IJCAI 2024 Papers — Page 5
International Joint Conference on Artificial Intelligence · 790 papers
Joint Multimodal Aspect Sentiment Analysis with Aspect Enhancement and Syntactic Adaptive Learning
Linlin Zhu (Xi'an Jiaotong University), Liang He (Xi'an Jiaotong University)
ClassificationRecognitionRepresentation LearningGraph Neural NetworkTransformerLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: This paper proposes the AESAL model, which utilizes aspect-oriented enhanced pre-training and syntax-distance based adaptive learning to achieve three tasks: joint multi-modal aspect sentiment analysis (JMASA), aspect extraction (MATE), and sentiment classification (MASC).
Joint Source Localization in Different Platforms via Implicit Propagation Characteristics of Similar Topics
Zhen Wang (Northwestern Polytechnical University), Xianghua Li (Northwestern Polytechnical University)
Recurrent Neural NetworkGraph Neural NetworkAuto EncoderText
🎯 What it does: This paper constructs a cross-platform cascade dataset for similar topics spreading on Twitter and Weibo, and proposes a dual-channel source localization framework (DSLF), achieving precise localization of cross-platform source nodes through technologies such as self-loop attention GCN, dual structure KL regularization, and exponential VAE.
Justifying Argument Acceptance with Collective Attacks: Discussions and Disputes
Giovanni Buraglio (Technische Universitaet Wien), Markus Ulbricht (Leipzig University)
🎯 What it does: This paper proposes the instantiation of knowledge bases under the context of collective attacks using SETAF, and defines corresponding discussion games and argument trees to generate more concise argument acceptance reasons.
KDDC: Knowledge-Driven Disentangled Causal Metric Learning for Pre-Travel Out-of-Town Recommendation
Yinghui Liu (Zhejiang University of Technology), Xiangjie Kong (Zhejiang University of Technology)
Recommendation SystemGraph Neural NetworkGraphTabular
🎯 What it does: Proposed the KDDC framework, which combines knowledge graph semantic representation, causal separation, and metric learning to model users' interests and herd behavior, achieving intent inference and recommendation for pre-travel POI (Points of Interest).
Kernel Readout for Graph Neural Networks
Jiajun Yu (China Agricultural University), Jicong Fan (Chinese University of Hong Kong)
ClassificationDrug DiscoveryGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposed a kernel-based graph pooling method called KerRead, which maps node embeddings to an infinite-dimensional enhanced space via kernel computations after using learnable adaptive centers.
KG-CoT: Chain-of-Thought Prompting of Large Language Models over Knowledge Graphs for Knowledge-Aware Question Answering
Ruilin Zhao (Huazhong University of Science and Technology), Guandong Xu (University of Technology Sydney)
Explainability and InterpretabilityComputational EfficiencyGraph Neural NetworkTransformerLarge Language ModelPrompt EngineeringTextGraphRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose the KG-CoT framework, combining a small-scale graph reasoning model with a large LLM. The framework leverages knowledge graphs to generate high-confidence chain-of-thought reasoning paths, providing interpretable and responsible knowledge support for LLMs.
Knowledge Compilation for Incremental and Checkable Stochastic Boolean Satisfiability
Che Cheng (National Taiwan University), Jie-Hong R. Jiang (National Taiwan University)
Explainability and InterpretabilityComputational EfficiencyBenchmark
🎯 What it does: Proposed a knowledge compilation method based on levelized dec-DNNF to implement incremental queries and verifiable SSAT solving in the SSAT solver SharpSSAT, and extended CPOG to generate verifiable proof logs.
KTCN: Enhancing Open-World Object Detection with Knowledge Transfer and Class-Awareness Neutralization
Xing Xi (South China University of Technology), Ronghua Luo (South China University of Technology)
Object DetectionConvolutional Neural NetworkTransformerImage
🎯 What it does: In the open-world object detection task, the KTCN framework is proposed, which utilizes the Segment Anything Model (SAM) to generate pseudo labels and improves the recall rate of unknown categories by combining the Dual Matching Label Assignment method with the Class-Awareness Neutralizer method.
Label Distribution Learning from Logical Label
Yuheng Jia (Southeast University), Jiahao Jiang (Southeast University)
ClassificationImage
🎯 What it does: Proposed an algorithm called DLDL that directly learns a label distribution model from logical labels.
Label-efficient Semantic Scene Completion with Scribble Annotations
Song Wang (Zhejiang University), Jianke Zhu (Zhejiang University)
SegmentationAutonomous DrivingKnowledge DistillationConvolutional Neural NetworkTransformerPoint CloudBenchmark
🎯 What it does: Established the ScribbleSC benchmark for semantic scene completion requiring only sparse scribble annotations, and proposed the Scribble2Scene weakly supervised method, which utilizes geometric self-supervised labelers and range-guided offline-to-online distillation to achieve 3D semantic occupancy prediction under sparse annotations.
Langshaw: Declarative Interaction Protocols Based on Sayso and Conflict
Munindar P. Singh (North Carolina State University), Amit K. Chopra (Lancaster University)
Text
🎯 What it does: Proposes the Langshaw language, which defines multi-agent interaction protocols declaratively, along with synchronous semantics and asynchronous compilation methods.
Large Language Model as a Policy Teacher for Training Reinforcement Learning Agents
Zihao Zhou (Zhejiang Lab), Bin Liu (Zhejiang Lab)
Knowledge DistillationTransformerLarge Language ModelReinforcement LearningPrompt EngineeringTextChain-of-Thought
🎯 What it does: This paper proposes the LLM4Teach framework, which utilizes a pre-trained large language model (LLM) as a teacher to provide uncertainty-aware soft instructions to a lightweight reinforcement learning (RL) student agent. Through policy distillation, the student rapidly acquires high-level planning capabilities in the early training phase and gradually transitions to relying solely on environmental feedback to improve performance in later stages.
Large Language Model Guided Knowledge Distillation for Time Series Anomaly Detection
Chen Liu (Zhejiang University), Wenchao Meng (Zhejiang University)
Anomaly DetectionKnowledge DistillationTransformerLarge Language ModelContrastive LearningTime Series
🎯 What it does: Propose AnomalyLLM, which trains a student network to detect time series anomalies through knowledge distillation combined with a pre-trained large language model as the teacher network.
Large Language Model-Enhanced Algorithm Selection: Towards Comprehensive Algorithm Representation
Xingyu Wu (Hong Kong Polytechnic University), Kay Chen Tan (Hong Kong Polytechnic University)
OptimizationRepresentation LearningRecurrent Neural NetworkTransformerLarge Language ModelTextBenchmark
🎯 What it does: Proposed a framework for algorithm selection based on large language models, AS-LLM, which matches by combining high-dimensional representations extracted from algorithm code text with problem features.
Large Language Models Are Not Strong Abstract Reasoners
Gaël Gendron (University of Auckland), Gillian Dobbie (University of Auckland)
TransformerSupervised Fine-TuningPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: This paper proposes a new benchmark for evaluating the abstract reasoning ability of Large Language Models (LLMs) and conducts a systematic assessment of the performance of various state-of-the-art LLMs on this benchmark.
Layered and Staged Monte Carlo Tree Search for SMT Strategy Synthesis
Zhengyang Lu (University of Waterloo), Vijay Ganesh (Georgia Institute of Technology)
OptimizationComputational EfficiencyBenchmark
🎯 What it does: This paper proposes Z3alpha, an automated SMT strategy synthesis method based on Monte Carlo Tree Search (MCTS), designed to generate efficient solving strategies on given instance sets.
Layered Graph Security Games
Jakub Cerny (Columbia University), Garud Iyengar (Columbia University)
OptimizationGraph
🎯 What it does: This paper proposes the Layered Graph Security Games (LGSG) framework, which compresses the strategy spaces of two parties into paths in a layered directed acyclic graph (DAG), addressing large-scale but structured security games.
Laying the Foundations for Solving FOND HTN Problems: Grounding, Search, Heuristics (and Benchmark Problems)
Mohammad Yousefi (Australian National University), Pascal Bercher (Australian National University)
OptimizationBenchmark
🎯 What it does: This paper proposes the first method capable of solving hierarchical task network (HTN) problems in fully observable non-deterministic (FOND) environments. The approach encompasses 'all-outcome determinization' compilation from FOND HTN domains to deterministic domains, a strong solution search algorithm based on AO*, heuristic evaluation leveraging deterministic HTN and classical planning heuristics, and corresponding solver implementation.
LEAP: Optimization Hierarchical Federated Learning on Non-IID Data with Coalition Formation Game
Jianfeng Lu (Wuhan University of Science and Technology), Yun Xin (Wuhan University of Science and Technology)
OptimizationFederated LearningConvolutional Neural NetworkImage
🎯 What it does: Propose a federated learning optimization method named LEAP, which utilizes coalition formation games and gradient projection to adaptively adjust edge server associations, bandwidth allocation, and transmission power in multi-layer edge federated learning, aiming to reduce cross-edge data heterogeneity and communication overhead.
Learning a Spiking Neural Network for Efficient Image Deraining
Tianyu Song (Dalian Polytechnic University), Jiyu Jin (Dalian Polytechnic University)
RestorationComputational EfficiencySpiking Neural NetworkImage
🎯 What it does: Proposed an efficient single-image de-raining model ESDNet based on spiking neural networks, addressing the challenges of binary activation information loss and training difficulties caused by rain pixels.
Learning Big Logical Rules by Joining Small Rules
Céline Hocquette (University of Oxford), Andrew Cropper (University of Oxford)
OptimizationExplainability and InterpretabilityImageTextBenchmark
🎯 What it does: Proposed and implemented JOINER, a constraint-solving based ILP system that learns large rules by joining small rules, enabling the learning of rules with more than 100 literals.
Learning Causally Disentangled Representations via the Principle of Independent Causal Mechanisms
Aneesh Komanduri (University of Arkansas), Xintao Wu (University of Arkansas)
Representation LearningFlow-based ModelAuto Encoder
🎯 What it does: Constructed the ICM-VAE framework based on ICM principles, achieving the learning of supervised causal decomposition representations
Learning Conditional Preference Networks: An Approach Based on the Minimum Description Length Principle
Pierre-François Gimenez, Jérôme Mengin
Recommendation SystemTabular
🎯 What it does: This study proposes an unsupervised learning method based on the Minimum Description Length (MDL) principle to automatically infer Conditional Preference Networks (CP-nets) from sales history data, and evaluates it on car configuration recommendation tasks.
Learning Embeddings for Sequential Tasks Using Population of Agents
Mridul Mahajan (Max Planck Institute for Software Systems), Adish Singla (Max Planck Institute for Software Systems)
Representation LearningReinforcement LearningAgentic AISequential
🎯 What it does: Proposes a framework based on information theory, using a diverse proxy family to estimate the mutual information between tasks, thereby learning interpretable, fixed-dimension task embedding vectors, capturing task similarity through inner product, and ranking task difficulty through norm;
Learning Fair Cooperation in Mixed-Motive Games with Indirect Reciprocity
Martin Smit (University of Amsterdam), Fernando P. Santos (University of Amsterdam)
Reinforcement Learning
🎯 What it does: This paper studies how fairness and cooperation can be achieved through indirect reciprocity in mixed-motive games with group identities; first, evolutionary game theory is used to identify social norms that can stably sustain cooperation and fairness, followed by validation of these norms' reachability under learning dynamics using independent Q-learning agents.
Learning Fair Representations for Recommendation via Information Bottleneck Principle
Junsong Xie (Hefei University Of Technology), Le Wu (Hefei University Of Technology)
Recommendation SystemGraph Neural NetworkContrastive LearningGraphTabular
🎯 What it does: Propose a fair representation learning method (FairIB) based on the Information Bottleneck principle to generate user representations in recommendation systems that retain collaborative filtering information while suppressing sensitive attribute information.
Learning from Long-Tailed Noisy Data with Sample Selection and Balanced Loss
Lefan Zhang (Nanjing University), Wei Wang (Nanjing University)
ClassificationImage
🎯 What it does: This paper proposes a robust method for training on long-tailed noisy data. It first divides data into clean-labeled and unlabelled sets through class-aware sample selection, then performs semi-supervised training using a balance loss based on model bias.
Learning Generalized Policies for Fully Observable Non-Deterministic Planning Domains
Till Hofmann (RWTH Aachen University), Hector Geffner (RWTH Aachen University)
OptimizationBenchmark
🎯 What it does: This paper proposes a general strategy learning method based on combinatorial optimization, capable of learning rule-based policies in fully observable non-deterministic (FOND) planning domains.
Learning Hierarchy-Enhanced POI Category Representations Using Disentangled Mobility Sequences
Hongwei Jia (Shandong University), Yongshun Gong (Shandong University)
Recommendation SystemRepresentation LearningRecurrent Neural NetworkTransformerSupervised Fine-TuningSequential
🎯 What it does: Proposed and implemented a semantic separation POI category embedding model (SD-CEM) based on decoupled mobile sequences, generating hierarchically enhanced POI category representations.
Learning in CubeRes Model Space for Anomaly Detection in 3D GPR Data
Xiren Zhou (University of Science and Technology of China), Huanhuan Chen (University of Science and Technology of China)
Anomaly DetectionRecurrent Neural NetworkPhysics Related
🎯 What it does: This paper proposes a method for three-dimensional ground-penetrating radar (3D GPR) anomaly detection within the CubeRes model space.
Learning Label Dependencies for Visual Information Extraction
Minghong Yao (University of Science and Technology of China), Jiuchang Wei (University of Science and Technology of China)
RecognitionTransformerVision Language ModelContrastive LearningMultimodalityBenchmark
🎯 What it does: This paper proposes a framework to enhance visual information extraction (VIE) performance by learning long-range dependencies between entity labels. The framework includes three Transformers: the Feature Transformer extracts text, layout, and image features; the Label Transformer (pre-trained as a language model) learns long-range transitions in the label sequence; and the Inference Transformer generates label sequences that conform to label dependencies during testing through iterative fixed-point approximation mechanisms.
Learning Label-Specific Multiple Local Metrics for Multi-Label Classification
Jun-Xiang Mao (Southeast University), Min-Ling Zhang (Southeast University)
ClassificationOptimizationImageBenchmark
🎯 What it does: Propose a Label-Specific Multiple Metric (LSMM) framework that learns semantic similarity in multi-label data by combining a global metric with multiple label-specific local metrics, thereby enhancing the performance of KNN-based multi-label classifiers.
Learning Logic Programs by Discovering Higher-Order Abstractions
Céline Hocquette (University of Oxford), Andrew Cropper (University of Oxford)
OptimizationRepresentation Learning
🎯 What it does: Proposes a high-order reconstruction problem, using STEVis to automatically discover and utilize high-order abstractions (such as map, filter, fold) to compress logic programs, thereby enhancing ILP learning performance.
Learning Low-Rank Tensor Cores with Probabilistic ℓ0-Regularized Rank Selection for Model Compression
Tianxiao Cao (ShanghaiTech University), Hiroshi Mamitsuka (Kyoto University)
CompressionImageText
🎯 What it does: Propose a method for automatic tensor rank selection using probabilistic ℓ0 regularization and approximate Bernoulli gating, jointly training deep network weights and tensor decomposition ranks to achieve high compression rates.
Learning Multi-Granularity and Adaptive Representation for Knowledge Graph Reasoning
Ziyu Shang (Southeast University), Yining Li (Southeast University)
Computational EfficiencyRepresentation LearningGraph Neural NetworkTransformerGraphBenchmark
🎯 What it does: Propose a multi-granularity adaptive representation framework called MulGA, which efficiently learns triplet, relation path, and subgraph representations required for knowledge graph reasoning through connected subgraphs.
Learning Pareto Set for Multi-Objective Continuous Robot Control
Tianye Shu (Southern University of Science and Technology), Hisao Ishibuchi (Southern University of Science and Technology)
OptimizationRobotic IntelligenceReinforcement LearningBenchmark
🎯 What it does: Propose a single Hypernet that learns the Pareto set for multi-objective continuous robot control problems, representing the entire Pareto front as a low-dimensional curve in a high-dimensional parameter space.
Learning Robust Classifiers with Self-Guided Spurious Correlation Mitigation
Guangtao Zheng (University of Virginia), Aidong Zhang (University of Virginia)
ClassificationVision Language ModelImage
🎯 What it does: Propose a self-guided, unlabeled pseudo-correlation elimination framework called LBC, which automatically leverages pre-trained vision-language models (VLMs) to extract image attributes, computes pseudo-correlation scores (spuriousness scores) between each attribute and category, clusters in the spuriousness embedding space to generate fine-grained labels, and trains a classifier less sensitive to pseudo-correlations and with higher robustness by reshaping the classification head into (K·C) classes, balanced sampling, and iterative updates.
Learning Spatial Similarity Distribution for Few-shot Object Counting
Yuanwu Xu (Nanjing University of Science and Technology), Haofeng Zhang (Nanjing University of Science and Technology)
Object DetectionMeta LearningConvolutional Neural NetworkImage
🎯 What it does: Propose a few-shot object counting network SSD based on 4D spatial similarity distribution, combining feature cross enhancement (FCE), dynamic image scaling, and generalization loss;
Learning to Solve Geometry Problems via Simulating Human Dual-Reasoning Process
Tong Xiao (Institute of Artificial Intelligence, Hefei Comprehensive National Science Center), Enhong Chen (Institute of Artificial Intelligence, Hefei Comprehensive National Science Center)
Explainability and InterpretabilityKnowledge DistillationRepresentation LearningRecurrent Neural NetworkTransformerLarge Language ModelVision Language ModelImageTextMultimodalityRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Proposed a DualGeoSolver that solves geometric problems by simulating the dual reasoning process of humans.
Learning Translations: Emergent Communication Pretraining for Cooperative Language Acquisition
Dylan Cope (King's College London), Peter McBurney (King's College London)
Reinforcement LearningText
🎯 What it does: In multi-agent collaborative environments, the Cooperative Language Acquisition Problem (CLAP) challenge is proposed, constructing a joiner agent to learn and adapt to the communication protocols of the target community.
Learning What to Monitor: Using Machine Learning to Improve past STL Monitoring
Andrea Brunello (University of Udine), Nicola Saccomanno (University of Udine)
Anomaly DetectionTime Series
🎯 What it does: This paper proposes combining runtime monitoring with machine learning for preemptive detection of system failures.
Learning with Posterior Sampling for Revenue Management under Time-varying Demand
Kazuma Shimizu (NEC Corporation), Shinji Nakadai (NEC Corporation)
OptimizationReinforcement LearningTabularTime SeriesFinance Related
🎯 What it does: This paper discusses the revenue management problem under time-varying demand, proposing an algorithm based on posterior sampling to optimize pricing and maximize revenue.
Learning-Based Tracking-before-Detect for RF-Based Unconstrained Indoor Human Tracking
Zhi Wu (University of Science and Technology of China), Yan Chen (University of Science and Technology of China)
Object TrackingConvolutional Neural NetworkRecurrent Neural NetworkImage
🎯 What it does: Propose NeuralTBD, an end-to-end deep learning-based Track-Before-Detect (TBD) framework for multi-person human trajectory tracking using RF radar in unconstrained indoor environments.
LeMeViT: Efficient Vision Transformer with Learnable Meta Tokens for Remote Sensing Image Interpretation
Wentao Jiang (Wuhan University), Bo Du (Wuhan University)
ClassificationRecognitionObject DetectionSegmentationTransformerImage
🎯 What it does: Propose a new visual Transformer architecture called LeMeViT, which achieves sparse representation in high-resolution remote sensing images by utilizing learnable meta tokens, and significantly reduces the computational complexity of self-attention through Dual Cross-Attention (DCA).
LeRet: Language-Empowered Retentive Network for Time Series Forecasting
Qihe Huang (University of Science and Technology of China), Yang Wang (University of Science and Technology of China)
TransformerLarge Language ModelTime Series
🎯 What it does: Proposes the LeRet framework, which splits time series into patches, employs RetNet to capture causal dependencies, pretrains on patch-level autoregressive tasks, leverages pre-trained large language models (LLMs) to extract time series-related language knowledge, and achieves language-enabled sequence prediction through two-stage cross-modal integration.
Let’s Start Over: Retraining with Selective Samples for Generalized Category Discovery
Zhimao Peng (Nankai University), Ming-Ming Cheng (Nankai University)
Representation LearningTransformerSupervised Fine-TuningContrastive LearningImage
🎯 What it does: In the generalized category discovery (GCD) task, a sample selection strategy based on nearest neighbor distance-aware label consistency is proposed. The selected high-purity samples are used as a pseudo-labeled set for retraining, thereby improving the quality of representation learning without new category labels.
LG-FGAD: An Effective Federated Graph Anomaly Detection Framework
Jinyu Cai (National University of Singapore), See-Kiong Ng (National University of Singapore)
Anomaly DetectionFederated LearningSafty and PrivacyKnowledge DistillationGraph Neural NetworkGenerative Adversarial NetworkGraph
🎯 What it does: Proposed a federated graph anomaly detection framework named LG-FGAD, which generates abnormal graphs at each client via a self-adversarial generator, detects anomalies using a discriminator, and achieves model personalization and collaborative learning through a local-global mutual information module and dual knowledge distillation.
LG-GNN: Local-Global Adaptive Graph Neural Network for Modeling Both Homophily and Heterophily
Zhizhi Yu (Tianjin University), Zhiyong Feng (Tianjin University)
Graph Neural NetworkGraph
🎯 What it does: Proposed a local-global adaptive graph neural network called LG-GNN for simultaneously modeling homogeneous and heterogeneous graphs.
Linear-Time Optimal Deadlock Detection for Efficient Scheduling in Multi-Track Railway Networks
Hastyn Doshi (IIT Bombay), Shivaram Kalyanakrishnan (IIT Bombay)
OptimizationGraph
🎯 What it does: Proposed a linear-time optimal deadlock detection rule R0 based on the construction of a 'next station graph,' and applied it to rolling planning and recovery scheduling in multi-track railway dispatching.
LitE-SNN: Designing Lightweight and Efficient Spiking Neural Network through Spatial-Temporal Compressive Network Search and Joint Optimization
Qianhui Liu (National University Of Singapore), Haizhou Li (Chinese University Of Hong Kong, Shenzhen)
Computational EfficiencyNeural Architecture SearchConvolutional Neural NetworkSpiking Neural NetworkImageAudio
🎯 What it does: Propose the LitE-SNN method, utilizing spatial compression with CompConv and temporal compression with stride search to achieve lightweight and efficient Spiking Neural Network (SNN) automated design.
LLM-based Multi-Level Knowledge Generation for Few-shot Knowledge Graph Completion
Qian Li (Beijing University of Posts and Telecommunications), Jianxin Li (Beihang University)
Knowledge DistillationRepresentation LearningMeta LearningTransformerLarge Language ModelPrompt EngineeringContrastive LearningMultimodalityGraph
🎯 What it does: This paper proposes the MuKDC framework, which utilizes LLM distillation to achieve multi-layer knowledge generation and consistency evaluation, addressing the long-tail scarcity problem in few-shot knowledge graph completion (FKGC).
LLMem: Estimating GPU Memory Usage for Fine-Tuning Pre-Trained LLMs
Taeho Kim (University of Colorado Boulder), Sangtae Ha (University of Colorado Boulder)
Computational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes LLMem, a framework that accurately predicts GPU memory consumption during fine-tuning of large language models (LLMs) in a multi-GPU environment and automatically selects the optimal distributed fine-tuning method based on the prediction results.
LLMs Can Find Mathematical Reasoning Mistakes by Pedagogical Chain-of-Thought
Zhuoxuan Jiang (Shanghai Business School), Dongsheng Li (Microsoft Research Asia)
RecognitionTransformerPrompt EngineeringTextChain-of-Thought
🎯 What it does: This paper proposes a zero-shot prompting strategy called Pedagogical Chain-of-Thought (PedCoT) to enable large language models (LLMs) to identify errors in mathematical reasoning steps.
LocMoE: A Low-overhead MoE for Large Language Model Training
Jing Li (Huawei Technologies Co., Ltd), Xin Chen (Huawei Technologies Co., Ltd)
Computational EfficiencyLarge Language ModelMixture of ExpertsText
🎯 What it does: This paper proposes LocMoE, a low-overhead Mixture-of-Experts routing and communication strategy, addressing the load imbalance and All-to-All communication latency issues in traditional MoE during large language model training.
Long Short-Term Dynamic Prototype Alignment Learning for Video Anomaly Detection
Chao Huang (Sun Yat-sen University), Yabo Liu (Harbin Institute of Technology)
Anomaly DetectionConvolutional Neural NetworkRecurrent Neural NetworkContrastive LearningVideo
🎯 What it does: Proposed the prototype-guided dynamic matching network (PDM-Net), which achieves long-distance frame prediction through long-term and short-term dynamic prototype alignment for video anomaly detection.
LongVQ: Long Sequence Modeling with Vector Quantization on Structured Memory
Zicheng Liu (Zhejiang University), Stan Z. Li (Westlake University)
Computational EfficiencyRepresentation LearningTransformerImageTextSequentialAudio
🎯 What it does: The paper proposes a hybrid model called LongVQ, which achieves linear time complexity for long sequence self-attention by compressing global information using vector quantization.
Look-ahead Search on Top of Policy Networks in Imperfect Information Games
Ondřej Kubíček (Czech Technical University in Prague), Viliam Lisý (Czech Technical University in Prague)
Reinforcement LearningBenchmark
🎯 What it does: Propose a framework named SePoT that achieves safe search at any public state during testing without requiring search during training, by training an additional critic network on policy gradient-based models (e.g., RNaD).
LSPAN: Spectrally Localized Augmentation for Graph Consistency Learning
Heng-Kai Zhang (Nanjing University), Yu-Feng Li (Nanjing University)
ClassificationRepresentation LearningGraph Neural NetworkContrastive LearningGraphBenchmark
🎯 What it does: Propose a graph data augmentation method called LSPAN, which performs local spectral perturbation in the frequency domain and applies it to a graph consistency learning framework, significantly improving node classification performance.
M2Beats: When Motion Meets Beats in Short-form Videos
Dongxiang Jiang (Beijing University of Posts and Telecommunications), Anlong Ming (Beijing University of Posts and Telecommunications)
RecognitionPose EstimationConvolutional Neural NetworkGraph Neural NetworkVideoMultimodalityAudio
🎯 What it does: Propose M2Beats, a framework for action beat detection and enhancement in short videos, which includes a high-quality action beat dataset AIST-M2B, a spatiotemporal graph convolutional network-based model M2BNet, and an algorithm for aligning and enhancing action beats with music beats.
Machine Unlearning via Null Space Calibration
Huiqiang Chen (University of Technology Sydney), Wanlei Zhou (City University of Macau)
Safty and PrivacyImage
🎯 What it does: Proposes a machine unlearning method called UNSC under zero-space calibration, which efficiently and accurately enables the model to forget specific samples when data deletion requests are made.
MacMic: Executing Iceberg Orders via Hierarchical Reinforcement Learning
Hui Niu (Tsinghua University), Jian Li (Tsinghua University)
Reinforcement LearningTime SeriesFinance Related
🎯 What it does: Proposes the MacMic framework based on hierarchical reinforcement learning to address the long-term (4-hour) iceberg order execution problem, decomposing the hierarchical MDP into two stages: volume scheduling and order placement, and learning multi-grained market representations through SHMM.
Maintaining Diversity Provably Helps in Evolutionary Multimodal Optimization
Shengjie Ren (Nanjing University), Chao Qian (Nanjing University)
OptimizationBenchmark
🎯 What it does: This paper introduces a diversity maintenance method based on solution space diversity into genetic algorithms and multi-objective evolutionary algorithms. Theoretical proofs and experimental validations demonstrate that this method can significantly accelerate the solving of multi-modal optimization problems (Jump and OneJumpZeroJump).
Make Bricks with a Little Straw: Large-Scale Spatio-Temporal Graph Learning with Restricted GPU-Memory Capacity
Binwu Wang (University of Science and Technology of China), Yang Wang (University of Science and Technology of China)
Computational EfficiencyRepresentation LearningConvolutional Neural NetworkGraph Neural NetworkGraphTime Series
🎯 What it does: To address large-scale spatiotemporal graph learning, this paper proposes the LarSTL framework, which divides the complete graph into multiple overlapping subgraphs, trains them sequentially according to the task flow, and achieves continuous learning through experience replay and feature adapters.
Make Graph Neural Networks Great Again: A Generic Integration Paradigm of Topology-Free Patterns for Traffic Speed Prediction
Yicheng Zhou (University of Macau), Pengyang Wang (University of Macau)
Knowledge DistillationGraph Neural NetworkTransformerTime Series
🎯 What it does: Propose a dual-layer cross-scale Transformer (DCST) that captures topology-agnostic patterns in traffic speed prediction by aggregating spatial and temporal scales, and injects topological regularization patterns from existing GNN models into DCST via a teacher-student distillation framework to achieve fusion of both types of patterns;
Making LLMs as Fine-Grained Relation Extraction Data Augmentor
Yifan Zheng (Southeast University), Zhi Fang (Beijing Institute of Computer Technology and Application)
Data-Centric LearningTransformerLarge Language ModelPrompt EngineeringTextRetrieval-Augmented Generation
🎯 What it does: This paper proposes the ConsistRE framework, which achieves data augmentation for relation extraction by leveraging large language models, keyword prompts, and syntactic dependency tree selection.
MARCO: A Memory-Augmented Reinforcement Framework for Combinatorial Optimization
Andoni I. Garmendia (University of the Basque Country), Alexander Mendiburu (University of the Basque Country)
OptimizationGraph Neural NetworkTransformerReinforcement LearningGraph
🎯 What it does: Proposed the MARCO framework, which integrates a memory module with construction and improvement methods from neural combinatorial optimization (NCO) to enhance search efficiency and solution quality.
Markov Constraint as Large Language Model Surrogate
Alexandre Bonlarron (Universite Cote Dazur), Jean-Charles Régin
GenerationComputational EfficiencyLarge Language ModelText
🎯 What it does: Propose the NgramMarkov constraint, integrating the probability of large language models (LLM) into Markov constraints to filter sentences during the text generation process
MARS: Multimodal Active Robotic Sensing for Articulated Characterization
Hongliang Zeng (South China University of Technology), Fang Li (South China University of Technology)
Pose EstimationRobotic IntelligenceConvolutional Neural NetworkTransformerReinforcement LearningImageMultimodalityPoint Cloud
🎯 What it does: The MARS framework achieves precise perception of joint position, orientation, and status by fusing RGB images with point clouds and combining an active sensing strategy;
MAS-SAM: Segment Any Marine Animal with Aggregated Features
Tianyu Yan (Dalian University of Technology), Huchuan Lu (Dalian University of Technology)
SegmentationTransformerSupervised Fine-TuningImage
🎯 What it does: Proposed the MAS-SAM framework, customizing the Segment Anything Model (SAM) for marine animal segmentation, primarily by incorporating LoRA and Adapter into the SAM encoder for low-parameter fine-tuning, constructing a Hypermap Extraction Module (HEM) to extract multi-scale features, and designing a Progressive Prediction Decoder (PPD) and Fusion Attention Module (FAM) to achieve multi-source feature fusion, ultimately achieving fine-grained segmentation results.
Massively Parallel Single-Source SimRanks in O(log N) Rounds
Siqiang Luo (Nanyang Technological University), Zulun Zhu (Nanyang Technological University)
Computational EfficiencyGraph
🎯 What it does: This paper proposes a distributed algorithm for single-source SimRank computation under the Massively Parallel Computation (MPC) model, which can complete the computation within O(log² log n) communication rounds, using sublinear space per machine, and provides an ϵ-absolute error guarantee;
MCM: Multi-condition Motion Synthesis Framework
Zeyu Ling (Zhejiang University), Weidong Geng (Zhejiang University)
GenerationData SynthesisTransformerVision Language ModelDiffusion modelTextMultimodalityAudio
🎯 What it does: Propose a multimodal (text + audio) conditional human action synthesis framework called MCM, incorporating a multi-dimensional attention Transformer model named MWNet in its backbone;
Mean Aggregator Is More Robust than Robust Aggregators under Label Poisoning Attacks
Jie Peng (Sun Yat-Sen University), Qing Ling (Sun Yat-Sen University)
Federated LearningConvolutional Neural NetworkImage
🎯 What it does: In distributed learning, this paper investigates the robustness of the mean aggregator and existing robust aggregators under label poisoning attacks, and proves that the mean aggregator is theoretically and practically superior when data heterogeneity is sufficiently large;
Mechanisms That Play a Game, Not Toss a Coin
Toby Walsh (UNSW Sydney)
OptimizationAgentic AI
🎯 What it does: Proposed a method to derandomize a random mechanism by letting agents play modulo arithmetic games, maintaining the normative advantages of the original random mechanism while achieving deterministic and audit-friendly mechanisms.
MediTab: Scaling Medical Tabular Data Predictors via Data Consolidation, Enrichment, and Refinement
Zifeng Wang (University Of Illinois Urbana Champaign), Jimeng Sun (University Of Illinois Urbana Champaign)
Data-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningTabularBiomedical DataElectronic Health Records
🎯 What it does: Propose MediTab, which leverages LLM to convert multi-source medical tables into unified text, achieving data merging, augmentation, and correction. A single model can be trained to make predictions on any table format without requiring fine-tuning.
MEDVOC: Vocabulary Adaptation for Fine-tuning Pre-trained Language Models on Medical Text Summarization
Gunjan Balde (Indian Institute of Technology Kharagpur), Niloy Ganguly (Indian Institute of Technology Kharagpur)
GenerationHyperparameter SearchData-Centric LearningTransformerSupervised Fine-TuningTextBiomedical Data
🎯 What it does: This paper proposes a dynamic vocabulary adaptation strategy called MEDVOC for fine-tuning pre-trained language models, thereby improving the quality of medical text summarization.
Memorizing Documents with Guidance in Large Language Models
Bumjin Park (KAIST AI), Jaesik Choi (KAIST AI)
Explainability and InterpretabilityRepresentation LearningTransformerLarge Language ModelText
🎯 What it does: Propose a document-based memory architecture and document-guided loss, enabling large language models to map document content to trackable memory entries during training.
Meta In-Context Learning Makes Large Language Models Better Zero and Few-Shot Relation Extractors
Guozheng Li (Southeast University), Zijie Xu (Southeast University)
ClassificationMeta LearningTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextTabular
🎯 What it does: Propose the MICRE framework, which utilizes meta-learning to train LLMs for context learning, aiming to enhance performance in zero-shot and few-shot relation extraction tasks.
Meta-Learning via PAC-Bayesian with Data-Dependent Prior: Generalization Bounds from Local Entropy
Shiyu Liu (University of Electronic Science and Technology of China), Hui Wang (Peng Cheng Laboratory)
Safty and PrivacyMeta LearningImageTime SeriesStochastic Differential Equation
🎯 What it does: The paper proposes a meta-learning framework based on PAC-Bayesian theory, which introduces data-dependent priors and utilizes local entropy optimization to derive a closed-form optimal posterior, thereby achieving efficient meta-learning without requiring nested optimization.
MetaISP: Efficient RAW-to-sRGB Mappings with Merely 1M Parameters
Zigeng Chen (National University Of Singapore), Xinchao Wang (National University Of Singapore)
Image TranslationRestorationConvolutional Neural NetworkImage
🎯 What it does: Propose MetaISP, an adaptive RAW-to-sRGB deep ISP model that dynamically adjusts network parameters and structure based on each input image, significantly improving reconstruction quality.
MetaJND: A Meta-Learning Approach for Just Noticeable Difference Estimation
Miaohui Wang (Shenzhen University), Wuyuan Xie (Shenzhen University)
CompressionMeta LearningTransformerMultimodality
🎯 What it does: Propose the MetaJND framework, which utilizes RGB, saliency, and depth tri-modal meta-learning for pixel-level JND prediction;
Metric Distortion with Elicited Pairwise Comparisons
Soroush Ebadian (University of Toronto), Evi Micha (Harvard University)
Recommendation SystemReinforcement Learning from Human FeedbackText
🎯 What it does: Studied how to design algorithms to effectively aggregate individual preferences with limited pairwise comparisons, and used a metric distortion framework to compare the effectiveness of different outcomes.
MFTraj: Map-Free, Behavior-Driven Trajectory Prediction for Autonomous Driving
Haicheng Liao (University of Macau), Chengzhong Xu (University of Macau)
Autonomous DrivingRecurrent Neural NetworkGraph Neural NetworkTransformerTime SeriesSequential
🎯 What it does: Proposed MFTraj, a behavior-driven trajectory prediction model that does not rely on high-definition maps.
MGCBS: An Optimal and Efficient Algorithm for Solving Multi-Goal Multi-Agent Path Finding Problem
Mingkai Tang (Hong Kong University of Science and Technology), Lujia Wang (Hong Kong University of Science and Technology (Guangzhou))
OptimizationGraph
🎯 What it does: Proposed MGCBS, a two-layer optimal algorithm to address the multi-target multi-agent path planning problem.
MICRO: Model-Based Offline Reinforcement Learning with a Conservative Bellman Operator
Xiao-Yin Liu (Chinese Academy of Sciences), Zeng-Guang Hou (University of Chinese Academy of Sciences)
Reinforcement LearningBenchmark
🎯 What it does: Proposed the MICRO model-based offline reinforcement learning algorithm, achieving a balance between performance and robustness by introducing a conservative Bellman operator;
Minimizing Weighted Counterfactual Regret with Optimistic Online Mirror Descent
Hang Xu (Institute of Automation, Chinese Academy of Sciences), Jian Cheng (Institute of Automation, Chinese Academy of Sciences)
OptimizationBenchmark
🎯 What it does: This paper proposes a new PDCFR+ algorithm that utilizes weighted online mirror descent (OMD) and its optimistic variant to minimize counting regret in weighted counting hypothesis uncertainty games.
MISA: MIning Saliency-Aware Semantic Prior for Box Supervised Instance Segmentation
Hao Zhu (Chinese Academy of Sciences), Feng Dai (University of Chinese Academy of Sciences)
SegmentationKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: Proposes the MISA framework, which enhances box-supervised instance segmentation performance by mining significant semantic priors from box-supervised semantic segmentation networks and leveraging cross-model guidance.
MLP-DINO: Category Modeling and Query Graphing with Deep MLP for Object Detection
Guiping Cao (Southern University of Science and Technology), Yaowei Wang (Peng Cheng Laboratory)
Object DetectionTransformerImage
🎯 What it does: Proposed MLP-DINO, which integrates a deep MLP backbone network with the DINO framework, and introduces two strategies, QICS and GQS, to decouple class prediction from box regression and balance the query distribution;
MMGNN: A Molecular Merged Graph Neural Network for Explainable Solvation Free Energy Prediction
Wenjie Du (University of Science and Technology of China), Yang Wang (University of Science and Technology of China)
Explainability and InterpretabilityDrug DiscoveryGraph Neural NetworkGraph
🎯 What it does: Proposed and implemented the MMGNN model for predicting the dissolution Gibbs free energy of solvent-solute systems.
MMVQA: A Comprehensive Dataset for Investigating Multipage Multimodal Information Retrieval in PDF-based Visual Question Answering
Yihao Ding (University of Melbourne), Soyeon Caren Han (University of Melbourne)
RetrievalTransformerVision Language ModelMultimodalityBiomedical DataBenchmark
🎯 What it does: Constructed the MMVQA dataset (multi-page, multi-modal PDF Visual Question Answering (VQA)), and proposed a joint coarse-to-fine entity retrieval framework capable of locating paragraphs, tables, figures, and other entities across pages in documents to answer questions.
Model Checking Causality
Tiago de Lima (University of Artois), Emiliano Lorini (Toulouse University)
🎯 What it does: Propose a causal-based modal language and provide its semantics and model checking methods
Model-Free Preference Elicitation
Carlos Martin (Carnegie Mellon University), Tuomas Sandholm (Carnegie Mellon University)
Recommendation SystemRecurrent Neural NetworkTabular
🎯 What it does: This paper proposes a model-free preference mining framework that directly learns user response probabilities to questions and the expected utility of recommendations through function approximation, achieving an expectation of information gain (EVOI)-driven multi-round querying strategy without relying on explicit Bayesian inference.
Modeling Personalized Retweeting Behaviors for Multi-Stage Cascade Popularity Prediction
Mingyang Zhou (Shenzhen University), Rui Mao (Shenzhen University)
Recommendation SystemConvolutional Neural NetworkGraph Neural NetworkTransformerLarge Language ModelTextGraph
🎯 What it does: This study proposes the CasMS framework for predicting the popularity of information diffusion from the message generation stage to the long-term stage.
Modeling Selective Feature Attention for Lightweight Text Matching
Jianxiang Zang (Shanghai University of International Business and Economics), Hui Liu (Shanghai University of International Business and Economics)
RetrievalRecurrent Neural NetworkText
🎯 What it does: Proposed Feature Attention (FA) and Selective Feature Attention (SFA) to enhance matching performance in lightweight text matching networks by modeling dependencies in the embedded feature layer.
MOSER: Learning Sensory Policy for Task-specific Viewpoint via View-conditional World Model
Shenghua Wan (Nanjing University), De-Chuan Zhan (Nanjing University)
Robotic IntelligenceReinforcement LearningWorld ModelImage
🎯 What it does: Propose a reinforcement learning framework MOSER based on a perspective-conditioned world model (VWM) and perception strategy, for automatically adjusting camera perspectives and completing tasks in visual RL;
Motion-Aware Heatmap Regression for Human Pose Estimation in Videos
Inpyo Song (Sungkyunkwan University), Jangwon Lee (Sungkyunkwan University)
Pose EstimationTransformerOptical FlowVideo
🎯 What it does: Designed a 2D human pose estimation framework based on video temporal information—Motion-Aware Heatmap Regression, which uses motion-aware heatmaps generated from joint motion vectors to guide keypoint regression.
MuEP: A Multimodal Benchmark for Embodied Planning with Foundation Models
Kanxue Li (Yunnan University), Xiaodong He (JD Explore Academy)
Robotic IntelligenceLarge Language ModelSupervised Fine-TuningVision Language ModelMultimodalityBenchmark
🎯 What it does: Proposed MuEP, a multi-modal embedded planning benchmark for evaluating embedded agents in complex scenarios with multi-round interactions;
Multi-Attention Based Visual-Semantic Interaction for Few-Shot Learning
Peng Zhao (Anhui University), Xiaochun Cao (Hong Kong Polytechnic University)
ClassificationRepresentation LearningMeta LearningConvolutional Neural NetworkContrastive LearningImageTextMultimodality
🎯 What it does: Proposes a multi-attention visual-semantic interaction method (MAVSI) for few-shot learning, which can simultaneously leverage semantic knowledge on both the support set and query set to enhance the discriminability of visual features.
Multi-Granularity Graph-Convolution-Based Method for Weakly Supervised Person Search
Haichun Tai (Xidian University), Xinbo Gao (Chongqing University of Posts and Telecommunications)
RecognitionObject DetectionRetrievalConvolutional Neural NetworkGraph Neural NetworkContrastive LearningImage
🎯 What it does: Propose a multi-granularity graph convolutional framework for weakly supervised person search, jointly optimizing detection and ReID while improving pseudo-label generation.
Multi-level Disentangling Network for Cross-Subject Emotion Recognition Based on Multimodal Physiological Signals
Ziyu Jia (Chinese Academy of Sciences), Tianzi Jiang (Chinese Academy of Sciences)
ClassificationRecognitionTransformerMultimodalityBiomedical Data
🎯 What it does: Proposed the MDNet multi-layer decoupling network for cross-subject emotion recognition, which simultaneously handles the consistency and heterogeneity of multimodal physiological signals as well as individual differences.
Multi-Modal Sarcasm Detection Based on Dual Generative Processes
Huiying Ma (Tianjin University), Longbiao Wang (Tianjin University)
ClassificationConvolutional Neural NetworkTransformerLarge Language ModelVision Language ModelMultimodality
🎯 What it does: This paper proposes a multimodal sarcasm detection method based on a dual generation process, which captures emotional inconsistencies between images and text through text-guided image generation features and image-guided text generation features. It utilizes a strong-weak modality mechanism to automatically adjust the weights of three modalities (text, image, modality inconsistency), thereby achieving sarcasm detection.