IJCAI 2024 Papers — Page 8
International Joint Conference on Artificial Intelligence · 790 papers
SVD-AE: Simple Autoencoders for Collaborative Filtering
Seoyoung Hong (Yonsei University), Noseong Park (Korea Advanced Institute of Science and Technology)
Recommendation SystemAuto EncoderTabular
🎯 What it does: Designed a linear autoencoder called SVD-AE based on truncated singular value decomposition (SVD) for closed-form solutions in collaborative filtering, achieving improved accuracy, efficiency, and noise robustness without iterative training.
SwiftThief: Enhancing Query Efficiency of Model Stealing by Contrastive Learning
Jeonghyun Lee (Korea University), Sangkyun Lee (Korea University)
ClassificationAdversarial AttackContrastive LearningImage
🎯 What it does: Propose SwiftThief, an efficient model stealing framework that leverages queried and unqueried data, soft supervised contrastive learning, and class imbalance priority sampling.
Symplectic Neural Gaussian Processes for Meta-learning Hamiltonian Dynamics
Tomoharu Iwata (NTT Corporation), Yusuke Tanaka (NTT Corporation)
Meta LearningRecurrent Neural NetworkTime SeriesSequentialPhysics Related
🎯 What it does: This paper proposes a meta-learning based Hamiltonian dynamics model that can quickly adapt to the dynamics of new systems with only a small amount of data.
Synthesizing Programmatic Policy for Generalization within Task Domain
Tianyi Wu (Fudan University), Wenyun Zhao (Fudan University)
Robotic IntelligenceMeta LearningRecurrent Neural NetworkReinforcement LearningBenchmark
🎯 What it does: Proposed a procedural strategy based on context-free grammar and recurrent neural networks, utilizing meta-learning to achieve task-domain generalization.
Tackling Stackelberg Network Interdiction against a Boundedly Rational Adversary
Tien Mai (Singapore Management University), Ayushman Kumar singh
OptimizationGraph
🎯 What it does: This paper addresses the Stackelberg network interdiction problem, assuming the adversary is a bounded rationality (dynamic quantized response) decision-maker, and proposes two solution methods: LiSD, a MILP solver based on path sampling and piecewise linear approximation, and DynP, an approach based on dynamic programming and gradient ascent.
TaD: A Plug-and-Play Task-Aware Decoding Method to Better Adapt LLMs on Downstream Tasks
Xinhao Xu (Tsinghua University), Guiguang Ding (Tsinghua University)
TransformerLarge Language ModelText
🎯 What it does: Improve the model's performance on downstream tasks by leveraging the difference in output probability distributions before and after fine-tuning, using task-aware decoding on a fine-tuned LLM.
TAI++: Text as Image for Multi-Label Image Classification by Co-Learning Transferable Prompt
Xiangyu Wu (Nanjing University of Science and Technology), Jianfeng Lu (Nanjing University of Science and Technology)
ClassificationTransformerPrompt EngineeringVision Language ModelContrastive LearningImage
🎯 What it does: Proposes a pseudo visual prompt (PVP) module and a transferable prompt co-learning strategy based on pre-trained vision-language models to achieve multi-label image classification without relying on large-scale annotated image data;
Task-Agnostic Self-Distillation for Few-Shot Action Recognition
Bin Zhang (Zhejiang University of Technology), Xiaofei He (Zhejiang University)
RecognitionKnowledge DistillationMeta LearningTransformerVideo
🎯 What it does: Propose a task-agnostic self-distillation framework combined with multi-dimensional distillation to enhance the generalization and fine-grained matching of few-shot action recognition;
Temporal Domain Generalization via Learning Instance-level Evolving Patterns
Yujie Jin (Peking University), Liantao Ma (Peking University)
Data SynthesisDomain AdaptationTabularTime SeriesStochastic Differential EquationOrdinary Differential Equation
🎯 What it does: Propose the CTOT framework, which generates temporal trajectories using instance-level optimal transport and predicts future domains through neural differential equations, enhancing Temporal Domain Generalization performance.
Temporal Graph ODEs for Irregularly-Sampled Time Series
Alessio Gravina (University of Pisa), Cesare Alippi (Swiss AI Lab IDSIA, Università della Svizzera italiana)
Computational EfficiencyRepresentation LearningGraph Neural NetworkTime SeriesOrdinary Differential Equation
🎯 What it does: Designed and verified a continuous-time graph neural network framework called TG-ODE that can handle irregularly sampled temporal graphs.
Temporal Inductive Logic Reasoning over Hypergraphs
Yuan Yang (Georgia Institute of Technology), Faramarz Fekri (Georgia Institute of Technology)
Autonomous DrivingGraph Neural NetworkVideoGraph
🎯 What it does: Propose an inductive logic reasoning framework called TILR, which models time intervals and higher-order relationships (hypergraphs), addressing the limitations of traditional ILP in temporal multi-relational contexts.
Temporal Knowledge Graph Extrapolation via Causal Subhistory Identification
Kai Chen (National University of Defense Technology), Aiping Li (National University of Defense Technology)
Explainability and InterpretabilityRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: Proposed a temporal knowledge graph extrapolation method called CSI based on causal sub-history identification, which uses attention masks to separate causal sub-histories from shortcut sub-histories in query-related history, and achieves backdoor adjustment through counterfactual intervention at the representation layer.
Ten Words Only Still Help: Improving Black-Box AI-Generated Text Detection via Proxy-Guided Efficient Re-Sampling
Yuhui Shi (Key Lab of Intelligent Information Processing of Chinese Academy of Sciences), Danding Wang (Key Lab of Intelligent Information Processing of Chinese Academy of Sciences)
Anomaly DetectionComputational EfficiencyTransformerLarge Language ModelText
🎯 What it does: This paper proposes a proxy-guided efficient resampling method called POGER for detecting AI-generated text (AIGT) in black-box environments, achieving classification by estimating word generation probabilities through resampling and combining them with context compensation.
TFCD: Towards Multi-modal Sarcasm Detection via Training-Free Counterfactual Debiasing
Zhihong Zhu (Peking University), Yefeng Zheng (Jarvis Research Center, Tencent YouTu Lab)
ClassificationTransformerLarge Language ModelVision Language ModelMultimodality
🎯 What it does: Proposed a training-agnostic counterfactual debiasing framework called TFCD for multimodal sarcasm detection, capable of eliminating statistical label bias and non-sarcastic word bias without retraining models or altering data distributions.
TFLOP: Table Structure Recognition Framework with Layout Pointer Mechanism
Minsoo Khang (Upstage AI), Teakgyu Hong (Upstage AI)
RecognitionTransformerVision Language ModelContrastive LearningImage
🎯 What it does: Proposed TFLOP, a table structure recognition framework based on the layout pointer mechanism, which directly predicts HTML structure using text region information, eliminating traditional bounding box prediction and post-processing steps.
TFWT: Tabular Feature Weighting with Transformer
Xinhao Zhang (Portland State University), Kunpeng Liu (Portland State University)
OptimizationRepresentation LearningData-Centric LearningTransformerReinforcement LearningTabular
🎯 What it does: Proposed a Transformer-based feature weighting method (TFWT) and refined the weights using reinforcement learning.
The Distortion of Threshold Approval Matching
Mohamad Latifian (University of Toronto), Alexandros A. Voudouris (University of Essex)
Optimization
🎯 What it does: This paper studies the matching problem when only the threshold approval information of agents towards items is known, providing upper and lower bounds on the distortion of deterministic and randomized mechanisms under one-sided matching and more general capacity/supply constraints, and establishing a tight relationship with the number of thresholds.
The Impact of Features Used by Algorithms on Perceptions of Fairness
Andrew Estornell (Washington University in Saint Louis), Yevgeniy Vorobeychik (Washington University in Saint Louis)
Explainability and Interpretability
🎯 What it does: This paper conducts a simulated hiring experiment on Amazon Mechanical Turk to study employers (selectors) and workers' perceptions of fairness and emotions when algorithms use different features (controllable vs. uncontrollable, directly vs. implicitly relevant).
The Orthogonality of Weight Vectors: The Key Characteristics of Normalization and Residual Connections
Zhixing Lu (Dalian University of Technology), Hongfei Lin (Dalian University of Technology)
Representation LearningConvolutional Neural NetworkTransformerImageText
🎯 What it does: Studied how combining normalization and residual connections in deep neural networks enhances the orthogonality of weight vectors, thereby improving feature learning capabilities.
The Transformation Logics
Alessandro Ronca (University of Oxford)
🎯 What it does: Propose a new temporal logic framework called Transformation Logics, which utilizes transformation operators based on semigroups/groups to unify and refine the expressive power of regular languages;
The Trembling-Hand Problem for LTLf Planning
Pian Yu (University of Oxford), Moshe Vardi (Rice University)
OptimizationRobotic IntelligenceReinforcement LearningTabular
🎯 What it does: This paper addresses the errors caused by agents making 'trembling-hand' mistakes during the execution of temporally extended goals (LTLf), studying how to generate strategies that maximize the probability of goal satisfaction in deterministic and non-deterministic planning domains.
Theoretical Study on Multi-objective Heuristic Search
Shawn Skyler (Ben Gurion University Of Negev), Carlos Hernandez Ulloa (Universidad San Sebastián)
OptimizationGraphBenchmark
🎯 What it does: This paper systematically studies multi-objective heuristic search (MOSA*) from a theoretical perspective, proposing to classify states into mandatory expansion, expandable, and non-expandable, and migrate this classification to search tree nodes; further defining a general framework MOSA*, analyzing the impact of node ordering functions (OF) and tie-breaking strategies (TB) on search, and experimentally comparing the performance of different OFs.
TIM: An Efficient Temporal Interaction Module for Spiking Transformer
Sicheng Shen (Chinese Academy of Sciences), Yi Zeng (Chinese Academy of Sciences)
ClassificationRecognitionSpiking Neural NetworkTransformerVideoTime Series
🎯 What it does: Proposes a plug-and-play time interaction module (TIM) to enhance the temporal information utilization efficiency of spiking transformers when processing time series data.
To Promote Full Cooperation in Social Dilemmas, Agents Need to Unlearn Loyalty
Chin-wing Leung (University of Warwick), Paolo Turrini (University of Warwick)
Reinforcement LearningGraph
🎯 What it does: This paper explores the evolution of cooperative behavior under different time scales by enabling agents to self-learn partner selection rules in a networked social dilemma game.
Tolerating Outliers: Gradient-Based Penalties for Byzantine Robustness and Inclusion
Latifa Errami (Mohammed VI Polytechnic University), El Houcine Bergou (Mohammed VI Polytechnic University)
OptimizationFederated LearningImage
🎯 What it does: Under the Byzantine attack scenario in federated/distributed SGD, we propose an aggregation method called F·LS based on linear scalarization, which maintains robustness against Byzantine adversaries while accommodating normal but outlier client updates.
Toward a Manifold-Preserving Temporal Graph Network in Hyperbolic Space
Viet Quan Le (VNU University of Engineering and Technology), Viet Cuong Ta (VNU University of Engineering and Technology)
Representation LearningGraph Neural NetworkGraphTime Series
🎯 What it does: Proposed a time series graph network (HMPTGN) that directly operates on a hypersphere, learning spatial and temporal relationships in dynamic graphs by preserving the hypersurface structure.
Toward Completing the Picture of Control in Schulze and Ranked Pairs Elections
Cynthia Maushagen (Heinrich-Heine-Universitat Dsseldorf), Tessa Seeger (Heinrich-Heine-Universitat Dsseldorf)
🎯 What it does: This paper systematically studies the computational complexity of Schulze and Ranked-Pairs voting rules under various election controls (deletion, replacement, precise multi-mode control, etc.), correcting flaws in previous proofs and extending to the unique winner model.
Towards a Framework for Learning of Algorithms: The Case of Learned Comparison Sorts
Philipp Kunz (University of Stuttgart), Marco Aiello (University of Stuttgart)
ClassificationOptimizationRepresentation LearningHyperparameter SearchTabular
🎯 What it does: Proposed a generic algorithm learning framework that models algorithms as learnable transfer operators using computational graphs and isomorphic algebra, with comparison sorting algorithms as a case study for learning and evaluation.
Towards a Pretrained Model for Restless Bandits via Multi-arm Generalization
Yunfan Zhao (Harvard University), Milind Tambe (Harvard University)
Reinforcement Learning
🎯 What it does: This paper develops a pre-trained RMAB model called PreFeRMAB, which supports multi-arm generalization, streaming arms joining/leaving, and nonlinear rewards in continuous states.
Towards a Principle-based Framework for Assessing the Contribution of Formulas on the Conflicts of Knowledge Bases
Badran Raddaoui (SAMOVAR, T'el'ecom SudParis, Institut Polytechnique de Paris), Said Jabbour (CRIL - University of Artois & CNRS)
Review/Survey Paper
🎯 What it does: This paper studies and proposes various local inconsistency measures based on knowledge base conflicts, establishing a complete evaluation framework.
Towards Automatic Composition of ASP Programs from Natural Language Specifications
Manuel Borroto Santana (University of Calabria), Francesco Ricca (University of Calabria)
AI Code AssistantTransformerText
🎯 What it does: This paper proposes a dataset named NL2CNL and a tool called NL2ASP, implementing a two-step process for automatically generating ASP programs from natural language descriptions;
Towards Counterfactual Fairness-aware Domain Generalization in Changing Environments
Yujie Lin (Tianjin University), Haifeng Chen (NEC Labs America)
Domain AdaptationExplainability and InterpretabilityRecurrent Neural NetworkAuto EncoderTabular
🎯 What it does: This paper proposes the DCFDG framework, achieving dual objectives of domain adaptation and causal fairness through causal decomposition and variational autoencoders.
Towards Dynamic Trend Filtering through Trend Point Detection with Reinforcement Learning
Jihyeon Seong (Korea Advanced Institute of Science and Technology), Jaesik Choi (Korea Advanced Institute of Science and Technology)
Reinforcement LearningTime SeriesFinance Related
🎯 What it does: Propose a dynamic trend filtering network called DTF-net based on reinforcement learning, which models trend point detection as a Markov Decision Process (MDP) and learns dynamic trend points (DTP) in a discrete action space, achieving more accurate trend interpolation.
Towards Dynamic-Prompting Collaboration for Source-Free Domain Adaptation
Mengmeng Zhan (University of Electronic Science and Technology of China), Xiaofeng Zhu (University of Electronic Science and Technology of China)
ClassificationDomain AdaptationTransformerPrompt EngineeringVision Language ModelImageBenchmark
🎯 What it does: Propose a dynamic prompt collaboration framework that combines a source pre-trained model with a large-scale vision-language model to address source-agnostic domain adaptation.
Towards Exact Computation of Inductive Bias
Akhilan Boopathy (Massachusetts Institute of Technology), Ila Fiete (Massachusetts Institute of Technology)
ImageBenchmark
🎯 What it does: Propose a practical method to directly estimate the inductive bias required for a given task: sample random hypotheses from the hypothesis space, estimate their error distribution on the test set, and calculate the negative logarithm of the probability of achieving a specified error threshold.
Towards Generalizable Neural Solvers for Vehicle Routing Problems via Ensemble with Transferrable Local Policy
Chengrui Gao (Nanjing University), Chao Qian (Nanjing University)
OptimizationTransformerReinforcement LearningGraphBenchmark
🎯 What it does: Propose a neural solver that integrates global and local strategies for the vehicle routing problem.
Towards Geometric Normalization Techniques in SE(3) Equivariant Graph Neural Networks for Physical Dynamics Simulations
Ziqiao Meng (Chinese University of Hong Kong), Irwin King (Chinese University of Hong Kong)
OptimizationGraph Neural NetworkGraphPhysics Related
🎯 What it does: Proposed a normalization layer called GEONORM that maintains SE(3) equivariance on large-scale particle systems and demonstrated its ability to stabilize the training of EGNN;
Towards Robust Multi-Label Learning against Dirty Label Noise
Yuhai Zhao (Northeastern University), Xingwei Wang (Northeastern University)
ClassificationData-Centric Learning
🎯 What it does: Propose a multi-label learning framework named NML for scenarios with mixed noise (Gaussian, sparse, subjective), and present the NMLD algorithm.
Towards Robust Trajectory Representations: Isolating Environmental Confounders with Causal Learning
Kang Luo (Hong Kong University of Science and Technology (Guangzhou)), Yuxuan Liang (Hong Kong University of Science and Technology (Guangzhou))
Explainability and InterpretabilityRepresentation LearningRecurrent Neural NetworkTransformerTabularSequential
🎯 What it does: By constructing a causal perspective trajectory representation framework (TrajCL), utilizing backdoor adjustment and environmental alignment modules to learn robust trajectory features that are not affected by geographical confounding.
Towards Sharper Generalization Bounds for Adversarial Contrastive Learning
Wen Wen (Huazhong Agricultural University), Hong Chen (Huazhong Agricultural University)
Representation LearningContrastive LearningImageTabular
🎯 What it does: This paper proposes and proves new high-probability generalization error upper bounds for unsupervised adversarial contrastive learning (ACL).
Towards Sharper Risk Bounds for Minimax Problems
Bowei Zhu (Renmin University of China), Yong Liu (Renmin University of China)
Optimization
🎯 What it does: Studied the generalization error and overfitting risk in non-convex-strongly concave min-max problems, providing tighter high-probability upper bounds.
Trade When Opportunity Comes: Price Movement Forecasting via Locality-Aware Attention and Iterative Refinement Labeling
Liang Zeng (Tsinghua University), Jian Li (Tsinghua University)
TabularTime SeriesFinance Related
🎯 What it does: This paper proposes the LARA framework for price trend prediction, primarily through local attention (LA-Attention) to automatically extract potential profitable samples and iterative label refinement (RA-Labeling) to correct noisy labels, ultimately improving prediction accuracy and actual returns.
Trusted Multi-view Learning with Label Noise
Cai Xu (Xidian University), Wei Zhao (Xidian University)
ClassificationRepresentation LearningImageTextTabular
🎯 What it does: Propose the Trusted Multi-view Noise Refining (TMNR) method, which utilizes evidence theory to generate beliefs and uncertainties at each view, designs view-specific noise association matrices, transforms original opinions into 'noise opinions' consistent with noisy labels, aggregates multi-view opinions via Dempster's rule, and trains the model using generalized maximum likelihood loss. Additionally, uncertainty-guided regularization and view consistency constraints are employed to enhance robustness against label noise.
Truth Table Net: Scalable, Compact & Verifiable Neural Networks with a Dual Convolutional Small Boolean Circuit Networks Form
Adrien Benamira (Nanyang Technological University), Bryan Hooi (National University of Singapore)
ClassificationExplainability and InterpretabilityComputational EfficiencyConvolutional Neural NetworkImageTabular
🎯 What it does: Designed a differentiable Truth Table Net (TTnet) that combines trainable convolutional networks with a form directly convertible to CNF logic gate circuits through learning a Truth Table (LTT) filter.
Truthful Interval Covering
Argyrios Deligkas (Royal Holloway University of London), Alexandros A. Voudouris (University of Essex)
Optimization
🎯 What it does: In this paper, the authors propose and study a new mechanism design problem called Truthful Interval Covering (TIC), which involves selecting covering intervals for a group of agents without using money to minimize the agents' costs;
TSESNet: Temporal-Spatial Enhanced Breast Tumor Segmentation in DCE-MRI Using Feature Perception and Separability
Jiezhou He (Xiamen University), Guojun Zhang (Xiamen University)
SegmentationConvolutional Neural NetworkContrastive LearningBiomedical DataMagnetic Resonance Imaging
🎯 What it does: Proposed an end-to-end Temporal-Spatial Enhanced Network (TSESNet) for precise segmentation of breast tumors in DCE-MRI sequences.
Two-stage Semi-supervised Speaker Recognition with Gated Label Learning
Xingmei Wang (Harbin Engineering University), Jinghan Liu (Harbin Engineering University)
RecognitionContrastive LearningAudio
🎯 What it does: Proposes a two-stage semi-supervised speaker recognition framework: first pre-trains using contrastive learning, then generates pseudo-labels through clustering and introduces gated label learning (GLL) for iterative semi-supervised training.
Unbiased Active Semi-supervised Binary Classification Models
JooChul Lee (Auburn University), Ziyang Wang (University of Connecticut)
ClassificationTabular
🎯 What it does: This paper proposes an unbiased estimator (AI-AEE) that combines semi-supervised inference in an active learning environment, using an imputation model that leverages both labeled and unlabeled data to estimate parameters of a binary classification model.
Unified Evidence Enhancement Inference Framework for Fake News Detection
Lianwei Wu (Northwestern Polytechnical University), Yongqiang Zhao (Peking University)
ClassificationExplainability and InterpretabilityGraph Neural NetworkTransformerTextRetrieval-Augmented Generation
🎯 What it does: Proposes a Unified Evidence-Enhanced Reasoning Framework (UEEI), achieving interpretable fake news detection through hierarchical conflict discovery, external evidence enhancement, and multi-perspective consistent reasoning.
Unified Physical-Digital Face Attack Detection
Hao Fang (Institute of Automation of Chinese Academy of Sciences), Zhen Lei (Institute of Automation of Chinese Academy of Sciences)
Anomaly DetectionTransformerPrompt EngineeringVision Language ModelVideoMultimodalityBenchmark
🎯 What it does: Developed a unified framework for physical and digital facial attack detection (UniAttackDetection) and released the first unified attack dataset (UniAttackData) that ensures ID consistency.
Unified Single-Stage Transformer Network for Efficient RGB-T Tracking
Jianqiang Xia (Intelligent Game and Decision Lab), Jian Zhao (China Telecom)
Object TrackingTransformerMultimodality
🎯 What it does: Designed a unified single-stage Transformer network called USTrack, integrating feature extraction, fusion, and relationship modeling for RGB-T tracking.
Unified Unsupervised Salient Object Detection via Knowledge Transfer
Yao Yuan (Nanjing University of Aeronautics and Astronautics), Jie Qin (Nanjing University of Aeronautics and Astronautics)
Object DetectionDomain AdaptationKnowledge DistillationConvolutional Neural NetworkImageVideo
🎯 What it does: Proposes a unified unsupervised salient object detection framework capable of migrating from natural static image (NSI) tasks to non-NSI tasks such as video SOD and remote sensing image SOD.
Unified View Imputation and Feature Selection Learning for Incomplete Multi-view Data
Yanyong Huang (Southwestern University of Finance and Economics), Fengmao Lv (Southwest Jiaotong University)
OptimizationRepresentation LearningData-Centric LearningGraph Neural NetworkMultimodality
🎯 What it does: Propose a unified learning framework called UNIFIER, which integrates missing view completion with unsupervised multi-view feature selection. It leverages dual local structures in both sample and feature spaces for adaptive graph learning and dynamically evaluates sample quality through semi-quadratic minimization.
UniM-OV3D: Uni-Modality Open-Vocabulary 3D Scene Understanding with Fine-Grained Feature Representation
Qingdong He (YouTu Lab, Tencent), Yunsheng Wu (YouTu Lab, Tencent)
SegmentationRepresentation LearningTransformerContrastive LearningImageTextMultimodalityPoint Cloud
🎯 What it does: This paper proposes a unified multimodal 3D scene understanding network called UniM-OV3D, which can recognize any category in open-vocabulary scenarios.
Unlearning during Learning: An Efficient Federated Machine Unlearning Method
Hanlin Gu (Webank), Qiang Yang (Webank)
Federated LearningImage
🎯 What it does: Propose a lightweight federated machine forgetting framework named FedAU, which can perform multi-client, multi-granularity (sample, class, client) forgetting operations simultaneously during the federated learning process.
Unlearning from Weakly Supervised Learning
Yi Tang (Southeast University), Min-Ling Zhang (Southeast University)
Safty and PrivacyImage
🎯 What it does: This paper proposes a machine forgetting method called UDRU for weakly supervised learning, achieving data forgetting by constructing a unified distribution target.
Unsupervised Anomaly Detection via Masked Diffusion Posterior Sampling
Di Wu (University of Electronic Science and Technology of China), Baihong Lin (University of Electronic Science and Technology of China)
Anomaly DetectionComputational EfficiencyDiffusion modelImage
🎯 What it does: Proposed an unsupervised anomaly detection method based on diffusion posterior sampling (MDPS)
Unsupervised Deep Graph Structure and Embedding Learning
Xiaobo Shen (Nanjing University of Science and Technology), Shirui Pan (Griffith University)
Computational EfficiencyRepresentation LearningAdversarial AttackGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposes an unsupervised graph structure learning method UPGNN (and its accelerated version AUPGNN), which reconstructs a clean graph by leveraging properties such as low rank, sparsity, and feature smoothing, and learns graph embeddings through graph mutual information maximization to enhance robustness under adversarial attacks.
Updates on the Complexity of SHAP Scores
Xuanxiang Huang (CNRS@CREATE), Joao Marques-Silva (ICREA, University of Lleida)
ClassificationExplainability and Interpretability
🎯 What it does: This paper systematically investigates the computational complexity of SHAP scores. First, it proves that computing SHAP scores for classifiers using voting mechanisms in random forests (RF) and tree ensembles (TE) is #P-hard; subsequently, it extends the polynomial-time computation methods for d-DNNF and read-once multi-valued decision trees under the knowledge compilation framework.
Using Large Language Models to Improve Query-based Constraint Acquisition
Younes Mechqrane (International Artificial Intelligence Center of Morocco University Mohammed VI Polytechnic), Ismail Elabbassi (International Artificial Intelligence Center of Morocco University Mohammed VI Polytechnic)
OptimizationTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: Proposed an active constraint acquisition framework named ACQNOGOODS that does not require bias, and integrated large language models to realize a constraint acquisition system with natural language feedback called LLMACQ.
VCC-INFUSE: Towards Accurate and Efficient Selection of Unlabeled Examples in Semi-supervised Learning
Shijie Fang (Peking University), Tong Lin (Peking University)
Computational EfficiencyData-Centric LearningAuto EncoderImage
🎯 What it does: Propose the VCC-INFUSE method, which separately employs Variational Confidence Calibration (VCC) and Unlabeled Sample Pruning Based on Influence Functions (INFUSE) to improve the quality of pseudo-labels and training efficiency in semi-supervised learning.
VCformer: Variable Correlation Transformer with Inherent Lagged Correlation for Multivariate Time Series Forecasting
Yingnan Yang (Shenzhen University), Jianyong Chen (Shenzhen University)
TransformerTime Series
🎯 What it does: Propose the VCformer model, combining Variable Correlation Attention (VCA) and Koopman Temporal Detector (KTD) to achieve multivariate time series forecasting.
Vertical Symbolic Regression via Deep Policy Gradient
Nan Jiang (Purdue University), Yexiang Xue (Purdue University)
OptimizationComputational EfficiencyAI Code AssistantRecurrent Neural NetworkReinforcement LearningTabularPhysics RelatedOrdinary Differential Equation
🎯 What it does: Propose a framework that combines vertical symbolic regression with deep policy gradients, capable of incrementally constructing symbolic equations in multivariate scenarios.
VF-Detector: Making Multi-Granularity Code Changes on Vulnerability Fix Detector Robust to Mislabeled Changes
Zhenkan Fu (Dalian Maritime University), He Jiang (Dalian University of Technology)
ClassificationRepresentation LearningData-Centric LearningConvolutional Neural NetworkRecurrent Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningTextSequential
🎯 What it does: This paper proposes VF-Detector, which automatically identifies vulnerability fixes in software repair commits using a framework that combines multi-granularity code embeddings with confidence learning for noise removal.
Vision-based Discovery of Nonlinear Dynamics for 3D Moving Target
Zitong Zhang (Renmin University of China), Hao Sun (Renmin University of China)
Object TrackingConvolutional Neural NetworkVideoPhysics RelatedOrdinary Differential Equation
🎯 What it does: Using videos captured by three cameras, combined with target tracking, coordinate transformation based on Rodrigues' rotation formula, and sparse regression based on spline functions, automatically infers the nonlinear dynamic equations of three-dimensional moving objects from raw videos.
Vision-fused Attack: Advancing Aggressive and Stealthy Adversarial Text against Neural Machine Translation
Yanni Xue (Beihang University), Xianglong Liu (Beihang University)
Adversarial AttackTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodality
🎯 What it does: Propose the Vision-fused Attack (VFA) framework, which leverages visual information and semantic space to generate more aggressive and harder-to-be-noticed-by-humans adversarial text for attacking neural machine translation models.
Visual Attention Prompted Prediction and Learning
Yifei Zhang (Emory University), Liang Zhao (Emory University)
ClassificationObject DetectionExplainability and InterpretabilityConvolutional Neural NetworkPrompt EngineeringImageBiomedical Data
🎯 What it does: Proposes a visual attention prompting prediction and learning framework that utilizes visual prompts to guide model inference while considering unprompted samples.
vMFER: Von Mises-Fisher Experience Resampling Based on Uncertainty of Gradient Directions for Policy Improvement
Yiwen Zhu (Zhejiang University), Changjie Fan (NetEase Fuxi AI Lab)
Reinforcement Learning
🎯 What it does: Proposed an empirical experience resampling method called vMFER based on gradient direction uncertainty to improve the policy optimization process under multi-Critic architectures.
VSGT: Variational Spatial and Gaussian Temporal Graph Models for EEG-based Emotion Recognition
Chenyu Liu (Nanyang Technological University), Yang Liu (Nanyang Technological University)
RecognitionGraph Neural NetworkAuto EncoderBiomedical Data
🎯 What it does: For EEG emotion recognition, the VSGT model is proposed, capturing spatial and cross-temporal dependencies through variational spatial encoder and Gaussian temporal encoder.
Vulnerabilities of Single-Round Incentive Compatibility in Auto-bidding: Theory and Evidence from ROI-Constrained Online Advertising Markets
Juncheng Li (Tsinghua University), Pingzhong Tang (Tsinghua University)
TabularFinance Related
🎯 What it does: This paper studies automatic bidding markets under ROI constraints, analyzes the vulnerability of single-round incentive compatibility (IC) in second-price auctions, and proves various undesirable properties and computational difficulties in this environment.
WeatherGNN: Exploiting Meteo- and Spatial-Dependencies for Local Numerical Weather Prediction Bias-Correction
Binqing Wu (Alibaba Group), Ling Chen (Zhejiang University)
Graph Neural NetworkTabularTime Series
🎯 What it does: This paper proposes WeatherGNN, which addresses the problem of local numerical weather prediction bias correction by leveraging graph neural networks to learn meteorological factors and spatial dependencies, achieving adaptive regional bias correction.
Weighted EF1 and PO Allocations with Few Types of Agents or Chores
Jugal Garg (University of Illinois at Urbana-Champaign), John Qin (University of Illinois at Urbana-Champaign)
Optimization
🎯 What it does: Investigated the existence and polynomial-time algorithms for achieving weighted fairness (wEF1) and Pareto optimality (PO) when allocating indivisible chores among heterogeneous agents with unequal weights (or shares).
Welfare Loss in Connected Resource Allocation
Xiaohui Bei (Nanyang Technological University), Warut Suksompong (National University of Singapore)
OptimizationGraph
🎯 What it does: The study investigates the worst-case loss caused by connectivity constraints on equilibrium welfare and total utility when allocating indivisible items on undirected graphs, and introduces the concepts of equilibrium price connectivity and utility price connectivity.
What Hides behind Unfairness? Exploring Dynamics Fairness in Reinforcement Learning
Zhihong Deng (University of Technology Sydney), Chengqi Zhang (University of Technology Sydney)
Reinforcement Learning
🎯 What it does: This paper systematically studies the long-term fairness issue in reinforcement learning from a causal inference perspective, proposing a new fairness concept called dynamic fairness, and combining it with traditional direct and indirect effect decomposition to obtain an identifiable formula; subsequently, based on these theories, the model-based reinforcement learning algorithm InsightFair is constructed, which can actively assess and correct unfairness caused by environmental dynamics during the learning process.
What Is Best for Students, Numerical Scores or Letter Grades?
Evi Micha (University of Toronto), Nisarg Shah (University of Toronto)
Tabular
🎯 What it does: Investigated the (de)motivational effects of numerical ratings versus uniform letter ratings on students' subsequent exam performance, and proposed a theoretical model based on students' true ability, score noise, and motivation coefficients.
What Makes Models Compositional? A Theoretical View
Parikshit Ram (IBM Research), Alexander G. Gray (Centaur AI Institute)
Explainability and InterpretabilityRepresentation LearningConvolutional Neural NetworkRecurrent Neural NetworkTransformer
🎯 What it does: This paper theoretically analyzes the model's expressiveness and systematic generalization by proposing a unified neuro-symbolic definition, viewing sequence processing models as combinations of input encoding, computation DAG, span processor, and readout function.
When Fairness Meets Privacy: Exploring Privacy Threats in Fair Binary Classifiers via Membership Inference Attacks
Huan Tian (University of Technology Sydney), Wanlei Zhou (City University of Macau)
Safty and PrivacyAdversarial AttackImage
🎯 What it does: Investigate the privacy risks of fairness interventions in binary classification models, evaluate the effectiveness of existing membership inference attacks (MIA), and propose a new FD-MIA attack method that leverages prediction differences between fair models and biased models to enhance privacy leakage rates.
Where Elegance Meets Precision: Towards a Compact, Automatic, and Flexible Framework for Multi-modality Image Fusion and Applications
Jinyuan Liu (Dalian University of Technology), Xin Fan (Dalian University of Technology)
Object DetectionSegmentationOptimizationHyperparameter SearchConvolutional Neural NetworkMultimodality
🎯 What it does: Proposed the Compact, Automatic and Flexible (CAF) framework, which unifies infrared-visible image fusion with downstream perception tasks (detection, segmentation) into a single objective through bi-level optimization, achieving mutually beneficial fusion and task learning.
Where to Mask: Structure-Guided Masking for Graph Masked Autoencoders
Chuang Liu (Wuhan University), Wenbin Hu (Wuhan University)
Representation LearningGraph Neural NetworkAuto EncoderGraph
🎯 What it does: This paper proposes a graph structure-guided masking strategy called StructMAE to improve the pre-training of graph masked autoencoders.
Who Looks like Me: Semantic Routed Image Harmonization
Jinsheng Sun (University of Science and Technology Beijing), Xiaojuan Ban (University of Science and Technology Beijing)
Image HarmonizationTransformerImage
🎯 What it does: Propose an image harmony model based on instance similarity, utilizing the Instance Similarity Evaluation Module (ISEM) and Style Transfer Block (STB) to achieve semantic routing, enhancing consistency between foreground and background.
Why Only Text: Empowering Vision-and-Language Navigation with Multi-modal Prompts
Haodong Hong (University of Queensland), Jiajun Liu (CSIRO Data61)
Object DetectionData SynthesisAutonomous DrivingTransformerPrompt EngineeringVision Language ModelMultimodality
🎯 What it does: Propose Vision-and-Language Navigation with Multi-modal Prompts (VLN-MP), enhancing navigation performance by incorporating image prompts into traditional VLN instructions;
With a Little Help from Language: Semantic Enhanced Visual Prototype Framework for Few-Shot Learning
Hecheng Cai (Sichuan University), Jiancheng Lv (Sichuan University)
ClassificationRepresentation LearningMeta LearningConvolutional Neural NetworkContrastive LearningImageMultimodalityBenchmark
🎯 What it does: Propose a plug-and-play SEVpro framework that injects semantic knowledge into the visual feature extractor during the pre-training phase to improve prototype learning in few-shot learning.
WPML3CP: Wasserstein Partial Multi-Label Learning with Dual Label Correlation Perspectives
Ximing Li (Jilin University), Jihong Ouyang (Jilin University)
ClassificationImageTextBiomedical Data
🎯 What it does: Propose a dual label correlation perspective partial multi-label learning method WPML CP 3 based on Wasserstein distance, which jointly learns label confidence and prediction model, and improves model performance through label correlation regularization.
WSRFNet: Wavelet-Based Scale-Specific Recurrent Feedback Network for Diabetic Retinopathy Lesion Segmentation
Xuan Li (Harbin Institute of Technology), Xiangqian Wu (Harbin Institute of Technology)
SegmentationConvolutional Neural NetworkRecurrent Neural NetworkImageBiomedical Data
🎯 What it does: Proposed a wavelet-based scale-specific recursive feedback network, WSRFNet, to enhance the quality of multi-scale features for diabetic retinopathy lesion segmentation.
X-Light: Cross-City Traffic Signal Control Using Transformer on Transformer as Meta Multi-Agent Reinforcement Learner
Haoyuan Jiang (Baidu Inc), Rui Zhao (SenseTime Research)
Autonomous DrivingOptimizationMeta LearningTransformerReinforcement LearningTime SeriesSequential
🎯 What it does: This paper proposes a meta-multi-agent traffic signal control framework named X-Light based on Transformer-on-Transformer, aiming to achieve cross-city transferable traffic light optimization.
Zero-shot High-fidelity and Pose-controllable Character Animation
Bingwen Zhu (Fudan University), Yu-Gang Jiang (Fudan University)
GenerationTransformerVision Language ModelDiffusion modelImageVideoTextMultimodality
🎯 What it does: Utilizing pre-trained diffusion models and pose control, achieving high-fidelity, pose-controllable animation generation from a single person image without training.
Zero-shot Learning for Preclinical Drug Screening
Kun Li (Macquarie University), Wenbin Hu (Macquarie University)
Domain AdaptationDrug DiscoveryGraph Neural NetworkTransformerBiomedical Data
🎯 What it does: Proposed and implemented a zero-shot learning plugin called MSDA to enhance the prediction performance of drug response prediction models for unknown drugs during the preclinical drug screening phase.
Zero-Shot Sketch Based Image Retrieval via Modality Capacity Guidance
Yanghong Zhou (Hong Kong Polytechnic University), P. Y. Mok (Hong Kong Polytechnic University)
RetrievalTransformerVision Language ModelContrastive LearningImage
🎯 What it does: Proposed a loss function based on modal capacity constraints to guide feature learning for zero-shot sketch retrieval;
ZeroDDI: A Zero-Shot Drug-Drug Interaction Event Prediction Method with Semantic Enhanced Learning and Dual-modal Uniform Alignment
Ziyan Wang (Huazhong Agricultural University), Wen Zhang (Huazhong Agricultural University)
Drug DiscoveryGraph Neural NetworkTransformerLarge Language ModelContrastive LearningTextGraph
🎯 What it does: Propose the ZeroDDI method for predicting unknown drug-drug interaction events (DDIE) without labeled samples in the training set.
Zeta*-SIPP: Improved Time-Optimal Any-Angle Safe-Interval Path Planning
Yiyuan Zou (Delft University of Technology), Clark Borst (Delft University of Technology)
OptimizationBenchmark
🎯 What it does: In dynamic obstacle environments, the Zeta*-SIPP algorithm was proposed and implemented, combining arbitrary-angle forward expansion and field-of-view (FoV) visibility checks to significantly accelerate TO-AA-SIPP.