IJCAI 2023 Papers — Page 6
International Joint Conference on Artificial Intelligence · 639 papers
RFENet: Towards Reciprocal Feature Evolution for Glass Segmentation
Ke Fan (Shanghai Jiao Tong University), Lizhuang Ma (Shanghai Jiao Tong University)
SegmentationConvolutional Neural NetworkImage
🎯 What it does: Propose a recursive feature evolution network called RFENet, specifically designed for the segmentation of glass-like objects.
Robust Image Ordinal Regression with Controllable Image Generation
Yi Cheng (Zhejiang University), Jian Wu (Zhejiang University)
ClassificationGenerationData SynthesisConvolutional Neural NetworkTransformerAuto EncoderImage
🎯 What it does: In image ordinal regression tasks, the controllable image generation framework CIG is proposed, utilizing generated synthetic samples to address class imbalance and class overlap issues;
Robust Reinforcement Learning via Progressive Task Sequence
Yike Li (Beijing Jiaotong University), Jiqiang Liu (Beijing Jiaotong University)
Reinforcement LearningImageSequential
🎯 What it does: Proposed a robust reinforcement learning framework in the form of max-expectation, and designed the Dynamic Robust RL (DRRL) framework, which generates and sorts evolutionary task sequences through genetic algorithms to achieve dynamic multi-task learning, thereby enhancing robustness and training stability.
Robust Steganography without Embedding Based on Secure Container Synthesis and Iterative Message Recovery
Ziping Ma (Peking University), Hui Fang (Loughborough University)
Safty and PrivacyGenerative Adversarial NetworkImage
🎯 What it does: Designed a robust steganographic method SI-SWE based on secure container synthesis and iterative message recovery, using a generator to map a key to a container image and a differential predictor to iteratively recover the message after attacks.
RuleMatch: Matching Abstract Rules for Semi-supervised Learning of Human Standard Intelligence Tests
Yunlong Xu (Shanghai Jiao Tong University), Ru-Yuan Zhang (Shanghai Jiao Tong University)
ClassificationExplainability and InterpretabilityData-Centric LearningImageBenchmark
🎯 What it does: Propose the RuleMatch semi-supervised learning framework, which utilizes rule-level data augmentation and rule consistency loss to train deep models on the RPM problem.
Runtime Analyses of Multi-Objective Evolutionary Algorithms in the Presence of Noise
Matthieu Dinot (cole Polytechnique), Sebastian Will (Ecole Polytechnique)
OptimizationBenchmark
🎯 What it does: This paper conducts the first mathematical runtime analysis of multi-objective evolutionary algorithms (MOEA) when the objective function is affected by noise, particularly examining the performance of the simple evolutionary multi-objective optimizer (SEMO) on the classic benchmark problem ONEMINMAX.
RZCR: Zero-shot Character Recognition via Radical-based Reasoning
Xiaolei Diao (Jilin University), Hao Xu (Jilin University)
RecognitionRepresentation LearningImageGraph
🎯 What it does: Proposes the RZCR zero-shot character recognition framework, combining radical extraction and knowledge graph reasoning to enhance the recognition performance of few-shot characters.
SAD: Semi-Supervised Anomaly Detection on Dynamic Graphs
Sheng Tian (Ant Group), Liang Chen (Sun Yat-sen University)
Anomaly DetectionGraph Neural NetworkContrastive LearningGraphFinance Related
🎯 What it does: Propose a semi-supervised anomaly detection framework named SAD for detecting node anomalies in dynamic graphs.
Safe Multi-agent Learning via Trapping Regions
Aleksander Czechowski (Delft University of Technology), Frans A. Oliehoek (Delft University of Technology)
Reinforcement Learning
🎯 What it does: This paper proposes utilizing the trapping region to provide safety guarantees for multi-agent learning and presents a verification algorithm based on binary partitioning and sampling.
Safe Reinforcement Learning via Probabilistic Logic Shields
Wen-Chi Yang (KU Leuven), Luc De Raedt (KU Leuven)
Convolutional Neural NetworkReinforcement Learning
🎯 What it does: Proposed a differentiable safe reinforcement learning method called Probabilistic Logic Policy Gradient (PLPG), achieving safety constraints in continuous action spaces by combining probabilistic logic programming with policy gradients.
Safety Verification and Universal Invariants for Relational Action Bases
Silvio Ghilardi (Universita degli Studi di Milano), Andrey Rivkin (Technical University of Denmark)
Safty and Privacy
🎯 What it does: Proposed a new Relational Action Basis (RAB) framework for modeling and verifying relational dynamic systems with arithmetic and universal quantification constraints.
Sample Efficient Model-free Reinforcement Learning from LTL Specifications with Optimality Guarantees
Daqian Shao (University of Oxford), Marta Kwiatkowska (University of Oxford)
OptimizationReinforcement Learning
🎯 What it does: Propose a model-free reinforcement learning framework that learns optimal policies on unknown Markov Decision Processes (MDPs) using Linear Temporal Logic (LTL) specifications, with theoretical optimality guarantees;
Sampling Ex-Post Group-Fair Rankings
Sruthi Gorantla (Indian Institute of Science), Anand Louis (Indian Institute of Science)
Recommendation SystemOptimizationTabular
🎯 What it does: Propose a random ranking distribution based on group fairness constraints and prove that this distribution is unique and supports ex-post fair ranking.
SAT-Based PAC Learning of Description Logic Concepts
Balder ten Cate (University of Amsterdam), Carsten Lutz (Leipzig University)
Benchmark
🎯 What it does: This paper proposes a description logic concept learning framework based on bounded fitting, implements a SAT-driven system called SPELL, and provides theoretical guarantees for PAC learning; meanwhile, it proves that traditional most specific/general fitting and refinement methods lack sample efficiency.
Scalable Communication for Multi-Agent Reinforcement Learning via Transformer-Based Email Mechanism
Xudong Guo (Tsinghua University), Wenhui Fan (Tsinghua University)
Recurrent Neural NetworkTransformerReinforcement Learning
🎯 What it does: Designed a Transformer-based Email Mechanism (TEM) to achieve local communication and message chaining forwarding, supporting scalable information sharing in multi-agent reinforcement learning.
Scalable Coupling of Deep Learning with Logical Reasoning
Marianne Defresne (Université Fédérale de Toulouse), Thomas Schiex (Université Fédérale de Toulouse)
OptimizationExplainability and InterpretabilityConvolutional Neural NetworkImageTabularBiomedical Data
🎯 What it does: Propose a hybrid architecture that combines any deep learning layer with a final discrete graphical model (pairwise GM), and design a differentiable E-NPLL loss function to learn how to solve NP-hard discrete reasoning or optimization problems from natural inputs.
Scalable Optimal Margin Distribution Machine
Yilin Wang (Huazhong University of Science and Technology), Hai Jin (Huazhong University of Science and Technology)
OptimizationComputational EfficiencyImageText
🎯 What it does: This paper proposes a scalable optimal margin distribution machine (SODM), which significantly improves training speed while maintaining generalization performance through a novel partitioning strategy and accelerated SVRG under a linear kernel.
Scalable Verification of Strategy Logic through Three-Valued Abstraction
Francesco Belardinelli (Imperial College London), Aniello Murano (University of Naples Federico II)
Graph
🎯 What it does: Propose a three-valued semantics for strategy logic, design a three-valued abstraction method based on this semantics, and implement and verify it within the MCMAS framework.
Scaling Goal-based Exploration via Pruning Proto-goals
Akhil Bagaria (Brown University), Tom Schaul (DeepMind)
Reinforcement LearningText
🎯 What it does: Propose a goal-driven reinforcement learning framework based on the 'proto-goal' space and Proto-Goal Evaluator (PGE), which automatically refines a large but meaningful proto-goal space into achievable, controllable, reachable, and reward-related goals. These goals are embedded into a goal-conditioned RL agent implemented with distributed R2D2 to address the exploration challenge in large-scale sparse reward environments.
Schelling Games with Continuous Types
Davide Bilò (University of L'Aquila), Jonas Schmidt (Hasso Plattner Institute, University of Potsdam)
Graph
🎯 What it does: The study is based on the continuous-type Schelling residential segregation game, proposing three cost models: average distance, maximum distance, and truncated threshold, and analyzing their equilibrium, price disorder, and price stability.
ScriptWorld: Text Based Environment for Learning Procedural Knowledge
Abhinav Joshi (Indian Institute of Technology Kanpur), Ashutosh Modi (Indian Institute of Technology Kanpur)
TransformerLarge Language ModelReinforcement LearningText
🎯 What it does: Developed the ScriptWorld text-based game environment and trained RL agents based on script knowledge within it.
Self-Recover: Forecasting Block Maxima in Time Series from Predictors with Disparate Temporal Coverage Using Self-Supervised Learning
Asadullah Hill Galib (Michigan State University), Lifeng Luo (Michigan State University)
Recurrent Neural NetworkAuto EncoderContrastive LearningTime Series
🎯 What it does: Propose the Self-Recover framework to predict time series block maxima by integrating self-supervised learning and residual learning with historical observations and process model predictions.
Self-supervised Graph Disentangled Networks for Review-based Recommendation
Yuyang Ren (Shanghai Jiao Tong University), Chenghu Zhou (Institute Of Geographical Sciences And Natural Resources Research Chinese Academy Of Sciences)
Recommendation SystemGraph Neural NetworkContrastive LearningText
🎯 What it does: Proposed a self-supervised graph decomposition network (SGDN) to decouple the latent intentions of user–item interactions in comment-based recommendations.
Self-Supervised Neuron Segmentation with Multi-Agent Reinforcement Learning
Yinda Chen (University of Science and Technology of China), Zhiwei Xiong (University of Science and Technology of China)
SegmentationTransformerReinforcement LearningAuto EncoderBiomedical Data
🎯 What it does: Designed and implemented a self-supervised masked image model based on multi-agent reinforcement learning (Decision-based MIM) to automatically identify optimal mask ratios and strategies. During pre-training, a Vision Transformer was trained with HOG and MSE reconstruction objectives, and a UNETR decoder was added for neuron instance segmentation.
Semantic-Aware Generation of Multi-View Portrait Drawings
Biao Ma (Hangzhou Dianzi University), Gang Xu (Hangzhou Dianzi University)
GenerationNeural Radiance FieldGenerative Adversarial NetworkImage
🎯 What it does: Designed a semantic-aware generative model SAGE for generating multi-view human face sketches or paintings;
Semi-supervised Domain Adaptation in Graph Transfer Learning
Ziyue Qiao (Jiangmen Laboratory of Carbon Science and Technology), Hui Xiong (Hong Kong University of Science and Technology)
ClassificationDomain AdaptationGraph Neural NetworkGenerative Adversarial NetworkTextGraph
🎯 What it does: This paper proposes a semi-supervised domain adaptation method called SGDA for graph transfer learning scenarios, which transfers labeled source graph knowledge to an unlabeled target graph to achieve node classification;
Semi-supervised Domain Adaptation via Prototype-based Multi-level Learning
Xinyang Huang (Beijing University of Posts and Telecommunications), Wenkai Chen (Beijing University of Posts and Telecommunications)
ClassificationDomain AdaptationConvolutional Neural NetworkImageBenchmark
🎯 What it does: Propose the Prototype-based Multi-level Learning (ProML) framework to fully utilize the limited target label samples in semi-supervised domain adaptation, enhancing model transfer performance through prototype aggregation, cross-domain alignment, and batch-level dual consistency.
SemiGNN-PPI: Self-Ensembling Multi-Graph Neural Network for Efficient and Generalizable Protein–Protein Interaction Prediction
Ziyuan Zhao (Institute for Infocomm Research), Xiaoli Li (Institute for Infocomm Research)
Drug DiscoveryGraph Neural NetworkGraphBiomedical Data
🎯 What it does: Propose a self-supervised multi-graph neural network, SemiGNN-PPI, for efficient and generalizable prediction of multi-type protein-protein interactions (PPI).
Sequence Learning Using Equilibrium Propagation
Malyaban Bal (Pennsylvania State University), Abhronil Sengupta (Pennsylvania State University)
ClassificationRecurrent Neural NetworkTextSequential
🎯 What it does: Combine Equilibrium Propagation with modern Hopfield networks to construct a convergent RNN model applicable to sequence classification.
Sequential Attention Source Identification Based on Feature Representation
Dongpeng Hou (Northwestern Polytechnical University), Xuelong Li (Northwestern Polytechnical University)
Anomaly DetectionRepresentation LearningRecurrent Neural NetworkGraph Neural NetworkGraphTime Series
🎯 What it does: Study the problem of multi-source rumor source identification, proposing a sequence-to-sequence framework called TGASI based on time series attention and graph attention, which uses infection snapshot sequences to locate the propagation source.
Sequential Recommendation with Probabilistic Logical Reasoning
Huanhuan Yuan (Soochow University), Lei Zhao (Soochow University)
Recommendation SystemRecurrent Neural NetworkTransformerSequential
🎯 What it does: Integrate deep learning with symbolic learning, proposing the SR-PLR framework, which employs Beta distribution for probabilistic logical reasoning in sequence recommendation, and generates recommendations by concatenating logical embeddings with feature embeddings.
SeRO: Self-Supervised Reinforcement Learning for Recovery from Out-of-Distribution Situations
Chan Kim (Seoul National University), Seong-Woo Kim (Seoul National University)
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed a self-supervised reinforcement learning method called SeRO to restore reliable robot behaviors in discrete distribution (OOD) states.
SF-PATE: Scalable, Fair, and Private Aggregation of Teacher Ensembles
Cuong Tran (Syracuse University), Pascal Van Hentenryck (Georgia Institute of Technology)
Federated LearningSafty and PrivacyMixture of ExpertsImageTabular
🎯 What it does: This paper proposes the SF-PATE framework, achieving simultaneous fairness and privacy through differential privacy voting on teacher ensemble models;
SGAT4PASS: Spherical Geometry-Aware Transformer for PAnoramic Semantic Segmentation
Xuewei Li (Zhejiang University), Xi Li (Zhejiang University)
SegmentationTransformerImage
🎯 What it does: Propose a spherical geometry-aware Transformer (SGAT4PASS) for panoramic semantic segmentation, enhancing robustness against 3D perturbations through three modules.
Shaken, and Stirred: Long-Range Dependencies Enable Robust Outlier Detection with PixelCNN++
Barath Mohan Umapathi (Indian Institute of Science), Devarajan Sridharan (Indian Institute of Science)
Anomaly DetectionConvolutional Neural NetworkImage
🎯 What it does: Analyze the likelihood bias in the PixelCNN++ generative model and propose two lightweight bijective transformations (Stirring and Shaking) to correct the bias caused by low-order dependencies, achieving robust unsupervised anomaly detection.
Shhh! The Logic of Clandestine Operations
Pavel Naumov (University of Southampton), Oliver Orejola (Tulane University)
🎯 What it does: Proposed formal semantics and a logical framework for clandestine actions, along with the corresponding logical system;
Simplification and Improvement of MMS Approximation
Hannaneh Akrami (Max Planck Institute for Informatics), Setareh Taki (Grubhub)
Optimization
🎯 What it does: This paper studies the problem of fairly allocating indivisible items among n agents with additive valuations using the fairness concept of maximin share (MMS). Since MMS allocations do not always exist, many studies have provided existence and algorithmic guarantees for approximate MMS allocations. This paper simplifies the analysis of the Garg-Taki algorithm and improves the existence guarantee to a factor of (3/4 + min(1/36, 3/16 n - 4)).
Simulation-Assisted Optimization for Large-Scale Evacuation Planning with Congestion-Dependent Delays
Kazi Ashik Islam (University of Virginia), Anil Vullikanti (University of Virginia)
OptimizationGraphTabular
🎯 What it does: This paper proposes an scalable hybrid integer programming and large neighborhood search (MIP-LNS) method for jointly optimizing evacuation routing and scheduling, and introduces a simulation-assisted (MIP-LNS-SIM) approach to model delays caused by congestion, aiming to minimize metrics such as average evacuation time and evacuation completion time.
Singularformer: Learning to Decompose Self-Attention to Linearize the Complexity of Transformer
Yifan Wu (Central South University), Min Li (Central South University)
Computational EfficiencyRepresentation LearningTransformerImageTextBiomedical Data
🎯 What it does: Propose Singularformer, which utilizes a neural network to learn the singular value decomposition (SVD) of the attention matrix, achieving linear complexity and low memory consumption in self-attention;
Sketch Recognition via Part-based Hierarchical Analogical Learning
Kezhen Chen (Northwestern University), Madeline Usher (Northwestern University)
RecognitionImage
🎯 What it does: Propose a sketch recognition method based on part hierarchical analogy learning (PHAL);
SLViT: Scale-Wise Language-Guided Vision Transformer for Referring Image Segmentation
Shuyi Ouyang (Zhejiang University), Lanfen Lin (Zhejiang University)
SegmentationTransformerLarge Language ModelVision Language ModelMultimodality
🎯 What it does: This paper proposes a Transformer-based referential image segmentation framework named SLViT, integrating visual-language encoders and cross-scale enhancement modules to precisely segment targets in images based on language expressions.
SmartBERT: A Promotion of Dynamic Early Exiting Mechanism for Accelerating BERT Inference
Boren Hu (Zhejiang University), Siliang Tang (Zhejiang University)
Computational EfficiencyTransformerContrastive LearningTextBenchmark
🎯 What it does: Propose SmartBERT by integrating dynamic early stopping and layer skipping mechanisms in BERT inference.
SMARTformer: Semi-Autoregressive Transformer with Efficient Integrated Window Attention for Long Time Series Forecasting
Yiduo Li (Harbin Institute of Technology Shenzhen), Zenglin Xu (Harbin Institute of Technology Shenzhen)
Computational EfficiencyRepresentation LearningTransformerTime Series
🎯 What it does: Proposed the SMARTformer model, introducing time-independent embedding, integrated window attention, and semi-autoregressive decoder to enhance the accuracy and efficiency of long-term time series forecasting.
Social Motivation for Modelling Other Agents under Partial Observability in Decentralised Training
Dung Nguyen (Deakin University), Truyen Tran (Deakin University)
Reinforcement Learning
🎯 What it does: In decentralized training within partially observable multi-agent environments, a social motivation reward is proposed to encourage agents to proactively approach other agents to enhance the quality of their behavioral models.
Solving Quantum-Inspired Perfect Matching Problems via Tutte-Theorem-Based Hybrid Boolean Constraints
Moshe Y. Vardi (Rice University), Zhiwei Zhang (Rice University)
OptimizationGraphBenchmark
🎯 What it does: Studied a hybrid Boolean constraint encoding method based on Tutte's theorem for solving quantum-inspired perfect matching problems.
Solving the Identifying Code Set Problem with Grouped Independent Support
Anna L.D. Latour (National University of Singapore), Kuldeep S. Meel (National University of Singapore)
OptimizationGraph
🎯 What it does: Studied the generalized identifying code set (GICS) problem for sensor placement in networks, and proposed a method to reduce it to the independent support problem of Boolean formulas through grouped independent support (GIS), thereby solving the minimal sensor set.
Some General Identification Results for Linear Latent Hierarchical Causal Structure
Zhengming Chen (Guangdong University of Technology), Ruichu Cai (Guangdong University of Technology)
🎯 What it does: This paper proposes a method for identifying latent variable hierarchies using second-order statistics and GIN conditions in a linear latent hierarchical causal model that does not require tree or triangular structure constraints and allows partial Gaussian noise, along with the corresponding algorithm.
Some Might Say All You Need Is Sum
Eran Rosenbluth (RWTH Aachen University), Martin Grohe (RWTH Aachen University)
Representation LearningGraph Neural NetworkGraph
🎯 What it does: This paper systematically studies the expressiveness of different aggregation functions (Sum, Mean, Max) in graph neural networks (GNNs), proposes a unified expressiveness framework, and proves that Sum aggregation cannot universally replace Mean or Max in various scenarios; however, Sum can approximate them when inputs are restricted to finite domains.
Sorting and Hypergraph Orientation under Uncertainty with Predictions
Thomas Erlebach (Durham University), Jens Schlöter (University of Bremen)
Optimization
🎯 What it does: This paper proposes a learning-augmented algorithm under explorable uncertainty, addressing sorting and hypergraph orientation problems. The algorithm provides near-zero query cost when predictions are accurate, while maintaining optimal worst-case query competitiveness when predictions are incorrect.
Spatial-Temporal Self-Attention for Asynchronous Spiking Neural Networks
Yuchen Wang (University of Electronic Science and Technology of China), Hong Qu (University of Electronic Science and Technology of China)
ClassificationRecognitionSpiking Neural NetworkTransformerImageAudio
🎯 What it does: Proposed a spatiotemporal self-attention mechanism (STSA) that maintains asynchronous characteristics and a relative position information bias (STRPB), constructing an STS-Transformer model based on these modules for event-driven spiking neural networks (SNNs);
Spatially Constrained Adversarial Attack Detection and Localization in the Representation Space of Optical Flow Networks
Hannah Kim (Duke University), Skyler Speakman (IBM Research Africa)
Adversarial AttackConvolutional Neural NetworkOptical FlowVideo
🎯 What it does: Proposed an unsupervised optical flow network adversarial patch attack detection and localization framework (SADL), which identifies and localizes attacks by performing spatially constrained subset scanning on inner features of pre-trained optical flow networks.
Spatially Covariant Lesion Segmentation
Hang Zhang (Cornell University), Jiahao Li (Cornell University)
SegmentationConvolutional Neural NetworkBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: Design and propose a spatial covariant pixel-aligned classifier (SCP) for medical image lesion segmentation, improving computational efficiency while maintaining or enhancing accuracy.
Specifying and Testing k-Safety Properties for Machine-Learning Models
Maria Christakis (MPI-SWS), Valentin Wüstholz (ConsenSys)
Safty and PrivacyImageTextTabularAudio
🎯 What it does: This paper proposes a new specification language called NOMOS for specifying and testing k-safety properties of machine learning models, and demonstrates its broad applicability across multiple domains.
Speeding Up Multi-Objective Hyperparameter Optimization by Task Similarity-Based Meta-Learning for the Tree-Structured Parzen Estimator
Shuhei Watanabe (University of Freiburg), Frank Hutter (University of Freiburg)
Hyperparameter SearchMeta LearningBenchmark
🎯 What it does: Propose a meta-learning method for multi-objective tree-structured Parzen estimator (MO-TPE), which weights high-quality configurations within tasks based on task similarity to accelerate hyperparameter optimization.
Sph2Pob: Boosting Object Detection on Spherical Images with Planar Oriented Boxes Methods
Xinyuan Liu (Key Laboratory of Intelligent Information Processing of Chinese Academy of Sciences, Institute of Computing Technology, Chinese Academy of Sciences), Feng Dai (Key Laboratory of Intelligent Information Processing of Chinese Academy of Sciences, Institute of Computing Technology, Chinese Academy of Sciences)
Object DetectionImage
🎯 What it does: The paper proposes a spherical-to-planar box transformation method called Sph2Pob, and based on this, introduces differentiable Sph2Pob-IoU and flexible, scalable Sph2Pob-Loss, significantly improving the performance of object detection on spherical images.
Spike Count Maximization for Neuromorphic Vision Recognition
Jianxiong Tang (Sun Yat-sen University), Lingxiao Yang (Sun Yat-sen University)
RecognitionComputational EfficiencySpiking Neural NetworkImage
🎯 What it does: Proposes an SNN training framework based on Output Spike Count Maximization (SCM), combining structural risk minimization and a specially designed spike count loss to form a two-stage iterative training algorithm.
Spotlight News Driven Quantitative Trading Based on Trajectory Optimization
Mengyuan Yang (Zhejiang University), MengHan Wang (eBay Inc.)
OptimizationConvolutional Neural NetworkTransformerSupervised Fine-TuningReinforcement LearningTextTabularTime SeriesFinance RelatedStochastic Differential Equation
🎯 What it does: Proposed a news-driven quantitative trading framework based on reinforcement learning called SpotlightTrader, which generates continuous and flexible trading decisions using a trajectory optimization model, and introduces illumination news screening and state trajectory modeling, combined with a training pipeline of offline pre-training and online fine-tuning.
SQuAD-SRC: A Dataset for Multi-Accent Spoken Reading Comprehension
Yixuan Tang (National University of Singapore), Anthony K.H: Tung
RecognitionTransformerSupervised Fine-TuningTextMultimodalityBenchmarkAudio
🎯 What it does: Constructed a large-scale, multi-accent naturally recorded speech reading comprehension dataset named SQuAD-SRC, and conducted experiments on question answering tasks involving multi-accent speech questions and textual context.
SS-BSN: Attentive Blind-Spot Network for Self-Supervised Denoising with Nonlocal Self-Similarity
Young-Joo Han (University of Seoul), Ha-Jin Yu (University of Seoul)
RestorationConvolutional Neural NetworkImage
🎯 What it does: Proposed a self-supervised blind spot network (SS-BSN) and introduced a lightweight self-similar attention (SS-Attention) to capture non-local self-similarity, enhancing the performance of self-supervised image denoising.
SSML-QNet: Scale-Separative Metric Learning Quadruplet Network for Multi-modal Image Patch Matching
Xiuwei Zhang (Northwestern Polytechnical University), Yanning Zhang (Northwestern Polytechnical University)
RetrievalConvolutional Neural NetworkContrastive LearningImageMultimodality
🎯 What it does: Designed and implemented a four-branch multimodal image patch matching network called SSML-QNet, enhancing cross-modal matching performance through scale-separated metric learning and attention mechanisms.
Stability and Generalization of lp-Regularized Stochastic Learning for GCN
Shiyu Liu (University of Electronic Science and Technology of China), Ming Li (Zhejiang Normal University)
ClassificationOptimizationGraph Neural NetworkGraph
🎯 What it does: This paper proposes a stochastic learning method using ℓp (1 < p ≤ 2) regularization in GCN and designs a Proximal SGD algorithm with approximate projection, investigating its stability and generalization performance.
StackFLOW: Monocular Human-Object Reconstruction by Stacked Normalizing Flow with Offset
Chaofan Huo (ShanghaiTech University), Jingya Wang (ShanghaiTech University)
Pose EstimationFlow-based ModelImage
🎯 What it does: This paper proposes a spatial relationship encoding based on Human-Object Offset, and utilizes Stacked Normalizing Flow to infer human-object relative poses from a single image. Subsequently, monocular 3D reconstruction of human-object interactions is achieved through optimization.
Statistically Significant Concept-based Explanation of Image Classifiers via Model Knockoffs
Kaiwen Xu (University Of Tsukuba), Jun Sakuma (Tokyo Institute Of Technology)
ClassificationExplainability and InterpretabilityAuto EncoderImage
🎯 What it does: This paper proposes a method that uses model knockoffs in concept-based explanations to control the false discovery rate (FDR), and further enhances the sparsity and interpretability of concept selection through concept sparse regularization (CSR).
Stochastic Feature Averaging for Learning with Long-Tailed Noisy Labels
Hao-Tian Li (Southeast University), Min-Ling Zhang (Southeast University)
ClassificationImage
🎯 What it does: This paper proposes a framework based on Stochastic Feature Averaging (SFA) to simultaneously address learning problems under long-tailed distributions and label noise.
Stochastic Population Update Can Provably Be Helpful in Multi-Objective Evolutionary Algorithms
Chao Bian (Nanjing University), Chao Qian (Nanjing University)
OptimizationBenchmark
🎯 What it does: This paper conducts a theoretical analysis of the population update mechanism in multi-objective evolutionary algorithms (MOEA), proposes and validates a randomized population update method, and proves that it can significantly reduce the expected runtime of SMS-EMOA on the OneJumpZeroJump benchmark problem;
StockFormer: Learning Hybrid Trading Machines with Predictive Coding
Siyu Gao (Shanghai Jiao Tong University), Xiaokang Yang (Shanghai Jiao Tong University)
TransformerReinforcement LearningTime SeriesFinance Related
🎯 What it does: Propose a hybrid trading agent called StockFormer that combines predictive coding with actor-critic reinforcement learning (RL) for investment decision-making in financial markets.
Strategic Adversarial Attacks in AI-assisted Decision Making to Reduce Human Trust and Reliance
Zhuoran Lu (Purdue University), Ming Yin (Purdue University)
Adversarial AttackConvolutional Neural NetworkImage
🎯 What it does: This paper investigates how attackers strategically deploy adversarial attacks to reduce human trust and reliance on AI models, and validates the impact of attack timing on trust degradation through human-machine experiments; subsequently, it proposes an attack decision framework based on input-output hidden Markov models, dynamically deciding when to launch attacks to maximize the attacker's benefits.
Strategic Resource Selection with Homophilic Agents
Jonathan Gadea Harder (University of Potsdam), Alexander Skopalik (University of Twente)
Optimization
🎯 What it does: Proposed and analyzed the Schelling Resource Selection Game (SRSG), studying strategy selection and equilibrium issues for similar agents under limited accessible resources
Strip Attention for Image Restoration
Yuning Cui (Technical University of Munich), Alois Knoll (Technical University of Munich)
RestorationConvolutional Neural NetworkImage
🎯 What it does: Proposed a Strip Attention Network (SANet) that achieves efficient image restoration through horizontal and vertical local attention mechanisms, replacing traditional global self-attention.
Structural Hawkes Processes for Learning Causal Structure from Discrete-Time Event Sequences
Jie Qiao (Guangdong University of Technology), Zhifeng Hao (Shantou University)
OptimizationExplainability and InterpretabilityTime Series
🎯 What it does: This paper proposes a Structured Hawkes Process (SHP) that learns the causal structure between event types by leveraging the immediate effects in discrete-time event sequences.
STS-GAN: Can We Synthesize Solid Texture with High Fidelity from Arbitrary 2D Exemplar?
Xin Zhao (Shandong Provincial Key Laboratory of Preparation and Measurement of Building Materials University of Jinan), Bo Yang (Quan Cheng Laboratory)
Data SynthesisConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: Propose a multi-scale solid texture synthesis framework based on generative adversarial networks (STS-GAN), which can generate high-fidelity 3D solid textures from arbitrary 2D texture samples.
Sub-Band Based Attention for Robust Polyp Segmentation
Xianyong Fang (Anhui University), Zhengyi Liu (Anhui University)
SegmentationConvolutional Neural NetworkTransformerBiomedical Data
🎯 What it does: This paper proposes using spectral domain subband attention and Transformer-enhanced convolutional encoder for colon polyp image segmentation.
SWAT: Spatial Structure Within and Among Tokens
Kumara Kahatapitiya (Stony Brook University), Michael S. Ryoo (Stony Brook University)
ClassificationSegmentationTransformerImage
🎯 What it does: Proposed structure-aware tokenization and mixing methods to preserve the internal spatial structure of tokens in visual Transformer/mixer models, achieving improvements across multiple models.
Synthesizing Resilient Strategies for Infinite-Horizon Objectives in Multi-Agent Systems
David Klaška (Masaryk University), Vojtěch Řehák (Masaryk University)
OptimizationReinforcement LearningGraph
🎯 What it does: This paper proposes a new objective function language FTRV and designs an algorithm based on differentiable programming and gradient descent to synthesize recoverable and stochastically stable infinite-horizon policies for multi-agent systems.
Targeting Minimal Rare Itemsets from Transaction Databases
Amel Hidouri (Universite d'Artois), Said Jabbour (Universite d'Artois)
TabularBenchmark
🎯 What it does: This paper introduces the concept of k-minimal rare itemsets (k-MRI) and designs a SAT-based framework along with decomposed encoding to efficiently enumerate all k-MRI from transactional databases.
TDG4Crowd:Test Data Generation for Evaluation of Aggregation Algorithms in Crowdsourcing
Yili Fang (Zhejiang Gongshang University), Xinyi Ding (Zhejiang Gongshang University)
GenerationData SynthesisAuto EncoderImageText
🎯 What it does: Propose an automated generation method named TDG4Crowd to generate comprehensive and balanced crowdsourced annotation datasets for evaluating the performance of aggregation algorithms.
Teacher Assistant-Based Knowledge Distillation Extracting Multi-level Features on Single Channel Sleep EEG
Heng Liang (Brainnetome Center, Institute of Automation, Chinese Academy of Sciences), Ziyu Jia (Brainnetome Center, Institute of Automation, Chinese Academy of Sciences)
ClassificationKnowledge DistillationTime SeriesBiomedical Data
🎯 What it does: Propose a generic sleep stage classification model distillation framework called SleepKD, which utilizes multi-level feature distillation and teacher assistant modules to compress the model.
Teaching What You Should Teach: A Data-Based Distillation Method
Shitong Shao (Beijing Institute of Technology), Xinxiao Wu (Beijing Institute of Technology)
ClassificationObject DetectionSegmentationKnowledge DistillationImage
🎯 What it does: This paper proposes a method called TST based on data distillation, which enhances the generalization capability of the student model by learning a differentiable data augmentation encoder with prior bias, and searching for augmented samples where the teacher excels but the student is weak.
Temporal Constrained Feasible Subspace Learning for Human Pose Forecasting
Gaoang Wang (Zhejiang University), Mingli Song (Zhejiang University)
Pose EstimationGraph Neural NetworkTime SeriesSequential
🎯 What it does: Propose a human pose prediction method based on Temporal Constraint Feasible Subspace Learning (TCSL), which uses subspace transformation and projection to ensure predicted poses strictly satisfy temporal constraints on velocity changes during inference.
Temporal Datalog with Existential Quantification
Matthias Lanzinger (University of Oxford), Przemysław A. Wałęga (University of Oxford)
Time Series
🎯 What it does: This paper proposes a language called DatalogMTL∃, which incorporates existential quantifiers into temporal Datalog, and defines and analyzes its natural semantics and uniform semantics.
Temporal Network Creation Games
Davide Bilò (University of L'Aquila), George Skretas (Hasso Plattner Institute)
OptimizationGraph
🎯 What it does: Proposed and analyzed the Temporal Network Creation Game (TRNCG), studying how self-interested agents construct temporal subgraphs by purchasing edges on a complete temporal host graph to ensure all nodes are mutually reachable via temporal paths;
Text-Video Retrieval with Disentangled Conceptualization and Set-to-Set Alignment
Peng Jin (Peking University), Jie Chen (Peking University)
RetrievalTransformerContrastive LearningVideoTextMultimodality
🎯 What it does: This paper proposes a method for achieving finer-grained text-video retrieval by decomposing video and text features into multi-dimensional latent concepts and performing adaptive pooling.
TG-VQA: Ternary Game of Video Question Answering
Hao Li (Peking University), Jie Chen (Peking University)
Knowledge DistillationTransformerVision Language ModelVideoTextMultimodality
🎯 What it does: For the video question answering (VideoQA) task, the TG-VQA model is proposed, which generates fine-grained visual-language alignment labels through ternary games (video, question, answer), and combines token merging with an alignment network to achieve more precise multimodal fusion and answer prediction.
The #DNN-Verification Problem: Counting Unsafe Inputs for Deep Neural Networks
Luca Marzari (University of Verona), Alessandro Farinelli (University of Verona)
Safty and Privacy
🎯 What it does: Studied the counting version of safety verification for deep neural networks, proposing exact and approximate counting methods.
The Computational Complexity of Single-Player Imperfect-Recall Games
Emanuel Tewolde (Carnegie Mellon University), Paul W. Goldberg (University of Oxford)
Optimization
🎯 What it does: This paper conducts theoretical research on single-player extensive-form games with incomplete recall, defining two equilibria based on causal decision theory and evidential decision theory, and relating them to global optima, subset optima, and KKT points in polynomial optimization, analyzing their computational complexity.
The Effects of AI Biases and Explanations on Human Decision Fairness: A Case Study of Bidding in Rental Housing Markets
Xinru Wang (Purdue University), Ming Yin (Purdue University)
Explainability and InterpretabilityTabularFinance Related
🎯 What it does: In the context of rental housing bidding, a randomized experiment was conducted using MTurk participants to investigate the impact of AI model bias levels and the provision of explanations on the fairness of human decision-making (both decision outcomes and processes), and further assess whether these effects persist after AI assistance ends.
The First Proven Performance Guarantees for the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) on a Combinatorial Optimization Problem
Sacha Cerf (Ecole Polytechnique Institut Polytechnique de Paris), Simon Wietheger (Hasso Plattner Institute University of Potsdam)
Optimization
🎯 What it does: Proposed the first mathematical runtime guarantee for NSGA-II on the two-objective minimum spanning tree problem, proving that it can find all extreme points within expected O(m w_max (2 log(n w_max))) steps.
The Hardness of Reasoning about Probabilities and Causality
Benito van der Zander (University of Lbeck), Maciej Liśkiewicz (University of Lbeck)
🎯 What it does: Studied the satisfiability problem for a language capable of expressing quantitative probabilistic reasoning and causal effects (do-operators), conducting a rigorous analysis from the perspective of computational complexity.
The Parameterized Complexity of Finding Concise Local Explanations
Sebastian Ordyniak (University of Leeds), Stefan Szeider (TU Wien)
Explainability and Interpretability
🎯 What it does: This study investigates the computational complexity of calculating minimal local explanations (Anchors) in black-box machine learning models, proving that it is NP-hard in general cases. It systematically explores the impact of various parameter combinations (Anchor size k, coverage size c, maximum difference δmax, treewidth tw, rankwidth rw, bipartite width tww, etc.) on the solvability of the problem from a parameterized complexity perspective, mapping out a complete complexity landscape.
Ties in Multiwinner Approval Voting
Łukasz Janeczko (AGH University), Piotr Faliszewski (AGH University)
🎯 What it does: This paper studies the computational complexity of determining whether an election outcome results in a tie and counting the number of tie election committees in multiwinner approval voting;
Timestamp-Supervised Action Segmentation from the Perspective of Clustering
Dazhao Du (Institute of Software, Chinese Academy of Science), Fuchun Sun (Institute of Software, Chinese Academy of Science)
SegmentationConvolutional Neural NetworkTransformerVideo
🎯 What it does: This paper proposes a timestamp-supervised video action segmentation framework based on clustering, which gradually generates high-quality pseudo-label sequences and trains the segmentation network through two modules: pseudo-label integration and iterative clustering.
TITAN : Task-oriented Dialogues with Mixed-Initiative Interactions
Sitong Yan (Xidian University), Guangneng Hu (Xidian University)
GenerationTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: In this study, the authors constructed a multi-domain task-oriented dialogue dataset called TITAN, which includes system-side proactive interaction strategies, and evaluated multiple baseline models for response generation and dialogue behavior prediction on it.
Topological Planning with Post-unique and Unary Actions
Guillaume Prévost (Académie Militaire de Saint-Cyr Coetquidan), Éric Jacopin (Hawkswell Studios)
OptimizationComputational EfficiencyGraph
🎯 What it does: This paper analyzes the planning requirements of NPCs in games and proposes three new solvable planning problem classes (SAS-PUC 0, SAS-PUC S 2, SAS-PUC ∗ 2). It also designs a linear-time topological planning algorithm called TopoPlan, which can rapidly generate NPC action plans while satisfying constraints such as actions having a single successor, unique successors, and each action appearing only once.
Totally Dynamic Hypergraph Neural Networks
Peng Zhou (Guangxi Normal University), Xiaofeng Zhu (Guangxi Normal University)
ClassificationGraph Neural NetworkPoint CloudGraph
🎯 What it does: Propose a fully dynamic hypergraph neural network (TDHNN) that can learn hyperedge feature distribution during training and dynamically adjust the number of hyperedges based on this distribution to achieve a more suitable hypergraph structure.
Toward Convex Manifolds: A Geometric Perspective for Deep Graph Clustering of Single-cell RNA-seq Data
Nairouz Mrabah (University of Quebec at Montreal), Abdoulaye Banire Diallo (University of Quebec at Montreal)
Representation LearningGraph Neural NetworkAuto EncoderGenerative Adversarial NetworkBiomedical Data
🎯 What it does: Designed and implemented a deep graph clustering model called scTCM for single-cell RNA-seq data, improving the geometry of the latent space by incorporating local flattening and global convexification mechanisms during pre-training and clustering stages.
Towards a Better Understanding of Learning with Multiagent Teams
David Radke (University of Waterloo), Kyle Tilbury (University of Waterloo)
Reinforcement Learning
🎯 What it does: This paper studies the impact of multi-agent team structure on individual learning processes, combining theoretical analysis and experiments to evaluate the optimal team size.
Towards Accurate Video Text Spotting with Text-wise Semantic Reasoning
Xinyan Zu (Fudan University), Xiangyang Xue (Fudan University)
RecognitionObject DetectionObject TrackingSuper ResolutionTransformerLarge Language ModelVision Language ModelContrastive LearningVideo
🎯 What it does: Proposed a video text recognition framework called VLSpotter, which combines text super-resolution, language models, and inter-text semantic reasoning to achieve end-to-end video text detection, tracking, and recognition.
Towards an Integrated View of Semantic Annotation for POIs with Spatial and Textual Information
Dabin Zhang (Shandong University), Meng Chen (Shandong University)
ClassificationGraph Neural NetworkTransformerTextGraph
🎯 What it does: This paper proposes a model called STPA that uses the geographic coordinates and names of static POIs to perform semantic annotation of POIs using only spatial and textual information.
Towards Collaborative Plan Acquisition through Theory of Mind Modeling in Situated Dialogue
Cristian-Paul Bara (University of Michigan), Joyce Chai (University of Michigan)
Convolutional Neural NetworkRecurrent Neural NetworkTransformerVision-Language-Action ModelVideoTextMultimodalityGraph
🎯 What it does: This study introduces the Collaborative Plan Acquisition task for the first time on the existing MindCraft benchmark, aiming to enable agents to predict and complete missing task knowledge through multimodal dialogue and visual information when facing partial plans from themselves and their partners, while exploring the integration of dialogue actions and Theory of Mind (ToM) modeling into this process.
Towards Generalizable Reinforcement Learning for Trade Execution
Chuheng Zhang (Microsoft Research), Li Zhao (Microsoft Research)
Reinforcement LearningTime SeriesFinance Related
🎯 What it does: This paper proposes a framework named Offline Reinforcement Learning with Dynamic Context (ORDC) to address the generalization challenges in learning trading execution strategies from limited offline market data, and provides theoretical generalization bounds. Based on this framework, two context aggregation methods (CASH, handcrafted statistical guidance, and CATE, end-to-end learning) and a high-fidelity simulator based on historical order book (LOB) data are designed. Experiments on simplified tasks and real stock LOB data validate the effectiveness of these methods in reducing overfitting, improving transaction cost performance, and enhancing generalization capabilities.