Neural NetworkTime Series
🎯 What it does: An adaptive confidence interval prediction method based on neural networks, NCC, is proposed for time series forecasting.
Neural Conjugate Flows: A Physics-Informed Architecture with Flow Structure
Arthur Bizzi (Instituto de Matemática Pura e Aplicada), João M. Pereira (Instituto de Matemática Pura e Aplicada)
Flow-based ModelTime SeriesBiomedical DataPhysics RelatedOrdinary Differential Equation
🎯 What it does: This paper designs Neural Conjugate Flows (NCF), which constructs a topologically conjugate flow structure through reversible Coupling Layers and affine flows, naturally satisfying physical constraints such as initial conditions and causality, and capable of approximating any ODE flow.
Neural Control and Certificate Repair via Runtime Monitoring
Emily Yu (Institute of Science and Technology Austria), Thomas A. Henzinger (Institute of Science and Technology Austria)
OptimizationRobotic IntelligenceReinforcement LearningTime SeriesBenchmark
🎯 What it does: A framework for detecting and repairing black-box neural network control strategies and certificates (barrier/Lyapunov) based on runtime monitoring is designed, utilizing monitoring alerts to collect violation states and retrain to enhance safety.
Neural Networks Perform Sufficient Dimension Reduction
Shuntuo Xu (East China Normal University), Zhou Yu (East China Normal University)
TabularTime Series
🎯 What it does: This study investigates the sufficient dimension reduction (SDR) capability of deep feedforward neural networks in regression tasks, proving that the weights of the first layer can approximate the central mean subspace, and provides statistical consistency and experimental validation.
Neural Reasoning for Sure Through Constructing Explainable Models
Tiansi Dong (University of Cambridge), Pietro Liò (University of Cambridge)
OptimizationExplainability and InterpretabilityRecurrent Neural NetworkLarge Language ModelPrompt Engineering
🎯 What it does: This paper proposes and implements the Hyperbolic Sphere Neural Network (HSphNN), a neural network that explicitly maps set-theoretic relationships into Poincaré ball space. It can determine the validity and satisfiability of Aristotelian syllogisms within a constant number of epochs by constructing spherical configurations, and it can interact with large language models (ChatGPT) to check and correct their reasoning results.
Neural Reasoning Networks: Efficient Interpretable Neural Networks with Automatic Textual Explanations
Stephen Carrow (IBM), Alexander G. Gray (Centaur AI Institute)
ClassificationExplainability and InterpretabilityComputational EfficiencyTabular
🎯 What it does: A Neural Reasoning Network (NRN) is proposed for table classification and can automatically generate interpretable textual explanations.
Neural Temporal Point Processes for Forecasting Directional Relations in Evolving Hypergraphs
Tony Gracious (Indian Institute of Science), Ambedkar Dukkipati (Indian Institute of Science)
Recurrent Neural NetworkGraph Neural NetworkGraphTime Series
🎯 What it does: This study investigates how to predict events in time-evolving hypergraphs with directionality and higher-order relationships.
Neural Variable-Order Fractional Differential Equation Networks
Wenjun Cui (Beijing Jiaotong University), Yidong Li (Beijing Jiaotong University)
Graph Neural NetworkImageGraphOrdinary Differential Equation
🎯 What it does: A Neural Variable-Order Fractional Differential Equation network (NvoFDE) is proposed, which embeds learnable variable-order fractional differential operators into neural networks for solving graph neural networks and variable-order fractional differential equations.
Neural-Symbolic Collaborative Distillation: Advancing Small Language Models for Complex Reasoning Tasks
Huanxuan Liao (Chinese Academy of Sciences), Jun Zhao (Chinese Academy of Sciences)
Explainability and InterpretabilityKnowledge DistillationTransformerLarge Language ModelTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Proposes Neural-Symbolic Collaborative Distillation (NesyCD), which analyzes the errors of small models (SLM) through large language models (LLM), generates and stores a symbolic knowledge base, and dynamically retrieves this knowledge during inference to enhance the performance of SLM on complex reasoning tasks.
NeuralFlix: A Simple While Effective Framework for Semantic Decoding of Videos from Non-invasive Brain Recordings
Jingyuan Sun (University of Manchester), Marie-Francine Moens (KU Leuven)
GenerationData SynthesisTransformerDiffusion modelContrastive LearningVideoMultimodalityMagnetic Resonance Imaging
🎯 What it does: A two-stage framework called NeuralFlix has been constructed to decode fMRI data and generate high-quality videos that are semantically consistent with the videos being watched.
New Compilation Languages Based on Restricted Weak Decomposability
Petr Illner (Charles University)
Graph
🎯 What it does: This paper proposes the positive-negative weakly separable DNNF variants pwDNNF and nwDNNF, modifies the Bella compiler to generate these circuits, and designs an acceleration method based on sub-circuit replication, demonstrating that nwDNNF performs excellently in MPE computation for two-layer large domain Bayesian networks.
NightHaze: Nighttime Image Dehazing via Self-Prior Learning
Beibei Lin (National University of Singapore), Robby T. Tan (National University of Singapore)
RestorationTransformerImage
🎯 What it does: The NightHaze method is proposed, which enhances the dehazing effect of nighttime images through self-prior learning and self-refinement.
NightReID: A Large-Scale Nighttime Person Re-Identification Benchmark
Yuxuan Zhao (Wuhan University), Mang Ye (Wuhan University)
RecognitionRetrievalTransformerImageBenchmark
🎯 What it does: A large-scale nighttime RGB pedestrian re-identification dataset, NightReID, is proposed, along with an EDA framework based on unsupervised image enhancement and distribution alignment, significantly improving nighttime Re-ID performance.
NLGT: Neighborhood-based and Label-enhanced Graph Transformer Framework for Node Classification
Xiaolong Xu (Nanjing University of Information Science and Technology), Wanchun Dou (Macquarie University)
ClassificationGraph Neural NetworkTransformerGraph
🎯 What it does: Proposes the NLGT framework for node classification, utilizing neighborhood sampling, label-enhanced feature fusion, and neighborhood mask attention.
NLSR: Neuron-Level Safety Realignment of Large Language Models Against Harmful Fine-Tuning
Xin Yi (East China Normal University), Liang He (East China Normal University)
Safty and PrivacyComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes the Neuron-Level Safety Realignment (NLSR) framework for untrained safety recovery in the context of safety degradation issues that arise in large language models (LLMs) during fine-tuning-as-a-service scenarios.
No More Tuning: Prioritized Multi-Task Learning with Lagrangian Differential Multiplier Methods
Zhengxing Cheng (University of California), Qingwen Liu (Alibaba Group)
Recommendation SystemOptimizationMixture of ExpertsVideo
🎯 What it does: A multi-task learning framework NMT is proposed, which does not require manual hyperparameter tuning. It utilizes the Lagrange multiplier method to implement task priority constraints, ensuring that high-priority tasks are not interfered with by low-priority tasks.
Noise-Injected Spiking Graph Convolution for Energy-Efficient 3D Point Cloud Denoising
Zikuan Li (Nanjing University of Aeronautics and Astronautics), Jun Wang (Nanjing University of Aeronautics and Astronautics)
RestorationComputational EfficiencyGraph Neural NetworkSpiking Neural NetworkPoint Cloud
🎯 What it does: A noise-injected spiking graph convolutional network is proposed for 3D point cloud denoising.
Noise-Resilient Symbolic Regression with Dynamic Gating Reinforcement Learning
Chenglu Sun (Tencent), Zixia Zhou (Stanford University)
OptimizationReinforcement Learning
🎯 What it does: This paper proposes a Noise-Robust Symbolic Regression (NRSR) method that utilizes reinforcement learning and a Noise Suppression Gating Module (NGM) to recover expressions from high-noise data.
NoiseHGNN: Synthesized Similarity Graph-Based Neural Network for Noised Heterogeneous Graph Representation Learning
Zhang Xiong, Beibei Yu (University of Technology Sydney)
Representation LearningGraph Neural NetworkContrastive LearningGraphBiomedical Data
🎯 What it does: A model named NoiseHGNN is proposed for node representation learning in heterogeneous graphs with noisy edges, and it mitigates the impact of erroneous edges through synthetic similarity graphs and contrastive learning.
Noisy Correspondence Rectification via Asymmetric Similarity Learning
Yunbo Wang (Central South University), Jianhai Chen (Zhejiang University)
RetrievalRecurrent Neural NetworkContrastive LearningImageTextMultimodality
🎯 What it does: This paper studies a method for image-text matching that addresses noise in multimodal data.
Noisy Label Calibration for Multi-View Classification
Shilin Xu (Sichuan University), Dezhong Peng (Sichuan National Innovation New Vision UHD Video Technology Co., Ltd.)
ClassificationTabular
🎯 What it does: A noise label calibration method NLC is proposed in multi-view learning, integrating cross-view maximum margin ranking, MixUp, and noise detection and calibration modules to achieve robust classification.
Noisy Node Classification by Bi-level Optimization Based Multi-Teacher Distillation
Yujing Liu (Guangxi Normal University), Xiaofeng Zhu (Guangxi Normal University)
ClassificationOptimizationKnowledge DistillationGraph Neural NetworkGraph
🎯 What it does: This paper proposes a multi-teacher distillation method based on dual-layer optimization, called BO-NNC, to handle noisy node labels in graph data.
NOMATTERXAI: Generating “No Matter What” Alterfactual Examples for Explaining Black-Box Text Classification Models
Tuc Van Nguyen (Indiana University), Thai Le (Indiana University)
ClassificationExplainability and InterpretabilityTransformerLarge Language ModelText
🎯 What it does: The NOMATTERXAI algorithm is proposed, which can automatically generate 'alterfactual' examples for text classifiers to assess the model's sensitivity and fairness regarding irrelevant features.
Non-Convex Tensor Recovery from Local Measurements
Tongle Wu (Pennsylvania State University), Jicong Fan (Chinese University of Hong Kong)
CompressionOptimizationVideo
🎯 What it does: A low tubal-rank tensor compressed sensing model based on local slice perception is proposed, and two non-convex iterative algorithms, Alt-PGD-Min and Alt-ScalePGD-Min, are provided to achieve global convergence.
Non-stochastic Budgeted Online Pricing with Semi-Bandit Feedback
Xiang Liu (Southeast University), Long Tran-Thanh (University of Warwick)
OptimizationReinforcement LearningTabular
🎯 What it does: This paper proposes and studies a non-random, budget-constrained procurement pricing scenario, utilizing a semi-bandit feedback framework to dynamically quote sellers and maximize the buyer's cumulative profit.
Normal-NeRF: Ambiguity-Robust Normal Estimation for Highly Reflective Scenes
Ji Shi (Peking University), Wenzhen Yue (Peking University)
GenerationData SynthesisNeural Radiance FieldImage
🎯 What it does: For high-reflective scenes, we propose Normal-NeRF, which achieves robust estimation of surface normals and renders high-quality images through the transmission gradient estimation method and a dual activation density module.
Normalize Then Propagate: Efficient Homophilous Regularization for Few-Shot Semi-Supervised Node Classification
Baoming Zhang (Nanjing University), Chongjun Wang (Nanjing University)
ClassificationGraph Neural NetworkGraph
🎯 What it does: A graph neural network named NormProp is proposed, which first normalizes node features to the unit sphere, then propagates through low-pass filtering. It encodes category information using the direction of node vectors and aggregates consistency using the Euclidean norm, while mining supervisory signals from unlabeled nodes through homogeneity regularization to enhance generalization performance in scenarios with few labels.
Novel View Synthesis Under Large-Deviation Viewpoint for Autonomous Driving
Xin Ma (Beijing University of Posts and Telecommunications), Chengwei Pan (Beihang University)
Data SynthesisAutonomous DrivingOptimizationImage
🎯 What it does: This paper addresses the challenge of view synthesis under large perspective shifts in autonomous driving scenarios by proposing an attention-based lighting model using neighboring views and a geometric optimization method based on planar homography to improve the rendering quality of 3D Gaussian Splatting (3D-GS).
Novelty-Guided Data Reuse for Efficient and Diversified Multi-Agent Reinforcement Learning
Yangkun Chen (Shenzhen International Graduate School Tsinghua University), Jiafei Lyu (Shenzhen International Graduate School Tsinghua University)
Reinforcement LearningTabular
🎯 What it does: To address the issues of low sample utilization and insufficient agent diversity in multi-agent reinforcement learning, a sample reuse method based on observation novelty called MANGER is proposed. It dynamically adjusts the policy update frequency for each agent and introduces shared and independent layers in the critic network to enhance agent diversity.
Nuance Matters: Probing Epistemic Consistency in Causal Reasoning
Shaobo Cui (École Polytechnique Fédérale de Lausanne), Boi Faltings (École Polytechnique Fédérale de Lausanne)
TransformerLarge Language ModelText
🎯 What it does: This study investigates the self-consistency of fine-grained mediating variables in causal reasoning by large language models (causal cognitive consistency) and proposes evaluation metrics.
Number Theoretic Accelerated Learning of Physics-Informed Neural Networks
Takashi Matsubara (Hokkaido University), Takaharu Yaguchi (Kobe University)
OptimizationComputational EfficiencyPhysics Related
🎯 What it does: A new training method called Good Lattice Training (GLT) is proposed to accelerate the learning process of Physics-Informed Neural Networks (PINNs), particularly in solving partial differential equations (PDEs).
NumbOD: A Spatial-Frequency Fusion Attack Against Object Detectors
Ziqi Zhou (Huazhong University of Science and Technology), Hai Jin (Huazhong University of Science and Technology)
Object DetectionAdversarial AttackImage
🎯 What it does: A model-agnostic spatial-frequency fusion adversarial attack, NumbOD, is designed to achieve imperceptible and global failure against object detectors.
Numerical Pruning for Efficient Autoregressive Models
Xuan Shen (Northeastern University), Jiuxiang Gu (Adobe Research)
GenerationCompressionOptimizationComputational EfficiencyTransformerLarge Language ModelImageText
🎯 What it does: A training-independent structured pruning method is proposed to compress autoregressive Transformer models through numerical scoring and compensation algorithms.
OAC: Output-adaptive Calibration for Accurate Post-training Quantization
Ali Edalati (Huawei Noah's Ark Lab), Vahid Partovi Nia (Huawei Noah's Ark Lab)
CompressionOptimizationTransformerLarge Language ModelText
🎯 What it does: A post-training quantization method based on model output cross-entropy error, called OAC, is proposed for compressing large language models.
OAMaskFlow: Occlusion-Aware Motion Mask for Scene Flow
Xiongfeng Peng (Samsung Research and Development Institute China), Qiang Wang (Samsung Research and Development Institute China)
Object DetectionSegmentationDepth EstimationAutonomous DrivingSimultaneous Localization and MappingOptical FlowImagePoint Cloud
🎯 What it does: A scene flow estimation framework called OAMaskFlow is proposed, which uses an occlusion-aware motion mask to supervise motion embedding and extends reliable 3D motion to occluded areas through an attention propagation module.
Object-level Geometric Structure Preserving for Natural Image Stitching
Wenxiao Cai (Southeast University), Wankou Yang (Southeast University)
Image TranslationSegmentationSimultaneous Localization and MappingImageBenchmark
🎯 What it does: A natural image stitching method based on object-level geometric structure preservation
ObjVariantEnsemble: Advancing Point Cloud LLM Evaluation in Challenging Scenes with Subtly Distinguished Objects
Qihang Cao (Shanghai Jiao Tong University), Huangxun Chen (Hong Kong University of Science and Technology)
Object DetectionRetrievalTransformerLarge Language ModelVision Language ModelPoint CloudBenchmark
🎯 What it does: This work constructs a customizable 3D point cloud evaluation benchmark, ObjVariantEnsemble, which combines object-level resources and scene-level scans to automatically generate 75k finely annotated scenes through LLM-VLM collaboration.
Occlusion-Embedded Hybrid Transformer for Light Field Super-Resolution
Zeyu Xiao (National University of Singapore), Wei Jia (Hefei University of Technology)
RestorationSuper ResolutionTransformerImage
🎯 What it does: A network called Occlusion-Embedded Hybrid Transformer (OHT) is proposed to achieve light field super-resolution;
Occlusion-Insensitive Talking Head Video Generation via Facelet Compensation
Yuhui Deng (South China University of Technology), Shengfeng He (Singapore Management University)
GenerationData SynthesisPose EstimationTransformerGenerative Adversarial NetworkOptical FlowVideo
🎯 What it does: A facial compensation mechanism based on facial semantic tokens is proposed, achieving occlusion-insensitive speaker video generation.
ODDN: Addressing Unpaired Data Challenges in Open-World Deepfake Detection on Online Social Networks
Renshuai Tao (Beijing Jiaotong University), Yao Zhao (Beijing Jiaotong University)
ClassificationAnomaly DetectionConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: A deepfake detection framework suitable for open-world social networks, ODDN, is proposed to address the issue of missing paired data between compressed images and original images.
Offline Multi-Agent Reinforcement Learning via In-Sample Sequential Policy Optimization
Zongkai Liu (Sun Yat-sen University), Xuetao Ding (Meituan)
Reinforcement LearningTabularSequential
🎯 What it does: This paper proposes an offline multi-agent reinforcement learning algorithm named InSPO, which utilizes sequential in-sample policy optimization to learn a joint policy for multiple agents, avoiding the out-of-distribution (OOD) joint action and local optimal convergence issues in offline scenarios.