AAAI 2024 Papers — Page 16
AAAI Conference on Artificial Intelligence · 2331 papers
On the Unstable Convergence Regime of Gradient Descent
Shuo Chen (Beijing Jiaotong University), Yao Zhao (Beijing Jiaotong University)
Optimization
🎯 What it does: This paper clarifies the phenomenon of 'unstable convergence' in gradient descent (GD) when the learning rate exceeds 2/L (i.e., the traditional convergence threshold) through theoretical analysis. It proves the existence of a forward-invariant open set U under any twice-differentiable function, such that initialization outside of U will first exhibit oscillations, then enter U and achieve monotonic convergence. An openness theorem regarding stable and unstable initialization is also provided, along with numerical experiments on several two-dimensional functions for validation.
On Unsupervised Domain Adaptation: Pseudo Label Guided Mixup for Adversarial Prompt Tuning
Fanshuang Kong (Beihang University), Yongyi Mao (University of Ottawa)
Domain AdaptationTransformerPrompt EngineeringText
🎯 What it does: This paper proposes a pseudo-label guided Mixup method (PL-Mix), which is combined with adversarial prompt tuning for unsupervised domain adaptation.
Once and for All: Universal Transferable Adversarial Perturbation against Deep Hashing-Based Facial Image Retrieval
Long Tang (Wuhan University), Yunming Zhang (Wuhan University)
RetrievalSafty and PrivacyAdversarial AttackMeta LearningConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: A universal transferable adversarial perturbation (UTAP) is proposed to prevent deep hashing-based facial image retrieval systems from leaking privacy.
One at a Time: Progressive Multi-Step Volumetric Probability Learning for Reliable 3D Scene Perception
Bohan Li (Shanghai Jiao Tong University), Wenjun Zeng (Shanghai Jiao Tong University)
SegmentationDepth EstimationConvolutional Neural NetworkDiffusion modelPoint Cloud
🎯 What it does: A voxel probability learning framework based on multi-step diffusion, VPD, is proposed to enhance the accuracy and robustness of multi-view stereo and semantic scene completion.
One Self-Configurable Model to Solve Many Abstract Visual Reasoning Problems
Mikołaj Małkiński (Warsaw University of Technology), Jacek Mańdziuk (AGH University of Krakow)
ClassificationRecognitionConvolutional Neural NetworkImage
🎯 What it does: A self-configuring unified model SCAR is proposed, capable of solving various single-choice abstract visual reasoning (AVR) problems without pre-assuming the task structure.
One Step Closer to Unbiased Aleatoric Uncertainty Estimation
Wang Zhang (Massachusetts Institute of Technology), Lam M. Nguyen (IBM Research)
ImageTabularOrdinary Differential Equation
🎯 What it does: A new unbiased data uncertainty (aleatoric) estimation method called Denoising Variance Attenuation (DVA) is proposed and validated, which approaches the true noise level through active denoising and normalized gradient descent on noise.
One Step Learning, One Step Review
Xiaolong Huang (Chongqing University of Technology), Xuesong Gao (Chongqing University of Technology)
ClassificationObject DetectionSegmentationDomain AdaptationConvolutional Neural NetworkTransformerSupervised Fine-TuningImage
🎯 What it does: A new fine-tuning method called OLOR is proposed, which combines weight rollback and hierarchical penalties to alleviate knowledge forgetting and improve performance on downstream visual tasks.
One-Step Forward and Backtrack: Overcoming Zig-Zagging in Loss-Aware Quantization Training
Lianbo Ma (Northeastern University), Qing Li (Peng Cheng Laboratory)
CompressionOptimizationConvolutional Neural NetworkImage
🎯 What it does: A one-forward and one-backward tracking quantization method (BLAQ) is proposed, which addresses the gradient 'sawtooth' problem and improves convergence speed by first performing a forward probing gradient at each step, followed by a backward update of the quantized gradient, thus solving issues present in traditional loss-aware quantization methods.
Online Boosting Adaptive Learning under Concept Drift for Multistream Classification
En Yu (Australian Artificial Intelligence Institute), Guangquan Zhang (Australian Artificial Intelligence Institute)
ClassificationDomain AdaptationTime Series
🎯 What it does: An online adaptive learning framework for multi-source data stream classification (OBAL) is proposed, and the AdaCOSA algorithm is designed to dynamically learn the covariance alignment and correlation between source streams and target streams.
Online Conversion Rate Prediction via Multi-Interval Screening and Synthesizing under Delayed Feedback
Qiming Liu (Chinese Academy of Sciences), Qing He (Chinese Academy of Sciences)
Recommendation SystemOptimizationTabularFinance Related
🎯 What it does: The MISS model is proposed, which achieves online CVR prediction through multi-interval selection and a lightweight synthesis framework, addressing the label bias problem caused by delayed feedback.
Online Markov Decision Processes Configuration with Continuous Decision Space
Davide Maran (Politecnico di Milano), Marcello Restelli (Politecnico di Milano)
OptimizationReinforcement LearningTabular
🎯 What it does: This paper studies the optimal methods for online configuration of episodic Markov decision processes in continuous decision spaces, with a particular focus on the interaction between the learner (configurator) and an agent with a fixed unknown policy.
Online Restless Multi-Armed Bandits with Long-Term Fairness Constraints
Shufan Wang, Jian Li (Stony Brook University)
OptimizationReinforcement LearningTabular
🎯 What it does: A RMAB model with long-term fairness constraints (RMAB-F) is proposed, and an online reinforcement learning algorithm Fair-UCRL is designed to achieve long-term reward maximization while satisfying fairness constraints under an unknown Markov decision process.
Online Sensitivity Optimization in Differentially Private Learning
Filippo Galli (Scuola Normale Superiore), Tommaso Cucinotta (Scuola Superiore Sant'Anna)
OptimizationSafty and PrivacyImageText
🎯 What it does: A method for dynamically optimizing the clipping threshold in differential privacy learning is proposed to improve the efficiency of model training and privacy protection.
Online Submodular Maximization via Online Convex Optimization
Tareq Si Salem (Northeastern University), Stratis Ioannidis (Northeastern University)
OptimizationTabular
🎯 What it does: This paper proposes a method to transform Online Submodular Optimization (OSM) into Online Convex Optimization (OCO), and achieves online submodular maximization problems based on WTP functions (including various applications such as influence maximization and facility location) under arbitrary base matrix constraints through randomized rounding and convex relaxation.
OntoFact: Unveiling Fantastic Fact-Skeleton of LLMs via Ontology-Driven Reinforcement Learning
Ziyu Shang (Southeast University), Ke Ji (Southeast University)
OptimizationReinforcement LearningGraph
🎯 What it does: The OntoFact framework is proposed, which automatically generates high-error-rate test cases based on ontology reinforcement learning to evaluate the factual omissions of LLMs.
Open-Set Facial Expression Recognition
Yuhang Zhang (Beijing University of Posts and Telecommunications), Weihong Deng (Beijing University of Posts and Telecommunications)
RecognitionConvolutional Neural NetworkImage
🎯 What it does: This paper proposes the Open-Set Facial Expression Recognition (Open-Set FER) task and designs a complete training framework that transforms it into a noise label detection problem, achieving efficient recognition of unknown expressions through pseudo-labels, attention map consistency, and cyclic training.
Open-Set Graph Domain Adaptation via Separate Domain Alignment
Yu Wang (LinkedIn Corporation), Sheng Li (University of Virginia)
Domain AdaptationGraph Neural NetworkGraph
🎯 What it does: This paper proposes the Open Set Graph Domain Adaptation (OS-GDA) task and designs a Separate Domain Alignment (SDA) framework to align known class nodes and cluster unknown class nodes simultaneously.
Open-Vocabulary Video Relation Extraction
Wentao Tian (Fudan University), Lechao Cheng (Zhejiang Lab)
GenerationRetrievalTransformerLarge Language ModelVision Language ModelVideoText
🎯 What it does: This paper proposes the Open Vocabulary Video Relation Extraction task (OVRE) and constructs the Moments-OVRE dataset, which contains 180K videos, aimed at generating all relation triples related to video actions.
Operationalizing Essential Characteristics of Creativity in a Computational System for Music Composition
Paul M. Bodily (Idaho State University), Dan Ventura (Brigham Young University)
GenerationTextAudio
🎯 What it does: A self-composing music creation system named Pop* has been designed and implemented, capable of generating complete pop music scores and audio based on social media tweets, operationalizing seven creative features: generation, knowledge representation, intention, aesthetics, domain knowledge, autonomy, and self-assessment.
Operator-Learning-Inspired Modeling of Neural Ordinary Differential Equations
Woojin Cho (Yonsei University), Noseong Park (Yonsei University)
ClassificationGenerationData SynthesisRecurrent Neural NetworkImageTime SeriesOrdinary Differential Equation
🎯 What it does: A new architecture for neural ordinary differential equations (NODE) based on branch Fourier neural operators (BFNO) is proposed to improve the expression of ODE functions.
Opponent-Model Search in Games with Incomplete Information
Junkang Li (NukkAI), Véronique Ventos (Normandie University)
OptimizationReinforcement Learning from Human Feedback
🎯 What it does: This paper proposes an opponent modeling search algorithm for two-player zero-sum games with incomplete information, capable of solving max-min strategies under the assumption of known opponent models.
Optical Flow for Spike Camera with Hierarchical Spatial-Temporal Spike Fusion
Rui Zhao (Peking University), Tiejun Huang (Peking University)
Spiking Neural NetworkOptical FlowVideo
🎯 What it does: This paper studies a method for optical flow estimation suitable for neuromorphic 'spike cameras' and proposes the HiST-SFlow network to achieve high-precision pixel-level motion matching.
Optimal Attack and Defense for Reinforcement Learning
Jeremy McMahan (University of Wisconsin-Madison), Qiaomin Xie (University of Wisconsin-Madison)
OptimizationAdversarial AttackReinforcement LearningTabular
🎯 What it does: In the reinforcement learning environment, researchers proposed a framework for attack and defense against full online manipulation attacks where the attacker can manipulate states, observations, actions, and rewards.
Optimal Makespan in a Minute Timespan! A Scalable Multi-Robot Goal Assignment Algorithm for Minimizing Mission Time
Aakash (Indian Institute of Technology Kanpur), Indranil Saha (Indian Institute of Technology Kanpur)
OptimizationRobotic IntelligenceSimultaneous Localization and MappingTabular
🎯 What it does: This paper proposes a scalable multi-robot task allocation algorithm, OM, which minimizes the makespan by solving only the necessary robot-target pairs through heuristic estimation and lazy path computation.
Optimal Mechanism in a Dynamic Stochastic Knapsack Environment
Jihyeok Jung (Seoul National University), Kiho Yoon (Korea University)
OptimizationReinforcement Learning
🎯 What it does: This study investigates the design of an optimal mechanism to maximize the expected revenue of sellers in a dynamic stochastic knapsack environment.
Optimal Quasi-clique: Hardness, Equivalence with Densest-k-Subgraph, and Quasi-partitioned Community Mining
Aritra Konar (KU Leuven), Nicholas D. Sidiropoulos (University of Virginia)
OptimizationGraph Neural NetworkGraph
🎯 What it does: This paper studies the Optimal Quasi-Clique (OQC) problem, proving that it is NP-hard on undirected unweighted graphs, and reveals its equivalence to the classic Densest-k-Subgraph (DkS) problem on the size-density frontier.
Optimal Survival Trees: A Dynamic Programming Approach
Tim Huisman (Delft University of Technology), Emir Demirović (Delft University of Technology)
OptimizationTabular
🎯 What it does: An optimal survival tree algorithm called SurTree based on dynamic programming is proposed, which can obtain a global optimal solution under the constraints of given tree depth and number of branching nodes, and significantly improves scalability through a specialized depth-two precomputation algorithm.
Optimal Transport with Cyclic Symmetry
Shoichiro Takeda (NTT Corporation), Kenta Niwa (NTT Corporation)
OptimizationComputational EfficiencyImage
🎯 What it does: A fast algorithm for optimal transport (OT) utilizing the cyclic symmetry structure of input data is proposed, focusing mainly on two important forms of OT: linear programming OT (LOT) and strongly convex regularized OT (SROT).
Optimal Transport with Tempered Exponential Measures
Ehsan Amid (Google DeepMind), Manfred K. Warmuth (Google Research)
OptimizationTabular
🎯 What it does: This study investigates the temperature regularization of optimal transport using the Temperature Exponential Measure (TEM) and proposes a scalable OT method that can achieve sparse solutions.
Optimised Storage for Datalog Reasoning
Xinyue Zhang (University of Oxford), Ian Horrocks (University of Oxford)
OptimizationComputational EfficiencyGraphTabularBenchmark
🎯 What it does: This paper proposes a multi-scheme framework that integrates custom storage structures (such as compressed storage for transitive closure and union rules) with standard materialization algorithms, achieving efficient storage and querying for Datalog reasoning.
Optimistic Model Rollouts for Pessimistic Offline Policy Optimization
Yuanzhao Zhai (National University of Defense Technology), Huaimin Wang (National University of Defense Technology)
OptimizationReinforcement LearningTabularBenchmark
🎯 What it does: A framework named ORPO is proposed, which utilizes optimistic model rolling (O-MDP) to generate richer offline trajectories, and then performs policy optimization in a conservative P-MDP to enhance the generalization and performance of offline reinforcement learning.
Optimistic Policy Gradient in Multi-Player Markov Games with a Single Controller: Convergence beyond the Minty Property
Ioannis Anagnostides (Carnegie Mellon University), Tuomas Sandholm (Carnegie Mellon University)
OptimizationReinforcement Learning
🎯 What it does: This paper proposes a new framework for describing optimistic policy gradient methods in multi-player Markov games with a single controller, proving convergence to a static ε-Nash equilibrium in O(1/ϵ²) iterations, particularly under the assumption of equilibrium collapse.
Optimistic Value Instructors for Cooperative Multi-Agent Reinforcement Learning
Chao Li (Nanjing University), Yang Gao (Nanjing University)
Reinforcement Learning
🎯 What it does: An optimized 'Optimistic Value Instructor' (OVI) algorithm is proposed in multi-agent reinforcement learning, which improves value decomposition learning to overcome the issue of relative overgeneralization by learning each agent's optimistic value function and incorporating two types of guidance constraints, thereby enhancing cooperative performance.
Optimize & Reduce: A Top-Down Approach for Image Vectorization
Or Hirschorn (Tel Aviv University), Shai Avidan (Tel Aviv University)
GenerationOptimizationDiffusion modelImage
🎯 What it does: A top-down iterative method based on DiffVG (Optimize & Reduce) is proposed to achieve image vectorization, aiming to reconstruct the input image with a minimal number of Bézier curves while ensuring editability and high reconstruction quality.
Optimizing ADMM and Over-Relaxed ADMM Parameters for Linear Quadratic Problems
Jintao Song (University of Birmingham), Jinming Duan (University of Birmingham)
OptimizationImageMagnetic Resonance Imaging
🎯 What it does: This paper proposes a method for automatically selecting the penalty parameter and relaxation parameter for ADMM and its over-relaxed variant in linear quadratic problems, significantly improving convergence speed.
Optimizing Local Satisfaction of Long-Run Average Objectives in Markov Decision Processes
David Klaška (Masaryk University), Vojtěch Řehák (Masaryk University)
OptimizationReinforcement LearningGraph
🎯 What it does: This paper presents the theory and algorithms for the local satisfaction optimization problem targeting long-term average objectives in Markov Decision Processes (MDP), and provides two practical algorithms (LocalEval for evaluation and LocalSynt for policy synthesis).
Optimizing the Optimization of Planning Domains by Automatic Action Schema Splitting
Mojtaba Elahi (Aalto University), Jussi Rintanen (Aalto University)
Optimization
🎯 What it does: An improved action splitting method in the planning domain automatically decomposes macro actions into smaller micro actions during the preprocessing phase by utilizing domain invariants, priority orders, and other information, significantly reducing the number of generated ground actions and alleviating preprocessing bottlenecks.
Orthogonal Dictionary Guided Shape Completion Network for Point Cloud
Pingping Cai (University of South Carolina), Song Wang (University of South Carolina)
RestorationGenerationAutonomous DrivingTransformerPoint Cloud
🎯 What it does: A novel Orthogonal Dictionary Guided Shape Completion Network (ODGNet) is proposed, achieving high-quality reconstruction of missing areas in point clouds through an improved Seed Generation U-Net and a learnable orthogonal shape dictionary.
OSFFNet: Omni-Stage Feature Fusion Network for Lightweight Image Super-Resolution
Yang Wang (Jiangnan University), Tao Zhang (Jiangnan University)
RestorationSuper ResolutionKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: A lightweight image super-resolution network OSFFNet is proposed, which enhances SR quality by fusing multi-level features.
Out of Thin Air: Exploring Data-Free Adversarial Robustness Distillation
Yuzheng Wang (Fudan University), Lizhe Qi (Fudan University)
Knowledge DistillationAdversarial AttackGenerative Adversarial NetworkImage
🎯 What it does: A method for data-free adversarial robustness distillation (DFARD) is studied, which allows small models to achieve adversarial robustness without the need for original data.
Out-of-Distribution Detection in Long-Tailed Recognition with Calibrated Outlier Class Learning
Wenjun Miao (Beihang University), Jin Zheng (Beihang University)
ClassificationRecognitionAnomaly DetectionConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: A method for OOD detection called COCL is proposed and implemented for long-tail recognition scenarios, addressing the confusion between head class and tail class samples with OOD samples.
OVD-Explorer: Optimism Should Not Be the Sole Pursuit of Exploration in Noisy Environments
Jinyi Liu (Tianjin University), Yang Sun (Northwestern Polytechnical University)
Robotic IntelligenceReinforcement Learning
🎯 What it does: A noise-aware optimistic exploration method for continuous control tasks is proposed—OVD-Explorer;
OWQ: Outlier-Aware Weight Quantization for Efficient Fine-Tuning and Inference of Large Language Models
Changhun Lee (POSTECH), Eunhyeok Park (POSTECH)
OptimizationComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: A mixed-precision weight quantization method based on activation outlier awareness, called OWQ, is proposed, along with a fine-tuning technique called WCT that only updates high-precision 'weak columns' for efficient inference and task adaptation in large language models.
p-Laplacian Adaptation for Generative Pre-trained Vision-Language Models
Haoyuan Wu (Chinese University of Hong Kong), Bei Yu (Chinese University of Hong Kong)
Graph Neural NetworkTransformerVision Language ModelImageMultimodality
🎯 What it does: This paper proposes a p-Laplacian-based adapter (p-adapter) that maps adapter tuning to graph information transmission on attention maps to address the issue of heterogeneous attention maps.
PA2D-MORL: Pareto Ascent Directional Decomposition Based Multi-Objective Reinforcement Learning
Tianmeng Hu (Central South University), Biao Luo (Central South University)
OptimizationReinforcement LearningSequential
🎯 What it does: A multi-strategy multi-objective reinforcement learning method PA2D-MORL is proposed, which utilizes Pareto ascent direction decomposition and an evolutionary framework to achieve Pareto strategy set approximation.
PAC-Bayes Generalisation Bounds for Dynamical Systems including Stable RNNs
Deividas Eringis (Aalborg University), Mihály Petreczky (University of Lille)
Recurrent Neural NetworkTime Series
🎯 What it does: This paper derives a set of PAC-Bayes generalization error bounds for predictors that can be viewed as discrete-time nonlinear dynamical systems (including stable recurrent neural networks) within a supervised time series learning framework, applicable to non-i.i.d. data and not growing with the prediction horizon.
Painterly Image Harmonization by Learning from Painterly Objects
Li Niu (Shanghai Jiao Tong University), Liqing Zhang (Shanghai Jiao Tong University)
Image TranslationImage HarmonizationAuto EncoderImage
🎯 What it does: This paper proposes a method for image harmonization based on learning painter styles, utilizing target objects in paintings to infer and synthesize the ideal style of the photographed object in the composite image, thereby achieving seamless integration of the painter's style background with the photographic subject.
PaintHuman: Towards High-Fidelity Text-to-3D Human Texturing via Denoised Score Distillation
Jianhui Yu (University of Sydney), Wayne Wu (University of Sydney)
GenerationData SynthesisDiffusion modelScore-based ModelMesh
🎯 What it does: This paper studies a zero-shot text-driven human 3D model texturing method called PaintHuman.
Pairwise-Label-Based Deep Incremental Hashing with Simultaneous Code Expansion
Dayan Wu (Institute of Information Engineering, Chinese Academy of Sciences), Weiping Wang (Institute of Information Engineering, Chinese Academy of Sciences)
RetrievalConvolutional Neural NetworkImage
🎯 What it does: A joint deep incremental hashing framework CEDIH is proposed, which can learn new categories and extend the hash code length without regenerating the original database codes.
Pandora’s Problem with Deadlines
Ben Berger (Offchain Labs), Federico Fusco (Sapienza University of Rome)
Optimization
🎯 What it does: Proposed a variant of the Pandora problem with Deadline and Time Slots, and provided an efficient threshold strategy that achieves a constant approximation solution.
Pano-NeRF: Synthesizing High Dynamic Range Novel Views with Geometry from Sparse Low Dynamic Range Panoramic Images
Zhan Lu (Nanyang Technological University), Xudong Jiang (Nanyang Technological University)
RestorationGenerationData SynthesisNeural Radiance FieldImage
🎯 What it does: A new method called Pano-NeRF is proposed, which synthesizes high dynamic range new views from sparse low dynamic range panoramic images, utilizing geometric information for reconstruction.
Panoptic Scene Graph Generation with Semantics-Prototype Learning
Li Li (National University of Singapore), Roger Zimmermann (National University of Singapore)
Object DetectionSegmentationGenerationDomain AdaptationGraph Neural NetworkTransformerContrastive LearningGraph
🎯 What it does: An adaptive data transmission framework ADTrans was designed and implemented to correct annotation biases in the Panoptic Scene Graph Generation dataset, enhancing the model's relationship prediction performance.
Pantypes: Diverse Representatives for Self-Explainable Models
Rune Kjærsgaard (Technical University of Denmark), Line Clemmensen (Physikalisch Technische Bundesanstalt)
Explainability and InterpretabilityAuto EncoderImage
🎯 What it does: This paper proposes Pantypes, which learns sparse diverse prototypes in the ProtoVAE framework using volume loss based on deterministic point processes, and implements dynamic pruning of prototypes.
ParaGuide: Guided Diffusion Paraphrasers for Plug-and-Play Textual Style Transfer
Zachary Horvitz (Columbia University), Kathleen McKeown (Columbia University)
GenerationData SynthesisTransformerDiffusion modelText
🎯 What it does: This paper studies an unsupervised text style transfer framework called ParaGuide based on diffusion models, which can achieve various style transfers during inference through gradient guidance.
Parallel Beam Search Algorithms for Domain-Independent Dynamic Programming
Ryo Kuroiwa (University of Toronto), J. Christopher Beck (University of Toronto)
OptimizationTabular
🎯 What it does: Three parallel beam search algorithms (shared beam search, hash distributed beam search 1 and 2) are proposed for domain-independent dynamic programming (DIDP), and a multi-threaded CABS solver is implemented, significantly improving the solving speed.
Parallel Empirical Evaluations: Resilience despite Concurrency
Johannes K. Fichte (Linköping University), Matthias Schlögel (TU Wien)
OptimizationTabular
🎯 What it does: This paper proposes a parallel experimental method based on memory hierarchy and cache partitioning to improve the reproducibility and stability of combinatorial optimization experiments.
Parallel Ranking of Ads and Creatives in Real-Time Advertising Systems
Zhiguang Yang (JD.com), Jingping Shao (JD.com)
Recommendation SystemOptimizationTransformerTabular
🎯 What it does: Proposed and implemented the Peri-CR parallel advertising and creative ranking architecture, and developed the offline joint optimization model JAC to enhance CTR prediction and creative selection.
Parallel Vertex Diffusion for Unified Visual Grounding
Zesen Cheng (Peking University), Jie Chen (Peking University)
Object DetectionSegmentationTransformerDiffusion modelImage
🎯 What it does: This paper proposes a Parallel Vertex Diffusion (PVD) method for unified vertex generation in visual localization tasks.
Parameterization of (Partial) Maximum Satisfiability above Matching in a Variable-Clause Graph
Vasily Alferov (Independent Researcher), Kirill Brilliantov (St. Petersburg Department of Steklov Mathematical Institute of the Russian Academy of Sciences)
Optimization
🎯 What it does: The algorithm for maximum satisfiability problem (MAXSAT) and partial maximum satisfiability problem (Partial MaxSAT) based on variable-clause graph maximum matching has been improved parametrically, achieving a better runtime of O*(2^{3k/2}), and this improvement has been extended to (n,3) and (n,4) MAXSAT.
Parameterized Approximation Algorithms for Sum of Radii Clustering and Variants
Xianrun Chen (Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences), Yong Zhang (Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences)
Optimization
🎯 What it does: This paper proposes a unified Fixed Parameter Tractable approximation (FPT-approximation) framework for solving the Sum of Radii (SoR) minimization problem and its fair (FairSoR) and matroid (MatroidSoR) variants.
Parameterized Projected Bellman Operator
Théo Vincent (German Research Center for AI), Carlo D'Eramo (Politecnico di Milano)
Reinforcement Learning
🎯 What it does: Proposed and learned the Projected Bellman Operator (PBO), replacing the traditional empirical Bellman operator to eliminate the projection step and reduce dependence on sampling; based on this, implemented two algorithms: ProFQI and ProDQN.
Pareto Front-Diverse Batch Multi-Objective Bayesian Optimization
Alaleh Ahmadianshalchi (Washington State University), Janardhan Rao Doppa (Washington State University)
OptimizationTabular
🎯 What it does: This paper proposes a new batch multi-objective Bayesian optimization method (PDBO), which adaptively selects acquisition functions and utilizes Determinantal Point Process (DPP) to achieve diversified search of the Pareto front.
PARSAC: Accelerating Robust Multi-Model Fitting with Parallel Sample Consensus
Florian Kluger (Leibniz University Hannover), Bodo Rosenhahn (Leibniz University Hannover)
OptimizationComputational EfficiencyConvolutional Neural NetworkReinforcement LearningImage
🎯 What it does: The PARSAC method is proposed, which uses deep networks to predict multiple sets of sample weights and inlier weights in one go, achieving multi-model geometric fitting within a parallel RANSAC framework.
Parsing All Adverse Scenes: Severity-Aware Semantic Segmentation with Mask-Enhanced Cross-Domain Consistency
Fuhao Li (Wuhan University of Science and Technology), Hong Zhang (Shanghai AI Lab)
SegmentationDomain AdaptationAutonomous DrivingImage
🎯 What it does: A unified model called PASS is proposed, which achieves cross-domain consistent semantic segmentation by utilizing severity awareness and style enhancement, covering all adverse scenarios such as fog, night, rain, and snow.
Partial Label Learning with a Partner
Chongjie Si (Shanghai Jiao Tong University), Wei Shen (Shanghai Jiao Tong University)
ClassificationImage
🎯 What it does: Introduce partner classifiers and mutual supervision mechanisms in Partial Label Learning (PLL) to correct erroneous labels and enhance de-fogging capabilities.
Partial Multi-View Clustering via Self-Supervised Network
Wei Feng (Xi'an Jiaotong University), Bo Dong (Xi'an Jiaotong University)
Representation LearningContrastive LearningImageText
🎯 What it does: A self-supervised network for partial multi-view clustering, PVC-SSN, is proposed, which jointly learns a unified and discriminative subspace representation through a multi-view contrastive encoder, self-expression layer, and decoder.
Participation Incentives in Approval-Based Committee Elections
Martin Bullinger (University of Oxford), Clara Mehler (Technical University of Munich)
🎯 What it does: This study proves that the ABC scoring rule satisfies group participation in approval-based committee elections, while most sequential rules severely violate participation; it also proposes and proves several avoidance strategies, demonstrating that determining whether abstaining from voting is beneficial for certain sequential rules is NP-hard.
Patch-Aware Sample Selection for Efficient Masked Image Modeling
Zhengyang Zhuge (Institute of Automation, Chinese Academy of Sciences), Jian Cheng (Institute of Automation, Chinese Academy of Sciences)
RecognitionObject DetectionSegmentationComputational EfficiencyTransformerImage
🎯 What it does: This paper proposes an efficient Masked Image Modeling pre-training method based on Patch-Aware Sample Selection (PASS), which significantly reduces the required number of training samples while maintaining or improving performance.
Patch-Wise Graph Contrastive Learning for Image Translation
Chanyong Jung (KAIST), Jong Chul Ye (KAIST)
Image TranslationGraph Neural NetworkGenerative Adversarial NetworkContrastive LearningImage
🎯 What it does: A patch-level graph contrastive learning framework based on graph neural networks is proposed to enhance the semantic correspondence between inputs and outputs in image translation tasks.
patchDPCC: A Patchwise Deep Compression Framework for Dynamic Point Clouds
Zirui Pan (Shandong University), Yao Liu (Rutgers University)
CompressionAuto EncoderPoint Cloud
🎯 What it does: For dynamic point cloud compression, a patchDPCC framework is proposed, which first divides the point cloud frames into patch groups with a fixed number of points, and then compresses them using a point-based deep network.
Patched Line Segment Learning for Vector Road Mapping
Jiakun Xu (Wuhan University), Nan Xue (Ant Group)
SegmentationAutonomous DrivingComputational EfficiencyConvolutional Neural NetworkAuto EncoderImage
🎯 What it does: A novel vector road mapping method is proposed by learning the Patch Line Segment (PaLiS) representation of local segments;
PathAsst: A Generative Foundation AI Assistant towards Artificial General Intelligence of Pathology
Yuxuan Sun (Zhejiang University), Lin Yang (Westlake University)
GenerationRetrievalTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelContrastive LearningMultimodalityBiomedical Data
🎯 What it does: PathAsst has been developed, a multimodal generative AI assistant for pathology, which integrates a dedicated visual encoder PathCLIP, the Vicuna-13B language model, as well as 8 specialized sub-models and a literature retrieval tool;
Paths, Proofs, and Perfection: Developing a Human-Interpretable Proof System for Constrained Shortest Paths
Konstantin Sidorov (Delft University of Technology), Emir Demirović (Delft University of Technology)
OptimizationExplainability and InterpretabilityGraph Neural NetworkGraph
🎯 What it does: A proof system for constrained shortest paths is proposed, and optimality proofs are constructed through specific reasoning rules in graph theory and A* search.
Pay Attention to Target: Relation-Aware Temporal Consistency for Domain Adaptive Video Semantic Segmentation
Huayu Mai (University of Science and Technology of China), Feng Wu (University of Science and Technology of China)
SegmentationDomain AdaptationOptical FlowVideo
🎯 What it does: A pseudo-label based domain adaptive video semantic segmentation framework PAT is proposed, combining target domain focalization and relation-aware temporal consistency.
Pay to (Not) Play: Monetizing Impatience in Mobile Games
Taylor Lundy (University of British Columbia), Kevin Leyton-Brown (Shanghai University of Finance and Economics)
🎯 What it does: This paper studies the pricing problem of using 'skip waiting time' to implement microtransactions in mobile games, proposing a value model based on players' perception of task difficulty, and modeling three types of players (completely sensitive, price sensitive, and insensitive).
PC-Conv: Unifying Homophily and Heterophily with Two-Fold Filtering
Bingheng Li (University of Electronic Science and Technology of China), Zhao Kang (University of Electronic Science and Technology of China)
ClassificationGraph Neural NetworkGraph
🎯 What it does: A two-layer filtering mechanism is proposed, combining heterogeneous graph heat kernels and low-pass filters to construct PC-Conv convolution and PCNet networks, achieving simultaneous processing of node classification for homogeneous and heterogeneous graphs.
PCE-Palm: Palm Crease Energy Based Two-Stage Realistic Pseudo-Palmprint Generation
Jianlong Jin (Hefei University of Technology), Wei Jia (Hefei University of Technology)
RecognitionGenerationData SynthesisGenerative Adversarial NetworkImage
🎯 What it does: A two-stage pseudo-palmprint generation method based on palmprint crease energy (PCE) is proposed to address the issue of large-scale data scarcity in palmprint recognition.
PDE+: Enhancing Generalization via PDE with Adaptive Distributional Diffusion
Yige Yuan (Chinese Academy of Sciences), Xueqi Cheng (Chinese Academy of Sciences)
ClassificationDomain AdaptationConvolutional Neural NetworkImageStochastic Differential EquationOrdinary Differential Equation
🎯 What it does: A generalized framework PDE+ based on partial differential equations (PDE) is proposed, which directly enhances the smoothness of neural network functions by introducing adaptive distribution diffusion (ADD) into the transport equation, thereby improving generalization ability to unseen distributions.
Peer Learning: Learning Complex Policies in Groups from Scratch via Action Recommendations
Cedric Derstroff (Technische Universitat Darmstadt), Stefan Kramer (Johannes Gutenberg Universität Mainz)
Robotic IntelligenceReinforcement LearningAgentic AI
🎯 What it does: A 'Peer Learning' framework is proposed, allowing a group of reinforcement learning agents to communicate with each other through action suggestions in independent environments, learning complex strategies from scratch.
Peer Neighborhood Mechanisms: A Framework for Mechanism Generalization
Adam Richardson (École Polytechnique Fédérale de Lausanne), Boi Faltings (École Polytechnique Fédérale de Lausanne)
TabularSequential
🎯 What it does: This paper proposes a framework called Peer Neighborhood Mechanisms to extend traditional Peer Prediction mechanisms from discrete distributions to any continuous distribution, and provides a specific implementation and proof of Peer Truth Neighborhood Extension (PTNE).
Percentile Risk-Constrained Budget Pacing for Guaranteed Display Advertising in Online Optimization
Liang Dai (Alibaba Group), Bo Zheng (Alibaba Group)
OptimizationTabular
🎯 What it does: Designed the RCPacing framework to ensure budget rhythm control and optimization for display advertising.
PerFedRLNAS: One-for-All Personalized Federated Neural Architecture Search
Dixi Yao (University of Toronto), Baochun Li (University of Toronto)
Federated LearningNeural Architecture SearchConvolutional Neural NetworkTransformerReinforcement LearningImage
🎯 What it does: Designed and implemented PerFedRLNAS, a framework that automatically searches for personalized model structures and weights for each client in a federated learning environment through reinforcement learning.
Performative Federated Learning: A Solution to Model-Dependent and Heterogeneous Distribution Shifts
Kun Jin (University of Michigan), Mingyan Liu (University of Michigan)
OptimizationFederated LearningImageTabular
🎯 What it does: Proposes an executable federated learning framework, studies the impact of model-driven distribution migration in federated learning, and provides corresponding PS (stable) and PO (optimal) solutions.
Permutation-Based Hypothesis Testing for Neural Networks
Francesca Mandel (University of Pennsylvania), Ian Barnett (University of Pennsylvania)
Biomedical DataElectronic Health Records
🎯 What it does: A hypothesis testing framework based on permutations is proposed, utilizing the output of neural networks to evaluate the partial derivatives of input features to assess the association between features and outcomes, including tests for nonlinear associations and general associations.
Personalized LoRA for Human-Centered Text Understanding
You Zhang (Yunnan University), Xuejie Zhang (Yunnan University)
ClassificationRecommendation SystemTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: For the Human-Centric Text Understanding (HCTU) task, a Pluggable Personalized Low-Rank Adapter (PLoRA) is proposed, which achieves lightweight personalized fine-tuning of the model by combining User Knowledge Injection (PKI) with LoRA, and addresses the cold start (zero-shot/few-shot) problem using techniques such as PDropout and MIM.
Personalized Reinforcement Learning with a Budget of Policies
Dmitry Ivanov (Technion), Omer Ben-Porat (Technion)
OptimizationReinforcement LearningSequential
🎯 What it does: This paper studies how to achieve personalized decision-making under a limited strategy budget in areas with high regulatory costs, proposing a representative MDP (r-MDP) framework and designing two deep reinforcement learning algorithms.
Perturbation-Invariant Adversarial Training for Neural Ranking Models: Improving the Effectiveness-Robustness Trade-Off
Yu-An Liu (Chinese Academy of Sciences), Xueqi Cheng (Chinese Academy of Sciences)
RetrievalAdversarial AttackTransformerText
🎯 What it does: This paper proposes a perturbation-invariant adversarial training (PIAT) for neural retrieval models, which enhances the balance between model performance and robustness on both natural and adversarial samples by simultaneously optimizing natural ranking loss and boundary ranking loss during training.
PG-LBO: Enhancing High-Dimensional Bayesian Optimization with Pseudo-Label and Gaussian Process Guidance
Taicai Chen (Nanjing University), Yang Gao (Nanjing University)
OptimizationAuto EncoderTabular
🎯 What it does: The PG-LBO method is proposed, which improves the potential space construction of VAE-BO in high-dimensional structural optimization using pseudo-labels and GP guidance.
PHFormer: Multi-Fragment Assembly Using Proxy-Level Hybrid Transformer
Wenting Cui (Xi'an Jiaotong University), Shaoyi Du (Xi'an Jiaotong University)
Pose EstimationComputational EfficiencyGraph Neural NetworkTransformerPoint Cloud
🎯 What it does: This paper proposes PHFormer, a proxy-level hybrid Transformer framework designed to address the Fragment Assembly problem by automatically predicting the 6-DoF transformations of each fragment.
Phoneme Hallucinator: One-Shot Voice Conversion via Set Expansion
Siyuan Shan (University of North Carolina at Chapel Hill), Junier B. Oliva (University of North Carolina at Chapel Hill)
GenerationData SynthesisTransformerAuto EncoderGenerative Adversarial NetworkAudio
🎯 What it does: A one-click voice conversion method called Phoneme Hallucinator is proposed, which utilizes a small amount of target speaker audio to generate diverse speaker features, thereby achieving one-shot VC.
PICNN: A Pathway towards Interpretable Convolutional Neural Networks
Wengang Guo (Tongji University), Wei Ye (Tongji University)
Explainability and InterpretabilityConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: Design a Pluggable Interpretation Path (PICNN) that allows standard CNNs to group the convolutional filters in the later layers into clusters corresponding to each category, thus making it directly interpretable while maintaining recognition ability.
Piecewise Linear Transformation – Propagating Aleatoric Uncertainty in Neural Networks
Thomas Krapf (University of Regensburg), Bernd Heinrich (University of Regensburg)
Convolutional Neural NetworkTabular
🎯 What it does: A new method called Piecewise Linear Transformation (PLT) is proposed, which can accurately propagate uncertainty to the outputs of neural networks with piecewise linear activation functions while preserving the complete dependencies of any input probability distribution.
Plug-In Diffusion Model for Sequential Recommendation
Haokai Ma (Shandong University), Zhanhui Kang (Tencent)
Recommendation SystemKnowledge DistillationRecurrent Neural NetworkTransformerDiffusion modelSequential
🎯 What it does: A pluggable extension framework PDRec is designed, which uses diffusion models as plugins to generate global user preferences, and combines historical behavior reweighting, positive sample augmentation, and noise-free negative sampling to enhance sequential recommendation performance.
PM-INR: Prior-Rich Multi-Modal Implicit Large-Scale Scene Neural Representation
Yiying Yang (Fudan University), Tao Chen (Fudan University)
GenerationData SynthesisNeural Radiance FieldImageTextMultimodalityPoint Cloud
🎯 What it does: This paper studies implicit neural representations of large-scale unbounded outdoor scenes and proposes the PM-INR framework, which utilizes multimodal priors from images, text, and 3D point clouds for extraction and fusion, and injects the fused priors into the sampling area to enhance the quality of novel view synthesis.
PMAC: Personalized Multi-Agent Communication
Xiangrui Meng (Peking University), Ying Tan (Peking University)
Robotic IntelligenceRecurrent Neural NetworkReinforcement LearningAgentic AISequential
🎯 What it does: The PMAC model is proposed to achieve personalized multi-agent communication with linear complexity;
PMET: Precise Model Editing in a Transformer
Xiaopeng Li (National University of Defense Technology), Jie Yu (National University of Defense Technology)
TransformerLarge Language ModelText
🎯 What it does: A precise model editing method called PMET is proposed, which utilizes the joint optimization of the multi-head self-attention (MHSA) and feed-forward network (FFN) hidden states in Transformers. It precisely updates the FFN weights using only the optimized FFN hidden states, achieving knowledge editing without changing the MHSA weights.
PMRC: Prompt-Based Machine Reading Comprehension for Few-Shot Named Entity Recognition
Jin Huang (Beijing University of Posts and Telecommunications), Yuanqiang Cai (Beijing University of Posts and Telecommunications)
RecognitionTransformerPrompt EngineeringText
🎯 What it does: A prompt-based machine reading comprehension (PMRC) model is proposed for few-shot named entity recognition, transforming the NER task into an MRC framework and inserting entity boundary markers in the input to predict all entities at once.
PNeRFLoc: Visual Localization with Point-Based Neural Radiance Fields
Boming Zhao (Zhejiang University), Zhaopeng Cui (Zhejiang University)
OptimizationNeural Radiance FieldSimultaneous Localization and MappingPoint Cloud
🎯 What it does: The PNeRFLoc framework is proposed, combining point-based NeRF with traditional 2D-3D matching to achieve a complete visual localization process from coarse localization to rendering optimization.
PNeSM: Arbitrary 3D Scene Stylization via Prompt-Based Neural Style Mapping
Jiafu Chen (Zhejiang University), Zhizhong Wang (Nanjing University of Science and Technology)
Image TranslationGenerationPrompt EngineeringNeural Radiance FieldImage
🎯 What it does: This paper proposes the PNeSM framework, which can transfer any style image to any 3D scene without the need to retrain for each scene, achieving visual stylization of 3D scenes.
PoetryDiffusion: Towards Joint Semantic and Metrical Manipulation in Poetry Generation
Zhiyuan Hu (National University of Singapore), Bryan Hooi (Nanyang Technological University)
GenerationDiffusion modelText
🎯 What it does: A poetry generation framework called PoetryDiffusion based on diffusion models has been constructed, which can achieve precise control over form (line count, rhyme, tonal patterns, etc.) and phonetics while maintaining semantic coherence.