AAAI 2023 Papers — Page 12
AAAI Conference on Artificial Intelligence · 1578 papers
Posterior Coreset Construction with Kernelized Stein Discrepancy for Model-Based Reinforcement Learning
Souradip Chakraborty (University of Maryland), Dinesh Manocha (University of Maryland)
Reinforcement Learning
🎯 What it does: A posterior conjugate subset construction model-based reinforcement learning framework KSRL is proposed, which is based on the Stein method and Kernelized Stein Discrepancy (KSD). It achieves sublinear convergence of Bayesian regret in continuous spaces and significantly reduces the complexity of posterior representation.
PPGenCDR: A Stable and Robust Framework for Privacy-Preserving Cross-Domain Recommendation
Xinting Liao (Zhejiang University), Chaochao Chen (Midea Group)
Recommendation SystemSafty and PrivacyGenerative Adversarial NetworkTabular
🎯 What it does: This paper proposes the PPGenCDR framework, which protects the privacy of source domain users while improving recommendation performance in the target domain.
Practical Cross-System Shilling Attacks with Limited Access to Data
Meifang Zeng (Xiamen University), Hui Li (PLA Strategic Support Force Information Engineering University)
Recommendation SystemAdversarial AttackGraph Neural NetworkContrastive LearningGraph
🎯 What it does: This paper proposes a cross-system attack framework called PC-Attack, which learns graph topology knowledge using self-supervised graph contrastive learning on publicly available recommendation system data, and fine-tunes the model with only a small amount of target system data (≤10%) to generate pseudo-user poisoning attacks on the target recommendation system.
Practical Markov Boundary Learning without Strong Assumptions
Xingyu Wu (University of Science and Technology of China), Huanhuan Chen (University of Science and Technology of China)
OptimizationTabular
🎯 What it does: A new Markov boundary (MB) learning algorithm called KMB is proposed, which is based on the kernel space conditional covariance operator and can effectively perform feature selection on real-world data that does not meet traditional assumptions.
Practical Parallel Algorithms for Submodular Maximization Subject to a Knapsack Constraint with Nearly Optimal Adaptivity
Shuang Cui (University of Science and Technology of China), Aakas Zhiyuli (Alibaba Group)
Recommendation SystemOptimizationGraphTabular
🎯 What it does: Two low-fitness parallel algorithms, ParSKP1 and ParSKP2, are designed for maximizing non-monotone submodular functions under knapsack constraints.
Predicate Invention for Bilevel Planning
Tom Silver (Massachusetts Institute of Technology), Joshua B. Tenenbaum
OptimizationRobotic IntelligenceReinforcement Learning
🎯 What it does: Learn new predicates from demonstration data and build abstract planning models based on them to achieve efficient two-layer planning.
Predict+Optimize for Packing and Covering LPs with Unknown Parameters in Constraints
Xinyi Hu (Chinese University of Hong Kong), Jimmy H.M. Lee (University of Wisconsin-Madison)
OptimizationTabularBenchmark
🎯 What it does: A new Predict+Optimize framework has been developed, allowing for constraints with unknown parameters and incorporating post-hoc correction functions and penalties.
Predicting Temporal Sets with Simplified Fully Connected Networks
Le Yu (Beihang University), Weifeng Lv (Beihang University)
Recommendation SystemRecurrent Neural NetworkGraph Neural NetworkSequential
🎯 What it does: A concise architecture is designed that uses only a Simplified Fully Connected Network (SFCN) to predict the next elements in a user's historical behavior sequence.
Predictive Exit: Prediction of Fine-Grained Early Exits for Computation- and Energy-Efficient Inference
Xiangjie Li (Shanghai Jiao Tong University), An Zou (Shanghai Jiao Tong University)
Computational EfficiencyConvolutional Neural NetworkImage
🎯 What it does: Proposes the Predictive Exit framework, which uses a low-cost prediction engine to anticipate network exit points in advance, thereby executing the offline layer only once and dynamically adjusting frequency/voltage to achieve savings in computation and energy consumption.
Predictive Multiplicity in Probabilistic Classification
Jamelle Watson-Daniels (Harvard University), Berk Ustun (University of California San Diego)
ClassificationOptimizationTabularBiomedical DataFinance Related
🎯 What it does: This paper studies the prediction multiplicity in probabilistic classification tasks, defining feasible prediction ranges, ambiguity, and disparity, and provides calculation methods.
Preference-Controlled Multi-Objective Reinforcement Learning for Conditional Text Generation
Wenqing Chen (Sun Yat-sen University), Yaohui Jin (Shanghai Jiao Tong University)
GenerationRecurrent Neural NetworkTransformerReinforcement LearningAuto EncoderText
🎯 What it does: A preference-controlled multi-objective reinforcement learning (PCMORL) framework is proposed for conditional text generation, explicitly optimizing semantic fidelity (CIDEr) and diversity (Self-CIDEr), and dynamically adjusting the trade-off between the two during the inference phase through the preference variable r.
Preserve Context Information for Extract-Generate Long-Input Summarization Framework
Ruifeng Yuan (Hong Kong Polytechnic University), Wenjie Li (Hong Kong Polytechnic University)
GenerationTransformerSupervised Fine-TuningText
🎯 What it does: Aiming at the extract-generate framework in long text summarization, a CAEG method is proposed to maintain local and global context through contextual prompts.
Preserving Structural Consistency in Arbitrary Artist and Artwork Style Transfer
Jingyu Wu (Alibaba-Zhejiang University Joint Institute of Frontier Technologies), Lingyun Sun
Image TranslationGenerationConvolutional Neural NetworkAuto EncoderGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes a Dual Style Transfer Module (DSTM) and an Edge Enhancement Module (EEM) to simultaneously extract the artist's style and the artwork's style from a single piece of art, generating high-quality, structurally consistent style transfer results under any style combination.
PrimeNet: Pre-training for Irregular Multivariate Time Series
Ranak Roy Chowdhury (University of California San Diego), Jingbo Shang (University of California San Diego)
ClassificationAnomaly DetectionRepresentation LearningTransformerContrastive LearningTime SeriesBiomedical DataElectronic Health Records
🎯 What it does: This paper proposes PrimeNet, a self-supervised pre-training model that utilizes time-sensitive contrastive learning and time-sensitive reconstruction tasks to learn representations of irregular multivariate time series, and fine-tunes it for downstream tasks.
Principled and Efficient Motif Finding for Structure Learning of Lifted Graphical Models
Jonathan Feldstein (Bennu.AI), Efthymia Tsamoura (Samsung AI)
Graph Neural NetworkGraph
🎯 What it does: This paper addresses the motif discovery problem in structure learning and proposes a new PRISM algorithm that can automatically identify structural patterns from data and accelerate model building.
Principled Data-Driven Decision Support for Cyber-Forensic Investigations
Soodeh Atefi (University of Houston), Aron Laszka (Pennsylvania State University)
Reinforcement Learning
🎯 What it does: A principled data-driven decision support framework based on Markov Decision Processes (MDP) is proposed to guide the selection of technical priorities in network forensic investigations.
Priori Anchor Labels Supervised Scalable Multi-View Bipartite Graph Clustering
Jiali You (Southwest University of Science and Technology), Yuancheng Yao (Southwest University of Science and Technology)
Graph Neural NetworkGraph
🎯 What it does: A scalable multi-view bidirectional graph clustering method SMGC based on prior anchor labels is proposed to address the issue of non-uniform sampling of anchor points.
Privacy Attacks on Schedule-Driven Data
Stephan A. Fahrenkrog-Petersen (Humboldt-Universitat zu Berlin), Matthias Weidlich (Humboldt-Universitat zu Berlin)
OptimizationSafty and PrivacyTabular
🎯 What it does: This paper presents a study on privacy attacks against publicly available scheduling data. It first constructs a threat model for public scheduling, then defines a distance-based privacy loss metric, and analyzes the theoretical characteristics of both information-free and information-rich attacks. Subsequently, it applies the inverse scheduling problem to single-machine total weighted completion time (TWCT) scheduling, providing a constraint satisfaction solving method, complexity analysis, and validating the effectiveness of information-rich attacks through large-scale synthetic scheduling experiments.
Probabilistic Generalization of Backdoor Trees with Application to SAT
Alexander Semenov (ITMO University), Ibragim Dzhiblavi (ITMO University)
OptimizationTabular
🎯 What it does: This paper extends the concept of strong backdoor sets (SBS) and introduces the probability backdoor tree (ρ-backdoor tree), providing its theoretical properties. It then implements an efficient search method based on evolutionary algorithms to quickly construct ρ-backdoor trees in SAT instances.
Probabilities of Potential Outcome Types in Experimental Studies: Identification and Estimation Based on Proxy Covariate Information
Ryusei Shingaki (Yokohama National University), Manabu Kuroki (Yokohama National University)
Tabular
🎯 What it does: A new identification condition and an unbiased estimation method for the probability of potential outcome types based on proxy covariates are proposed, and their consistency and asymptotic normality are verified through numerical experiments.
Probability Guided Loss for Long-Tailed Multi-Label Image Classification
Dekun Lin (Chengdu Institute of Computer Applications)
ClassificationConvolutional Neural NetworkImage
🎯 What it does: This paper addresses the long-tail multi-label image classification problem and proposes a Probability Guided Loss (PG loss) to enhance classification performance by controlling the probability growth and the gap between positive and negative probabilities during the training process.
Probably Approximate Shapley Fairness with Applications in Machine Learning
Zijian Zhou (National University of Singapore), Bryan Kian Hsiang Low (National University of Singapore)
Tabular
🎯 What it does: A concept of probabilistic approximate Shapley fairness is proposed, and a Greedy Active Estimation (GAE) algorithm is designed to improve the fairness and accuracy of Shapley value estimation under a limited sample budget.
Progress and Limitations of Deep Networks to Recognize Objects in Unusual Poses
Amro Abbas (African Institute For Mathematical Sciences), Stéphane Deny (Aalto University)
RecognitionObject DetectionPose EstimationConvolutional Neural NetworkTransformerContrastive LearningImage
🎯 What it does: This paper systematically evaluates the robustness of 38 mainstream deep networks in recognizing objects under unusual poses by constructing a synthetic dataset called ObjectPose, and explores the impact of dataset size, network size, training strategies, and multiple transformations on performance.
Progressive Bayesian Inference for Scribble-Supervised Semantic Segmentation
Chuanwei Zhou (Nanjing University of Science and Technology), Zhen Cui (Nanjing University of Science and Technology)
SegmentationConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a Progressive Bayesian Inference (PBI) framework that alternates between optimizing a segmenter and a Bayesian inference module, using sparse scribble annotations to infer the semantic distribution of unlabeled pixels, thereby enhancing the performance of scribble-based semantic segmentation.
Progressive Deep Multi-View Comprehensive Representation Learning
Cai Xu (Xidian University), Xiangyu Song (Swinburne University of Technology)
Representation LearningSupervised Fine-TuningTabular
🎯 What it does: A Progressive Deep Multi-View Fusion (PDMF) framework is proposed, which learns to assist complete representation during the pre-training phase and captures the consistency and complementarity between different views. In the fine-tuning phase, view-specific encoders are learned, and Multi-View Sparse Batch Normalization (MSBN) is used to achieve alignment and fusion of view-specific representations, ultimately resulting in a comprehensive multi-view representation.
Progressive Few-Shot Adaptation of Generative Model with Align-Free Spatial Correlation
Jongbo Moon (Sungkyunkwan University), Jae-Pil Heo (Sungkyunkwan University)
GenerationDomain AdaptationGenerative Adversarial NetworkOptical FlowImage
🎯 What it does: A progressive unaligned spatially correlated GAN adaptation method is proposed for target domains with very few samples;
Progressive Multi-View Human Mesh Recovery with Self-Supervision
Xuan Gong (University at Buffalo), Ziyan Wu (United Imaging Intelligence)
Data SynthesisPose EstimationMesh
🎯 What it does: A self-supervised multi-view human mesh recovery method based on synthetic data is proposed, which first extracts a 2D intermediate representation and then projects it into a 3D unified space for regression, ultimately obtaining the SMPL human mesh.
Progressive Neighborhood Aggregation for Semantic Segmentation Refinement
Ting Liu (Northwestern Polytechnical University), Yanning Zhang (Beijing Jiaotong University)
SegmentationConvolutional Neural NetworkImage
🎯 What it does: This paper proposes an Advanced Neighborhood Aggregation (PNA) framework, which aims to refine the coarse predictions of semantic segmentation step by step while maintaining the overall structural integrity of the network, utilizing the spatial structural information and detail information of multi-scale features from the backbone network.
ProKD: An Unsupervised Prototypical Knowledge Distillation Network for Zero-Resource Cross-Lingual Named Entity Recognition
Ling Ge (Beihang University), Jihong Liu (Beihang University)
RecognitionKnowledge DistillationTransformerContrastive LearningText
🎯 What it does: For the zero-resource cross-lingual named entity recognition task, we propose an unsupervised prototype knowledge distillation network (ProKD), which enhances the language-agnostic knowledge of the teacher network through prototype alignment driven by contrastive learning, and employs prototype self-training in the student network to integrate teacher priors to improve the learning of target language-specific knowledge.
Prompt-Augmented Linear Probing: Scaling beyond the Limit of Few-Shot In-Context Learners
Hyunsoo Cho (Seoul National University), Taeuk Kim (Hanyang University)
ClassificationTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: This paper proposes a hybrid method that combines linear probing and in-context learning—Prompt-Augmented Linear Probing (PALP). By adding templates or demonstrations before the input, it improves the representation of language models, thereby enhancing classification performance in black-box tuning scenarios.
Prompting Neural Machine Translation with Translation Memories
Abudurexiti Reheman (Northeastern University), Jingbo Zhu (NiuTrans Research)
Domain AdaptationTransformerPrompt EngineeringTextRetrieval-Augmented Generation
🎯 What it does: This paper proposes a method that utilizes retrieved translation memory (TM) as a prompt during NMT decoding, concatenating it with the source and target sentences, first forcing the generation of the target TM, and then automatically generating the translation while keeping the original NMT model unchanged.
Properties of Position Matrices and Their Elections
Niclas Boehmer (Technische Universitaet Berlin), Tomasz Wąs (Pennsylvania State University)
Tabular
🎯 What it does: This paper studies the properties of elections given a position matrix, proving that counting the elections that realize this matrix is #P-complete. It proposes a sampling algorithm without hard guarantees and provides decision algorithms and necessary conditions for structured domains (unimodal, group-separation) and the Condorcet winner problem. Finally, it evaluates the diversity implied by the position matrix through experiments.
Proportional Decisions in Perpetual Voting
Martin Lackner (Vienna University of Technology), Jan Maly (University of Amsterdam)
🎯 What it does: This paper studies the proportionality issues in the long-term decision-making framework of perpetual voting, proposing new proportionality axioms (lower/upper quotas for closed groups) and conducting axiomatic analysis and comparison of different classes of rules (winning/losing weight methods, Perpetual Consensus, Perpetual Phragmén, etc.).
Proportionality in Approval-Based Participatory Budgeting
Markus Brill (University of Warwick), Jannik Peters (Technische Universität Berlin)
🎯 What it does: This paper studies the proportional representativeness axiom in participatory budgeting (PB) based on approval voting, proposing a unified framework of satisfaction functions that meet this axiom, and theoretically analyzes the realizability of proportional axioms such as EJR and PJR under different satisfaction functions.
Prototypical Fine-Tuning: Towards Robust Performance under Varying Data Sizes
Yiqiao Jin (Georgia Institute of Technology), Xing Xie (Microsoft Research Asia)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Prototypical Fine-tuning (PFit) is proposed, which combines non-parametric prototypical networks with large pre-trained language models, automatically adjusting model capacity to accommodate different data scales, especially improving prediction performance in low-resource scenarios.
Prototypical Partial Optimal Transport for Universal Domain Adaptation
Yucheng Yang (Xi'an Jiaotong University), Jian Sun (Xi'an Jiaotong University)
Domain AdaptationContrastive LearningImage
🎯 What it does: A unified domain adaptation method based on minimum batch prototype partial optimal transport (m-PPOT) is proposed, which re-weights the source prototypes and target samples using transport plans to distinguish between common categories and unknown categories.
Provable Detection of Propagating Sampling Bias in Prediction Models
Pavan Ravishankar (New York University), Daniel B. Neill (New York University)
Tabular
🎯 What it does: The mechanism of differential sampling bias propagation from the data stage to the prediction stage is studied, and a testable threshold is provided. The theoretical results are then validated on the COMPAS and New York City Police Department stop-and-frisk data.
Provable Pathways: Learning Multiple Tasks over Multiple Paths
Yingcong Li (University of California), Samet Oymak (University of Michigan)
Tabular
🎯 What it does: This study investigates the theory and methods for learning task-specific representations in hypernetwork paths and provides a generalization error upper bound.
Provably Efficient Causal Model-Based Reinforcement Learning for Systematic Generalization
Mirco Mutti (Politecnico di Milano), Marcello Restelli (Politecnico di Milano)
OptimizationReinforcement LearningTabular
🎯 What it does: This paper proposes a reinforcement learning algorithm based on causal models, which learns the shared causal structure across environments through reward-free interactions in a limited number of environments, thereby obtaining an approximately optimal policy through planning in unknown environments.
Provably Efficient Primal-Dual Reinforcement Learning for CMDPs with Non-stationary Objectives and Constraints
Yuhao Ding (University of California Berkeley), Javad Lavaei (University of California Berkeley)
OptimizationReinforcement Learning
🎯 What it does: This paper proposes a safe reinforcement learning algorithm PROPD-PPO for non-stationary constrained Markov decision processes (CMDP), which can achieve dynamic scheduling and constraint satisfaction under time-varying rewards, transitions, and constraints.
Proximal Stochastic Recursive Momentum Methods for Nonconvex Composite Decentralized Optimization
Gabriel Mancino-Ball (Rensselaer Polytechnic Institute), Jie Chen (IBM Research)
OptimizationConvolutional Neural NetworkTabular
🎯 What it does: A single-cycle decentralized proximal stochastic recursive momentum algorithm named DEEPSTORM is proposed for solving non-convex stochastic composite optimization problems.
ProxyBO: Accelerating Neural Architecture Search via Bayesian Optimization with Zero-Cost Proxies
Yu Shen (Peking University), Bin Cui (Kuaishou Technology)
OptimizationNeural Architecture SearchTabular
🎯 What it does: A Bayesian optimization framework named ProxyBO is proposed, which utilizes zero-cost proxies to accelerate neural architecture search.
Pseudo Label-Guided Model Inversion Attack via Conditional Generative Adversarial Network
Xiaojian Yuan (University of Science and Technology of China), Yang Zhang (CISPA Helmholtz Center for Information Security)
GenerationAdversarial AttackGenerative Adversarial NetworkImage
🎯 What it does: A pseudo-label based conditional GAN model inverse attack method (PLG-MI) is proposed, which first generates pseudo-labels using a top-n strategy on public data to train a cGAN, and then searches for private images of specified categories in the latent space.
PUnifiedNER: A Prompting-Based Unified NER System for Diverse Datasets
Jinghui Lu (SenseTime Research), Fei Tan (SenseTime Research)
RecognitionTransformerPrompt EngineeringText
🎯 What it does: This paper proposes a prompt-based unified named entity recognition system (PUnifiedNER) that can handle up to 37 entity types from different domains at once, enabling on-demand entity recognition.
PUPS: Point Cloud Unified Panoptic Segmentation
Shihao Su (Zhejiang University), Xi Li (Zhejiang University)
Object DetectionSegmentationAutonomous DrivingTransformerPoint Cloud
🎯 What it does: A PUPS framework is proposed, which uses point-level classifiers to directly predict semantic categories and instance segmentation in point clouds, resulting in a unified panoramic segmentation outcome.
Purifier: Defending Data Inference Attacks via Transforming Confidence Scores
Ziqi Yang (Zhejiang University), Kui Ren (Zhejiang University)
Adversarial AttackAuto EncoderGenerative Adversarial NetworkImage
🎯 What it does: A defense framework named PURIFIER is proposed, which transforms the confidence vector of the target model to make member samples indistinguishable from non-member samples in terms of confidence distribution, thereby suppressing membership inference, model inversion, and attribute inference attacks.
Q-functionals for Value-Based Continuous Control
Samuel Lobel (Brown University), George Konidaris (University of Massachusetts)
Robotic IntelligenceReinforcement LearningSequential
🎯 What it does: Proposed the Q-functional architecture, which uses state-based coefficients and action basis functions to quickly and parallelly evaluate the value of multiple actions, thereby replacing traditional policy networks;
Quality-Aware Self-Training on Differentiable Synthesis of Rare Relational Data
Chongsheng Zhang (Henan University), Ji Liu (Henan University)
ClassificationData SynthesisGenerative Adversarial NetworkTabular
🎯 What it does: A Quality-Aware Self-Training (QAST) framework is proposed to generate high-quality synthetic samples through GAN on scarce relational datasets and automatically label them with pseudo-labels to alleviate the class imbalance problem.
Quantized Feature Distillation for Network Quantization
Ke Zhu (Nanjing University), Jianxin Wu (Nanjing University)
ClassificationObject DetectionSegmentationKnowledge DistillationTransformerSupervised Fine-TuningImage
🎯 What it does: A Quantized Feature Distillation (QFD) method is proposed for low-bit quantization training.
Quantum Multi-Agent Meta Reinforcement Learning
Won Joon Yun (Korea University), Joongheon Kim (Korea University)
Meta LearningReinforcement LearningPhysics Related
🎯 What it does: A quantum multi-agent meta reinforcement learning framework QM2ARL is designed, which first uses angle parameters for meta-learning in stages, and then uses pole parameters for local fine-tuning;
Quantum Multi-Armed Bandits and Stochastic Linear Bandits Enjoy Logarithmic Regrets
Zongqi Wan (Institute of Computing Technology, Chinese Academy of Sciences), Xiaoming Sun (Institute of Computing Technology, Chinese Academy of Sciences)
OptimizationReinforcement Learning from Human FeedbackPhysics Related
🎯 What it does: This paper proposes quantum versions of the multi-armed bandit (MAB) and stochastic linear bandit (SLB) algorithms, utilizing quantum Monte Carlo (QMC) and adaptive staging techniques, proving that an expected cumulative penalty of O(poly(log T)) can be achieved within a time frame of T.
Quantum-Inspired Representation for Long-Tail Senses of Word Sense Disambiguation
Junwei Zhang (Tianjin University), Fengyu Guo (Tianjin Normal University)
Representation LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes a quantum-inspired representation method that maps rare word meanings in word sense disambiguation to superposition states in Hilbert space, and based on this, constructs the QR-WSD model for disambiguating long-tail word meanings.
Query Your Model with Definitions in FrameNet: An Effective Method for Frame Semantic Role Labeling
Ce Zheng (Peking University), Baobao Chang (Peking University)
ClassificationRecognitionTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: A query-based framework AGED is proposed, using the FrameNet framework and frame element definitions as natural language queries to accomplish frame semantic role labeling.
Query-Aware Quantization for Maximum Inner Product Search
Jin Zhang (University of Science and Technology of China), Enhong Chen (University of Science and Technology of China)
RetrievalRecommendation SystemOptimizationContrastive LearningTabularSequential
🎯 What it does: This paper proposes a query-aware quantization method and provides a unified quantization framework to improve the approximation quality of maximum inner product search (MIPS).
Question Decomposition Tree for Answering Complex Questions over Knowledge Bases
Xiang Huang (Nanjing University), Yuzhong Qu (Nanjing University)
GenerationTransformerLarge Language ModelText
🎯 What it does: Proposes a Question Decomposition Tree (QDT) and a two-stage generation method called Clue-Decipher, constructs the QDTrees dataset, and designs the QDTQA question-answering system based on it;
RADIANT: Radar-Image Association Network for 3D Object Detection
Yunfei Long (Michigan State University), Punarjay Chakravarty (Ford Motor Company)
Object DetectionDepth EstimationAutonomous DrivingConvolutional Neural NetworkImageMultimodalityPoint Cloud
🎯 What it does: A radar-image joint 3D object detection network called RADIANT is proposed, which predicts the 3D offset from radar return points to the target center through training the network, achieving the association and fusion of radar depth and monocular detection results, significantly improving depth estimation accuracy and overall detection performance.
RAFaRe: Learning Robust and Accurate Non-parametric 3D Face Reconstruction from Pseudo 2D&3D Pairs
Longwei Guo (Nanjing University), Xun Cao (Nanjing University)
GenerationDepth EstimationNeural Radiance FieldImagePoint CloudMesh
🎯 What it does: A non-parametric single-view 3D face reconstruction method is proposed, using hierarchical implicit SDF and trained with pseudo 2D & 3D datasets.
Random Walk Conformer: Learning Graph Representation from Long and Short Range
Pei-Kai Yeh (National Taiwan University), Ming-Syan Chen (National Taiwan University)
Representation LearningGraph Neural NetworkTransformerGraph
🎯 What it does: A framework combining random walk encoding with Transformer and convolution, called Random Walk Conformer, is proposed to simultaneously capture global relationships and local subgraph patterns in graphs.
Rank Aggregation Using Scoring Rules
Niclas Boehmer (Technische Universität Berlin), Dominik Peters (Université Paris Dauphine-PSL)
🎯 What it does: This paper studies three types of ranking aggregation methods based on scoring rules: score-based ranking, Sequential-Winner (removing candidates in order of highest score), and Sequential-Loser (removing candidates in order of lowest score).
RankDNN: Learning to Rank for Few-Shot Learning
Qianyu Guo (Fudan University), Weifeng Ge (Fudan University)
ClassificationMeta LearningImage
🎯 What it does: A new Few-Shot learning framework called RankDNN is proposed, which transforms multi-class tasks into binary relevance ranking problems and utilizes vector-Kronecker product encoding of image triplets and an MLP classifier for training.
Rawlsian Fairness in Online Bipartite Matching: Two-Sided, Group, and Individual
Seyed Esmaeili (University of Maryland), John P. Dickerson (University of Maryland)
OptimizationTabular
🎯 What it does: This paper designs an online bipartite matching algorithm that achieves Rawlsian fairness for drivers (offline party) and passengers (online party) while maintaining platform revenue.
Reachability Games Modulo Theories with a Bounded Safety Player
Marco Faella (University of Naples Federico II), Gennaro Parlato (University of Molise)
Reinforcement LearningBenchmark
🎯 What it does: This paper proposes and studies logic reachability games (GMTs) with a limited number of actions for safe players, achieving automated solving by directly reducing the win/loss determination and strategy synthesis problems to the satisfiability/unsatisfiability of a Constraint Horn Clause (CHC) system.
Reactive Synthesis of Dominant Strategies
Benjamin Aminof (University of Vienna), Sasha Rubin (University of Sydney)
Optimization
🎯 What it does: The study investigates dominant strategies under LTL / LTLf environmental specifications, providing existence determination, complexity analysis, and a unified optimal algorithm, and compares it with enforcing strategies and best-effort strategies.
READ: Large-Scale Neural Scene Rendering for Autonomous Driving
Zhuopeng Li (Zhejiang University), Jianke Zhu (Zhejiang University)
Autonomous DrivingNeural Radiance FieldPoint Cloud
🎯 What it does: A neural rendering method called READ is proposed, which can render large-scale autonomous driving scenes in real-time on a regular PC. It is capable of generating high-quality multi-view images from sparse point clouds and supports scene editing and stitching.
Real or Fake Text?: Investigating Human Ability to Detect Boundaries between Human-Written and Machine-Generated Text
Liam Dugan (University of Pennsylvania), Chris Callison-Burch (University of Pennsylvania)
ClassificationGenerationTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper studies the boundary detection problem between machine-generated text and human-written text, using the gamified platform RoFT for annotation collection.
Real-World Deep Local Motion Deblurring
Haoying Li (Zhejiang University), Zhihai Xu (Zhejiang University)
RestorationConvolutional Neural NetworkOptical FlowImage
🎯 What it does: The first real local motion blur dataset, ReLoBlur, is proposed, and based on this dataset, a local blur-aware gated network (LBAG) is designed for local deblurring of single images.
Reconstructing an Epidemic Outbreak Using Steiner Connectivity
Ritwick Mishra (University of Virginia), Anil Vullikanti (University of Virginia)
GraphBiomedical DataElectronic Health Records
🎯 What it does: This study investigates the problem of reconstructing the epidemic transmission process based on observed infected nodes and proposes a maximum likelihood estimation (MLE) method called CASCADEMLE.
Recovering the Graph Underlying Networked Dynamical Systems under Partial Observability: A Deep Learning Approach
Sérgio Machado (University of Coimbra), Augusto Santos (Instituto de Telecomunicações)
Convolutional Neural NetworkGraphTime Series
🎯 What it does: The study focuses on recovering the network structure in partially observable linear stochastic network dynamic systems using observed node time series data.
Recurrent Structure Attention Guidance for Depth Super-resolution
Jiayi Yuan (Nanjing University of Science and Technology), Jian Yang (Nanjing University of Science and Technology)
RestorationDepth EstimationSuper ResolutionConvolutional Neural NetworkRecurrent Neural NetworkImage
🎯 What it does: A recursive structure attention-guided deep super-resolution (RSAG) framework is proposed, which uses a deep contrast network (DCN) to adaptively separate high-frequency/low-frequency components, and employs a recursive structure attention (SA) block to fuse the latest depth predictions with image features, enhancing low-frequency edges through HF&LF feature fusion.
Reducing ANN-SNN Conversion Error through Residual Membrane Potential
Zecheng Hao (Peking University), Zhaofei Yu (Peking University)
ClassificationOptimizationSpiking Neural NetworkImage
🎯 What it does: This paper proposes an optimization strategy based on the residual membrane potential, significantly reducing the uneven errors in the conversion from ANN to SNN.
Reducing Domain Gap in Frequency and Spatial Domain for Cross-Modality Domain Adaptation on Medical Image Segmentation
Shaolei Liu (Fudan University), Manning Wang (Fudan University)
SegmentationDomain AdaptationKnowledge DistillationTransformerImageBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: An unsupervised domain adaptation framework is designed, utilizing dual image transfer in the frequency domain (NSCT) and spatial domain (histogram matching) along with multi-teacher distillation to enhance the performance of medical image segmentation in unannotated target domains.
Reducing Sentiment Bias in Pre-trained Sentiment Classification via Adaptive Gumbel Attack
Jiachen Tian (Tianjin University), Zhiyong Feng (Tianjin University)
ClassificationAdversarial AttackTransformerLarge Language ModelSupervised Fine-TuningMixture of ExpertsText
🎯 What it does: An adaptive Gumbel-attack classifier (Gater) is proposed, which generates diverse emotional noise from pre-trained language models and integrates it during training to reduce emotional bias in sentiment classification.
Referring Expression Comprehension Using Language Adaptive Inference
Wei Su (Zhejiang University), Xi Li (Zhejiang University)
RecognitionObject DetectionTransformerVision Language ModelImageText
🎯 What it does: A language-adaptive dynamic subnetwork (LADS) framework is proposed, which can dynamically extract a minimized subnetwork from the global REC network for referential expression understanding based on the referential expression.
Refined Semantic Enhancement towards Frequency Diffusion for Video Captioning
Xian Zhong (Wuhan University of Technology), Mang Ye (University of Central Florida)
GenerationTransformerDiffusion modelVideoText
🎯 What it does: A new video subtitle generation model RSFD is proposed, which improves the generation of low-frequency words through frequency-aware diffusion and divergent semantic supervision.
ReGANIE: Rectifying GAN Inversion Errors for Accurate Real Image Editing
Bingchuan Li (ByteDance), Zili Yi (ByteDance)
Image TranslationRestorationGenerationGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes a two-stage framework called ReGANIE, which uses specialized networks to achieve image editing and correct GAN inversion errors, ultimately enabling high-quality editing of real images.
Reinforced Approximate Exploratory Data Analysis
Shaddy Garg (Adobe Research), Arjun Kashettiwar (Adobe Research)
Reinforcement LearningTabular
🎯 What it does: This paper proposes a sampling strategy selection framework based on deep reinforcement learning, APPROXEDA, to balance query latency and approximation error in interactive exploratory data analysis, preventing deviation from the analysis path.
Reinforcement Causal Structure Learning on Order Graph
Dezhi Yang (Shandong University), Maozu Guo (Beijing University of Civil Engineering and Architecture)
TransformerReinforcement LearningGraph
🎯 What it does: A reinforcement learning framework RCL-OG based on ordered graphs is proposed to approximate the posterior distribution of DAG topological sorting and conduct causal structure learning.
Reinforcement Learning for Branch-and-Bound Optimisation Using Retrospective Trajectories
Christopher W. F. Parsonson (University College London), Thomas D. Barrett (InstaDeep)
OptimizationGraph Neural NetworkReinforcement LearningTabular
🎯 What it does: Proposed the retro branching method, which trains a reinforcement learning agent for branching decisions by constructing subtree trajectories through backtracking;
Reject Decoding via Language-Vision Models for Text-to-Image Synthesis
Fuxiang Wu (Guangdong Provincial Key Laboratory of Robotics and Intelligent System, Shenzhen Institute of Advanced Technology, Chinese University of Hong Kong), Jun Cheng (Guangdong Provincial Key Laboratory of Robotics and Intelligent System, Shenzhen Institute of Advanced Technology, Chinese University of Hong Kong)
GenerationData SynthesisTransformerVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: A Reject Decoding algorithm is proposed, which utilizes a minimal multimodal language-vision model to preemptively eliminate low-quality paths during the generation process, thereby expanding the search space and improving the text-to-image generation quality while maintaining the same computational load.
Relation-Aware Language-Graph Transformer for Question Answering
Jinyoung Park (Korea University), Hyunwoo Kim
TransformerLarge Language ModelTextMultimodalityGraph
🎯 What it does: A question answering framework called QAT (Question Answering Transformer) is proposed, which combines a language model with a knowledge graph. It constructs Meta-Path token pairs to perform relation-centric encoding of multi-hop relationships in the KG, and incorporates Relation-Aware Self-Attention (RASA) and Cross-Modal Relative Position Bias into the self-attention of the Transformer, enabling dynamic interaction and reasoning of multimodal information.
Relational Program Synthesis with Numerical Reasoning
Céline Hocquette (University of Oxford), Andrew Cropper (University of Oxford)
🎯 What it does: A new ILP system called NUMSYNTH is proposed, which can learn interpretable programs containing numerical constants through the combination of SMT solvers, relational learning, and numerical reasoning in an infinite domain.
Reliable Robustness Evaluation via Automatically Constructed Attack Ensembles
Shengcai Liu (Southern University of Science and Technology), Ke Tang (Southern University of Science and Technology)
Adversarial AttackImage
🎯 What it does: This paper proposes AutoAE, an algorithm for automatically constructing attack sets (AE) to reliably assess adversarial robustness.
REMIT: Reinforced Multi-Interest Transfer for Cross-Domain Recommendation
Caiqi Sun (Ant Group), Linjian Mo (Ant Group)
Recommendation SystemReinforcement LearningTabular
🎯 What it does: This study investigates multi-interest transfer and selection in cross-domain recommendation to address the cold start problem.
RenewNAT: Renewing Potential Translation for Non-autoregressive Transformer
Pei Guo (Soochow University), Min Zhang (Soochow University)
GenerationComputational EfficiencyKnowledge DistillationTransformerText
🎯 What it does: The RenewNAT framework is proposed, which first generates potential translations in a one-time decoding process and then refreshes them through a masked language model submodule, integrating the advantages of fully parallel and iterative NAT.
Repair Is Nearly Generation: Multilingual Program Repair with LLMs
Harshit Joshi (Microsoft), Ivan Radiček (Microsoft)
AI Code AssistantTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: This paper proposes a multilingual program error correction engine RING based on a large code language model (Codex), employing a three-stage prompt-based strategy of error localization, code transformation, and candidate ranking.
Rephrasing the Reference for Non-autoregressive Machine Translation
Chenze Shao (Chinese Academy of Sciences), Yang Feng (Tencent Inc)
OptimizationKnowledge DistillationTransformerReinforcement LearningText
🎯 What it does: This paper proposes the introduction of a rephraser module in non-autoregressive machine translation (NAT) training, which rephrases reference sentences using the output of NAT, thereby providing a more suitable training objective for NAT.
RePreM: Representation Pre-training with Masked Model for Reinforcement Learning
Yuanying Cai (Tsinghua University), Longbo Huang (Tsinghua University)
Representation LearningTransformerReinforcement LearningContrastive LearningSequential
🎯 What it does: Using offline trajectories for unsupervised mask prediction pre-training, resulting in a state encoder that can be directly used for downstream reinforcement learning.
Representation Learning by Detecting Incorrect Location Embeddings
Sepehr Sameni (University of Bern), Paolo Favaro (Adobe Research)
SegmentationRepresentation LearningTransformerContrastive LearningImage
🎯 What it does: This paper studies a self-supervised learning loss called DILEMMA, which enhances the shape discrimination ability of visual representations by detecting positional errors in image patches.
Representation with Incomplete Votes
Daniel Halpern (Harvard University), Manuel Wüthrich (Harvard University)
OptimizationRepresentation LearningText
🎯 What it does: This paper studies the problem of handling incomplete voting in committee elections on online citizen participation platforms and proposes an adaptive query algorithm to achieve fair representation.
RESDSQL: Decoupling Schema Linking and Skeleton Parsing for Text-to-SQL
Haoyang Li (Renmin University of China), Hong Chen (Renmin University of China)
GenerationOptimizationTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes a Text-to-SQL framework called RESDSQL that separates pattern matching from syntax parsing. It first ranks and filters database schemas using a cross-encoder, and then generates an SQL skeleton followed by filling in specific fields using a seq2seq model, significantly improving the understanding and generation of complex queries.
Resilient Binary Neural Network
Sheng Xu (Beihang University), Jinhu Lu
ClassificationObject DetectionConvolutional Neural NetworkImageText
🎯 What it does: This paper proposes the Resilient Binary Neural Network (ReBNN), which suppresses weight oscillation in binary networks through parameterized scaling factors and weighted reconstruction loss, thereby enhancing training stability and performance.
Resolving Task Confusion in Dynamic Expansion Architectures for Class Incremental Learning
Bingchen Huang (Fudan University), Zuxuan Wu (Fudan University)
ClassificationKnowledge DistillationImage
🎯 What it does: This paper proposes a solution to the task confusion problem in category incremental learning under a dynamically scalable architecture, introducing the Task-Related Incremental Learning (TCIL) framework;
Resource Sharing through Multi-Round Matchings
Yohai Trabelsi (Bar-Ilan University), Daniel J. Rosenkrantz (University at Albany - State University of New York)
OptimizationTabular
🎯 What it does: This paper proposes a theoretical framework and algorithm for the multi-round matching (multi-resource sharing) problem, addressing situations where traditional matching fails to meet demands.
Restructuring Graph for Higher Homophily via Adaptive Spectral Clustering
Shouheng Li (Australian National University), Qing Wang
ClassificationOptimizationGraph Neural NetworkGraph
🎯 What it does: This paper proposes a graph reconstruction method based on adaptive spectral clustering, which reconstructs the adjacency matrix using learnable pseudo-feature weights, significantly improving the node classification performance of classical GNNs while maintaining the original structure of graphs with low homogeneity.
Rethinking Alignment and Uniformity in Unsupervised Image Semantic Segmentation
Daoan Zhang (Southern University of Science and Technology), Jianguo Zhang (Southern University of Science and Technology)
SegmentationConvolutional Neural NetworkTransformerImage
🎯 What it does: A new unsupervised image semantic segmentation framework called Semantic Attention Network (SAN) is proposed, which achieves dynamic alignment between pixels and semantics through a pixel-level encoder, a semantic generator, and a Semantic Attention (SEAT) module, using image reconstruction as supervision.
Rethinking Data Augmentation for Single-Source Domain Generalization in Medical Image Segmentation
Zixian Su (University of Liverpool), Jie Sun (Xi'an Jiaotong-Liverpool University)
SegmentationDomain AdaptationImageBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: Research on data augmentation methods for single-source domain generalization (SDG) in medical image segmentation
Rethinking Data-Free Quantization as a Zero-Sum Game
Biao Qian (Hefei University of Technology), Meng Wang (Hefei University of Technology)
Data SynthesisOptimizationConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: To address the problem of Data-Free Quantization (DFQ), a generative method called AdaSG based on sample adaptability is proposed. It utilizes a zero-sum game composed of a generator and a quantization network to produce synthetic samples beneficial to the quantization network, thereby restoring the performance of the quantized model without accessing real data.
Rethinking Disparity: A Depth Range Free Multi-View Stereo Based on Disparity
Qingsong Yan (Wuhan University), Fei Deng (Wuhan University)
Depth EstimationRecurrent Neural NetworkOptical FlowImage
🎯 What it does: A new multi-view stereo matching method called DispMVS is proposed, which constructs a 2D cost volume on the image plane using disparity flow (E-flow) and iteratively updates depth through multi-view geometry without the need for a predefined depth range.
Rethinking Interpretation: Input-Agnostic Saliency Mapping of Deep Visual Classifiers
Naveed Akhtar (University of Western Australia), Mohammad Amir Asim Khan Jalwana (University of Western Australia)
Explainability and InterpretabilityAdversarial AttackConvolutional Neural NetworkImage
🎯 What it does: This paper proposes an input-agnostic saliency mapping method that utilizes model gradient accumulation and projection to generate visualizations, revealing the model's geometric understanding of concepts.
Rethinking Rotation Invariance with Point Cloud Registration
Jianhui Yu (University of Sydney), Weidong Cai (University of Sydney)
ClassificationRetrievalTransformerPoint Cloud
🎯 What it does: This paper proposes a rotation-invariant learning framework based on point cloud registration.