AAAI 2026 Papers — Page 28
AAAI Conference on Artificial Intelligence · 4149 papers
Pansharpening for Thin-Cloud Contaminated Remote Sensing Images: A Unified Framework and Benchmark Dataset
Songcheng Du (Northwest Polytechnical University), Qiang Shen (Aberystwyth University)
RestorationSuper ResolutionConvolutional Neural NetworkTransformerImageBenchmark
🎯 What it does: This study investigates high-resolution fusion (pansharpening) of remote sensing images under thin cloud contamination, proposing a unified end-to-end framework called Pan-TCR to simultaneously address cloud removal and resolution enhancement.
Paper Folding Puzzles: Can Multimodal Large Language Models Perform Spatial Reasoning?
Dibin Zhou (Hangzhou Normal University), Fuchang Liu (University of Nottingham Ningbo China)
TransformerLarge Language ModelPrompt EngineeringVision Language ModelMultimodalityBenchmark
🎯 What it does: Proposed the Paper Folding Puzzles (PFP) benchmark to evaluate the ability of multimodal large language models in reasoning from 2D to 3D space.
ParaDySe: A Parallel Strategy Switching Framework for Dynamic Sequences in Transformer-based Large Language Models
Zhixin Ou (National University of Defense Technology), Baihui Liu (National University of Defense Technology)
Computational EfficiencyTransformerLarge Language ModelTextBiomedical Data
🎯 What it does: Propose the ParaDySe framework to achieve dynamic parallelism strategy hot switching at the Transformer level, addressing out-of-memory (OOM) issues and communication bottlenecks for variable-length sequences;
Parallel Training Time-to-First-Spike Spiking Neural Networks
Kaiwei Che (Peking University), Yonghong Tian (Peking University)
ClassificationSpiking Neural NetworkImage
🎯 What it does: Designed a fully parallel Time-to-First-Spike (TTFS) neural network training framework and proposed a temporal selection decoder based on membrane potential.
Parallelizable Riemannian Alternating Direction Method of Multipliers for Non-convex Pose Graph Optimization
Xin Chen (Beihang University), Liqun Qi (Beihang University)
Pose EstimationOptimizationSimultaneous Localization and MappingGraphBenchmark
🎯 What it does: Propose a fully parallel Riemannian ADMM algorithm called PRADMM with closed-form solutions for subproblems, designed for large-scale pose graph optimization (PGO)
ParaMETA: Towards Learning Disentangled Paralinguistic Speaking Styles Representations from Speech
Haowei Lou (University of New South Wales), Lina Yao (University of New South Wales)
ClassificationRecognitionGenerationRepresentation LearningConvolutional Neural NetworkRecurrent Neural NetworkTransformerVision Language ModelTextMultimodalityAudio
🎯 What it does: Propose the ParaMETA framework to learn and control disentangled speaking style embeddings in speech, supporting multi-task recognition and TTS control with text/speech prompts.
Parameter Merging with Gradient-Guided Supermasks in Online Continual Learning
Benliu Qiu (University of Electronic Science and Technology of China), Hongliang Li (University of Electronic Science and Technology of China)
Computational EfficiencyKnowledge DistillationImage
🎯 What it does: Proposed a gradient-guided supermask parameter fusion method that linearly merges parameters of old and new models directly in online continual learning;
Parameter-, Memory-, Time-Efficient Multi-Task Dense Vision Adaptation
Haiming Yao (Tsinghua University), Wei You (Huawei Technologies Co. Ltd)
SegmentationComputational EfficiencyTransformerMixture of ExpertsImage
🎯 What it does: Propose a dual-branch structure for multi-task dense visual adaptation framework, combining Bi-Directional Task Adaptation (BDTA) module and Mixture of Task Experts (MoTE) to achieve triple-efficient adaptation of pre-trained Swin Transformer in terms of parameters, GPU memory, and training time;
Parameter-Free Clustering via Self-Supervised Consensus Maximization
Lijun Zhang (National University of Defense Technology), Xinwang Liu (National University of Defense Technology)
Auto EncoderContrastive LearningImage
🎯 What it does: Proposes a completely parameter-free hierarchical clustering framework SCMax, integrating self-supervised consensus maximization to automatically generate and evaluate clustering structures.
Parameter-Free Fine-tuning via Redundancy Elimination for Vision Foundation Models
Jiahuan Long (Shanghai Jiao Tong University), Chao Ma (Shanghai Jiao Tong University)
ClassificationSegmentationDepth EstimationComputational EfficiencyTransformerImageBiomedical Data
🎯 What it does: Proposed a parameter-free tuning method that enhances downstream task performance by identifying and replacing redundant channels in visual foundation models.
Parameter-free Optimal Rates for Nonlinear Semi-Norm Contractions with Applications to Q-Learning
Ankur Naskar (Indian Institute of Science), Vijay Gupta (Purdue University)
Reinforcement Learning
🎯 What it does: This paper studies the Polyak-Ruppert averaging of nonlinear semi-norm convergent iterations, proving that an optimal convergence rate of ˜O(1/√t) can be achieved under parameter-independent step sizes, and applies this result to synchronous average reward Q-learning and asynchronous exponential discount Q-learning;
Parameterized Abstract Interpretation for Transformer Verification
Pei Huang (Stanford University), Clark Barrett (Stanford University)
Safty and PrivacyExplainability and InterpretabilityTransformerText
🎯 What it does: Propose two parameterizable linear abstract domains for more precise over-approximation of inner products in Transformer self-attention modules, thereby improving the formal verification effectiveness of neural networks.
Parametric Pareto Set Learning for Expensive Multi-Objective Optimization
Ji Cheng (City University of Hong Kong), Qingfu Zhang (City University of Hong Kong)
OptimizationBenchmark
🎯 What it does: Propose a Parametric Pareto Set Learning with Multi-Objective Bayesian Optimization (PPSL-MOBO) framework that can learn the entire Pareto set varying with parameters in one go, infer Pareto solutions in real-time for any parameter, and significantly reduce the demand for expensive function evaluations.
Pareto-Based Heterogeneous Knowledge Distillation for MLPs on Graphs
Wenrui Zhao, Chuxu Zhang (University of Utah)
Knowledge DistillationRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: Designed an HGKD framework that distills the structure and predictive knowledge of heterogeneous graph neural networks (HGNN) into a graph-structure-free MLP student model, eliminating the dependency on graph neighbor information during inference.
Pareto-Grid-Guided Large Language Models for Fast and High-Quality Heuristics Design in Multi-Objective Combinatorial Optimization
Ha Minh Hieu (Hanoi University of Science and Technology), Huynh Thi Thanh Binh (FPT Software AI Center)
OptimizationLarge Language ModelPrompt EngineeringBenchmark
🎯 What it does: Propose an automatic heuristic design framework named MPaGE based on large language models (LLM) for multi-objective combinatorial optimization problems (MOCOP), guiding heuristic evolution through Pareto front grid and semantic clustering to balance solution quality and runtime efficiency.
ParetoHqD: Fast Offline Multiobjective Alignment of Large Language Models Using Pareto High-Quality Data
Haoran Gu (Xidian University), Yaochu Jin (Victoria University of Wellington)
OptimizationReinforcement Learning from Human FeedbackLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposes a two-stage supervised fine-tuning method based on Pareto high-quality data (ParetoHqD) to achieve multi-objective large language model (LLM) alignment;
PARS: Partial-Label-Learning-inspired Recommender Systems
Shanshan Ye (University of Technology Sydney), Jie Lu (University of Technology Sydney)
Recommendation SystemTransformerSequential
🎯 What it does: This paper proposes a framework called PARS that constructs a recommendation system using only user browsing history, addressing the problem of recommendation without explicit purchase labels.
Partial Action Replacement: Tackling Distribution Shift in Offline MARL
Yue Jin (University of Warwick), Giovanni Montana (University of Warwick)
Reinforcement Learning
🎯 What it does: To address the joint action distribution shift problem in offline multi-agent reinforcement learning, this paper proposes Partial Action Replacement and Soft-Partial Conservative Q-Learning (SPaCQL) algorithms, leveraging the decomposition properties of the behavior policy to reduce distribution shift.
Partial Fairness Awareness: Belief-Guided Strategic Mechanism for Strategic Agents
Xinpeng Lv (National University of Defense Technology), Haotian Wang (Harbin Institute of Technology)
ClassificationOptimizationTabular
🎯 What it does: This paper proposes a partial fairness-aware framework and designs a belief-guided mechanism to address the fairness information exposure challenge in strategic classification.
Partially Shared Concept Bottleneck Models
Delong Zhao (Harbin Institute of Technology), Jun Yu (Harbin Institute of Technology)
ClassificationExplainability and InterpretabilityLarge Language ModelVision Language ModelContrastive LearningImage
🎯 What it does: Developed a partially shared concept bottleneck model (PS-CBM) that achieves interpretable image classification through multimodal concept generation, concept sharing strategies, and concept efficiency metrics.
PartialNet: Compute Less, Perform Better
Haiduo Huang (Xi'an Jiaotong University), Pengju Ren (Xi'an Jiaotong University)
ClassificationObject DetectionSegmentationComputational EfficiencyConvolutional Neural NetworkImage
🎯 What it does: Propose PartialNet, which achieves fewer parameters and lower FLOPs without accuracy degradation by proportionally splitting feature maps across channels and separately performing convolution and attention.
PASA: Progressive-Adaptive Spectral Augmentation for Automated Auscultation in Data-Scarce Environments
Ying Wang (Macao Polytechnic University), Xiaochen Yuan (Macao Polytechnic University)
Anomaly DetectionData-Centric LearningConvolutional Neural NetworkReinforcement LearningBiomedical DataAudio
🎯 What it does: Proposes Progressive-Adaptive Spectral Augmentation (PASA), a reinforcement learning framework that treats data augmentation as a Markov Decision Process (MDP), for adaptive augmentation of automated auscultation data;
PASE: Leveraging the Phonological Prior of WavLM for Low-Hallucination Generative Speech Enhancement
Xiaobin Rong (Nanjing University), Jing Lu (Cisco Systems)
RestorationGenerationTransformerGenerative Adversarial NetworkAudio
🎯 What it does: Propose a generative speech enhancement framework called PASE based on the pre-trained WavLM model, leveraging its inherent phonetic prior to alleviate hallucination issues under low signal-to-noise ratio (SNR) conditions;
PaSE: Prototype-aligned Calibration and Shapley-based Equilibrium for Multimodal Sentiment Analysis
Kang He (Wuhan University), Donghong Ji (Wuhan University)
ClassificationMultimodality
🎯 What it does: Propose the PaSE framework to address modality competition in multimodal sentiment analysis;
PASS: Probabilistic Agentic Supernet Sampling for Interpretable and Adaptive Chest X-Ray Reasoning
Yushi Feng (University of Hong Kong), Lequan Yu (University of Hong Kong)
Explainability and InterpretabilityComputational EfficiencyTransformerReinforcement LearningAgentic AIContrastive LearningMultimodalityBiomedical Data
🎯 What it does: Designed and implemented the PASS framework, utilizing probabilistic proxy supersampling to achieve interpretable, adaptive, and computationally cost-controllable multimodal chest X-ray reasoning systems.
PatchET: Learning Enzyme Temperature Properties Through Patch-Based Neural Architectures
Ziqi Zhang (Jiangnan University), Zhaohong Deng (Macquarie University)
Protein Structure PredictionConvolutional Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningSequentialBiomedical DataBenchmark
🎯 What it does: Propose the PatchET model, which employs a patch-based two-stage architecture to directly predict the optimal temperature, stability, and temperature range of enzymes from amino acid sequences, and release newly constructed benchmark datasets for optimal temperature (10,371 samples) and temperature range (1,818 samples).
PATexGS: Perceptual-Adaptive Texture Scheduling for Visual Coherence in Textured Gaussian Splatting
Yuesong Wang (Huazhong University of Science and Technology), Tao Guan (Huazhong University of Science and Technology)
Gaussian SplattingImage
🎯 What it does: Designed an adaptive texture scheduling framework PATexGS to enhance visual coherence in texture Gaussian point rendering.
PathFLIP: Fine-grained Language-Image Pretraining for Versatile Computational Pathology
Fengchun Liu (Harbin Institute of Technology), Yongbing Zhang (Harbin Institute of Technology)
ClassificationRetrievalTransformerLarge Language ModelVision Language ModelContrastive LearningImageTextMultimodalityBiomedical Data
🎯 What it does: Propose PathFLIP, a fine-grained language-image pre-training framework specifically designed for whole slide images (WSI), which achieves precise alignment between images and text without requiring region-level annotations.
PathMind: A Retrieve-Prioritize-Reason Framework for Knowledge Graph Reasoning with Large Language Models
Yu Liu (Institute of Information Engineering, Chinese Academy of Sciences), Yanan Cao (Peking University)
Computational EfficiencyRepresentation LearningGraph Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningTextGraphBenchmarkRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose the PathMind framework, which integrates LLM with knowledge graphs for retrieval-priority-reasoning, guiding LLM to complete question-answering reasoning through critical inference paths.
Patho-AgenticRAG: Towards Multimodal Agentic Retrieval-Augmented Generation for Pathology VLMs via Reinforcement Learning
Wenchuan Zhang (Sichuan University), Hong Bu (Sichuan University)
GenerationRetrievalReinforcement LearningAgentic AIVision Language ModelMultimodalityBiomedical DataRetrieval-Augmented Generation
🎯 What it does: This work proposes a multimodal retrieval-augmented generation framework called Patho-AgenticRAG for pathological vision-language models.
Patho-R1: A Multimodal Reinforcement Learning-Based Pathology Expert Reasoner
Wenchuan Zhang (Sichuan University), Hong Bu (Sichuan University)
Data-Centric LearningReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningVision Language ModelContrastive LearningImageTextMultimodalityBiomedical DataChain-of-Thought
🎯 What it does: Constructed a high-quality multimodal dataset centered on pathology textbooks, and trained two pathology-specific vision-language models, Patho-CLIP and Patho-R1;
PathRAG: Pruning Graph-based Retrieval Augmented Generation with Relational Paths
Boyu Chen (Beijing University of Posts and Telecommunications), Cheng Yang (Beijing University of Posts and Telecommunications)
GenerationRetrievalTransformerPrompt EngineeringTextGraphAgriculture RelatedRetrieval-Augmented Generation
🎯 What it does: This paper proposes the PathRAG method, which enhances the generation quality of large language models by retrieving key relationship paths from graph indexes and converting them into text prompts.
Paths Not Taken: Structure-Based Pruning in PSDD Learning and Inference
Cory Butz (University of Regina), Camilla Lewis (University of Regina)
Computational EfficiencyTextGraphBenchmark
🎯 What it does: This paper proposes a new method that utilizes the determinism of PSDD for parameter learning, structured pruning inference, and parallel circuit evaluation.
PatientVLM Meets DocVLM: Pre-Consultation Dialogue Between Vision-Language Models for Efficient Diagnosis
K Lokesh (Indian Institute of Technology Jodhpur), Anand Mishra (All India Institute of Medical Sciences New Delhi)
Supervised Fine-TuningPrompt EngineeringVision Language ModelBiomedical DataChain-of-Thought
🎯 What it does: Propose a pre-diagnosis dialogue framework (PCDF) that uses two vision-language models (DocVLM and PatientVLM) to simulate doctor questioning and patient responses, generating image-dialogue-diagnosis triplets for fine-tuning diagnostic models.
Pb4U-GNet: Resolution-Adaptive Garment Simulation via Propagation-before-Update Graph Network
Aoran Liu (University of Sydney), Zhiyong Wang (University of Tokyo)
GenerationGraph Neural NetworkMeshGraphPhysics Related
🎯 What it does: Proposes Pb4U-GNet, a graph neural network that decouples propagation and updates, for resolution-adaptive clothing simulation;
PBR3DGen: A VLM-Guided Mesh Generation with High-Quality PBR Texture
Xiaokang Wei (Hong Kong Polytechnic University), Yan Luximon (Hong Kong Polytechnic University)
GenerationData SynthesisTransformerVision Language ModelDiffusion modelNeural Radiance FieldAuto EncoderImageTextMesh
🎯 What it does: This paper proposes PBR3DGen, a two-stage 3D mesh and PBR material generation framework capable of directly generating high-quality relightable 3D assets from a single image or text prompt.
PC-CrossDiff: Point-Cluster Dual-Level Cross-Modal Differential Attention for Unified 3D Referring and Segmentation
Wenbin Tan (Xiamen University), Yanyun Qu (Xiamen University)
SegmentationTransformerVision Language ModelPoint Cloud
🎯 What it does: Propose a unified two-layer cross-modal differential attention framework named PC-CrossDiff, which can simultaneously accomplish 3D object localization (3DREC) and segmentation (3DRES) tasks.
PC-Flow: Preference Alignment in Flow Matching via Classifier
Shaomeng Wang (Nanjing University of Science and Technology), Jinhui Tang (Nanjing Forestry University)
GenerationFlow-based ModelImageStochastic Differential Equation
🎯 What it does: Proposes PC-Flow, a method that achieves human preference alignment within the Flow Matching framework without requiring a reference model or full-model fine-tuning.
PCFormer: Accelerating Privacy-preserving Transformer Inference by Partition and Combination
Bo Zeng (Wuhan University), Run Wang (Wuhan University)
Safty and PrivacyComputational EfficiencyTransformerTextBenchmark
🎯 What it does: Propose the PCFormer framework, which partitions and merges nonlinear redundancies in Transformers during privacy inference to reduce computational and communication costs in HE/MPC.
PCGS: Progressive Compression of 3D Gaussian Splatting
Yihang Chen (Shanghai Jiao Tong University), Jianfei Cai (Monash University)
CompressionGaussian Splatting
🎯 What it does: Proposed the PCGS (Progressive Compression of 3D Gaussian Splatting) framework, which can generate multi-level progressive compressed bitstreams after a single training session;
PCoKG: Personality-aware Commonsense Reasoning with Debate
Weijie Li (Soochow University), Guodong Zhou (Soochow University)
GenerationTransformerLarge Language ModelSupervised Fine-TuningTextGraph
🎯 What it does: Constructed a Personality-aware Commonsense Knowledge Graph (PCoKG), systematically organizing quadruples composed of events, MBTI personality types, reasoning dimensions, and reasoning outcomes, while enhancing knowledge quality through a multi-round LLM debate mechanism; simultaneously verifying its effectiveness in personalized dialogue generation tasks.
PCSR: Pseudo-label Consistency-Guided Sample Refinement for Noisy Correspondence Learning
Zhuoyao Liu (Sichuan University), Shudong Huang (Sichuan University)
RetrievalData-Centric LearningImageTextMultimodality
🎯 What it does: To address the noise correspondence problem in cross-modal retrieval, the PCSR framework is proposed, which subdivides training samples into three categories: clean, rewritable, and ambiguous through pseudo-label consistency. Adaptive optimization strategies are then applied to each category to enhance model robustness under noisy data.
PDE-Driven Spatiotemporal Generative Modeling for Multilead ECG Synthesis
Yakir Yehuda (Technion-Israel Institute of Technology), Kira Radinsky (Technion-Israel Institute of Technology)
GenerationData SynthesisGenerative Adversarial NetworkTime SeriesBiomedical DataElectrocardiogram
🎯 What it does: This study proposes a physics-informed partial differential equation GAN (PhysioPDE-GAN) for synthesizing high-fidelity 12-lead ECG data;
PEFT-BoA: Parameter-Efficient Fine-Tuning with Bag-of-Adapters for Multi-Modal Object Re-identification
Hongchao Li, YongLong Luo (Anhui Normal University)
RecognitionTransformerSupervised Fine-TuningVision Language ModelMultimodality
🎯 What it does: Proposes a parameter-efficient fine-tuning framework PEFT-BoA based on CLIP, achieving multi-modal object ReID through three lightweight adapters.
PEGNet: A Physics-Embedded Graph Network for Long-Term Stable Multiphysics Simulation
Can Yang (Xiamen University), Min Jiang (Xiamen University)
Graph Neural NetworkMeshGraphBiomedical DataPhysics Related
🎯 What it does: Proposed a physics-embedded graph network called PEGNet for long-term stable multi-physics simulations, primarily addressing the issues of error accumulation and insufficient physical consistency in traditional data-driven models.
PEOAT: Personalization-Guided Evolutionary Question Assembly for One-Shot Adaptive Testing
Xiaoshan Yu, Xingyi Zhang (Anhui University)
Optimization
🎯 What it does: Propose the One-time Adaptive Testing (OAT) task and design a personalized item assembly framework PEOAT using evolutionary algorithms.
PEOCH: Online Cross-Modal Hashing with Semi-Supervised Streaming Data Driving Prototype Evolution
Xiao Kang (Shandong University), Yilong Yin (Shandong University)
RetrievalMultimodalityBenchmark
🎯 What it does: Propose a semi-supervised online cross-modal hashing method called PEOCH, which generates consistent hash codes through streaming data-driven prototype evolution
PEOD: A Pixel-Aligned Event-RGB Benchmark for Object Detection Under Challenging Conditions
Luoping Cui (Beijing University of Posts and Telecommunications), Chuang Zhu (Beijing University of Posts and Telecommunications)
Object DetectionMultimodalityBenchmark
🎯 What it does: This paper proposes PEOD, a 1280×720 pixel-aligned event-RGB dataset designed for object detection under harsh conditions such as extreme lighting and high speeds.
PepCCD: A Contrastive Conditioned Diffusion Framework for Target-Specific Peptide Generation
Jun Zhang (Shenzhen University), Zexuan Zhu (Shenzhen University)
GenerationDrug DiscoveryTransformerLarge Language ModelSupervised Fine-TuningDiffusion modelContrastive LearningBiomedical Data
🎯 What it does: Propose a contrastive learning conditional diffusion framework (PepCCD) that can generate targeted specific peptides based solely on the target protein sequence, eliminating the need for structural information.
Perceive More with Less: LiDAR Point Cloud Compression at Just Recognizable Distortion for 3D Scene Understanding
Miaohui Wang (Shenzhen University), Yun Song
CompressionConvolutional Neural NetworkTransformerPoint Cloud
🎯 What it does: This paper proposes a LiDAR point cloud compression method for machine vision systems, defines a recognizable compression distortion threshold, and constructs an end-to-end compression framework based on this threshold.
Perceive, Act and Correct: Confidence Is Not Enough for Hyperspectral Classification
Muzhou Yang (Nanjing University of Aeronautics and Astronautics), Mingqiang Wei (Nanjing University of Aeronautics and Astronautics)
ClassificationImage
🎯 What it does: Propose the CABIN framework, which enhances hyperspectral image classification under low annotation scenarios through a perception-action-correction loop.
Perceiving the Knowledge Boundary: Uncertainty-Guided Exploration and Imagination for World Models
Zhenxian Liu (Peking University), Yonghong Tian (Peking University)
Reinforcement LearningWorld ModelImageVideoBenchmark
🎯 What it does: The study leverages uncertainty in world models to perceive knowledge boundaries and uses these boundaries to guide exploration and filter hallucinatory rollouts, thereby improving the performance of reinforcement learning based on world models.
Perception in Plan: Coupled Perception and Planning for End-to-End Autonomous Driving
Bozhou Zhang (Fudan University), Li Zhang (Fudan University)
Autonomous DrivingTransformerImageMultimodality
🎯 What it does: Proposed the VeteranAD framework, adopting the 'perception-in-plan' paradigm, embedding the perception module into the planning process, utilizing multi-modal anchor trajectories as planning priors, performing localization-aware and autoregressive trajectory planning, and constructing an end-to-end autonomous driving system.
Perceptual Quality Assessment of 3D Gaussian Splatting: A Subjective Dataset and Prediction Metric
Zhaolin Wan (Harbin Institute of Technology), Debin Zhao (Harbin Institute of Technology)
Graph Neural NetworkAuto EncoderGaussian SplattingPoint CloudBenchmark
🎯 What it does: Constructed a subjective quality assessment dataset specifically for 3D Gaussian Splatting (3DGS) called 3DGS-QA, and proposed a no-reference quality prediction model named GSOQA;
PeriUn: Enhancing Unlearning by Selectively Forgetting Peripheral Samples
Hee Bin Yoo (Seoul National University), Byoung-Tak Zhang (Seoul National University)
ClassificationConvolutional Neural NetworkImageBenchmark
🎯 What it does: Propose a selective forgetting method called PeriUn based on peripheral samples to approximate retrained models and reduce catastrophic forgetting.
Permutation Equivariant Framelet-based Hypergraph Neural Networks
Ming Li (Zhejiang Normal University), Pietro Lio (Beijing Normal University)
ClassificationGraph Neural NetworkImageGraph
🎯 What it does: Propose a new permutation equivariant framelet-based hypergraph neural network, PEF-HNN, for multi-scale feature extraction and learning on hypergraphs.
Persistent Autoregressive Mapping with Traffic Rules for Autonomous Driving
Shiyi Liang (Xi'an Jiaotong University), Xing Wei (Amap, Alibaba Group)
Autonomous DrivingTransformerLarge Language ModelVision Language ModelImageMultimodality
🎯 What it does: Propose an end-to-end autoregressive framework called PAMR for simultaneously generating lane vectors and traffic rules, achieving rule persistence through a caching mechanism;
Persistent Backdoor Attacks Under Continual Fine-Tuning of LLMs
Jing Cui (University Of Chinese Academy Of Sciences), Junge Zhang (Inria)
Adversarial AttackTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Optimize gradient alignment on trigger words before the release of large language models, enabling the implanted backdoor to remain efficient during subsequent user continuous fine-tuning processes;
Personality-guided Public-Private Domain Disentangled Hypergraph-Former Network for Multimodal Depression Detection
Changzeng Fu (Northeastern University), Chaoran Liu (Northeastern University)
ClassificationGraph Neural NetworkTransformerLarge Language ModelGenerative Adversarial NetworkContrastive LearningMultimodalityTime Series
🎯 What it does: Proposed a Hypergraph Former network (P3HF) based on personalized guidance and public-private domain decoupling for multimodal depression detection.
Personalize Anything for Free with Diffusion Transformer
Haoran Feng (Tsinghua University), Lu Sheng (Renmin University)
GenerationTransformerDiffusion modelFlow-based ModelImageBenchmark
🎯 What it does: This paper proposes an untrained differential transformer (DiT) personalized image generation framework called 'Personalize Anything,' achieving high-fidelity subject reconstruction while accommodating editing, layout, and multi-subject synthesis through resolution-agnostic token replacement at specific positions during the reverse process.
Personalize Before Retrieve: LLM-based Personalized Query Expansion for User-Centric Retrieval
Yingyi Zhang (Dalian University of Technology), Xiangyu Zhao (City University of Hong Kong)
RetrievalTransformerLarge Language ModelTextGraph
🎯 What it does: This paper proposes the PBR (Personalize-Before-Retrieve) framework, which enhances user-centered retrieval effectiveness by generating personalized query expansions through LLM before retrieval.
Personalize Your Gaussian: Consistent 3D Scene Personalization from a Single Image
Yuxuan Wang (Nanyang Technological University), Hanwang Zhang (Nanyang Technological University)
GenerationData SynthesisSupervised Fine-TuningRectified FlowGaussian SplattingImageMesh
🎯 What it does: Achieve personalized editing of 3D Gaussian Splatting scenes using a single reference image, constructing a three-stage consistency propagation framework CP-GS that progresses from coarse to fine.
Personalized Federated Graph-Level Clustering Network
Jingxin Liu, Yue Yang (Hainan University)
Federated LearningGraph Neural NetworkAuto EncoderGraph
🎯 What it does: Proposes PERFECT, a personalized federated graph-level clustering framework for multi-client structural heterogeneity, which utilizes privacy-preserving representative samples to enable cross-client information sharing and enhances local clustering performance through clustering gradient optimization.
Personalized Federated Learning with Bidirectional Communication Compression via One-Bit Random Sketching
Jiacheng Cheng, Kaiyuan Feng (Knowin AI)
Federated LearningComputational EfficiencyImage
🎯 What it does: Proposed the pFed1BS framework to achieve personalized federated learning with extremely low communication costs, using 1-bit random sketch compression for bidirectional communication and introducing symbol regularization to align local models with the global model;
Perspective from a Broader Context: Can Room Style Knowledge Help Visual Floorplan Localization?
Bolei Chen (Central South University), Jianxin Wang (Central South University)
Convolutional Neural NetworkContrastive LearningSimultaneous Localization and MappingImagePoint Cloud
🎯 What it does: Propose an unsupervised room style discriminator trained with clustering constraints, leveraging room style information to enhance visual floor plan localization and alleviate localization uncertainty caused by repetitive structural layouts.
PerTouch: VLM-Driven Agent for Personalized and Semantic Image Retouching
Zewei Chang (Nankai University), Chongyi Li (Samsung Electronics)
RestorationSegmentationAgentic AIVision Language ModelDiffusion modelImage
🎯 What it does: Proposes a unified diffusion model framework called PerTouch, which can achieve semantic-level fine-grained image retouching while maintaining aesthetic consistency globally.
Perturb Your Data: Paraphrase-Guided Training Data Watermarking
Pranav Shetty (Jpmorgan Ai Research), Manuela Veloso (Jpmorgan Ai Research)
Safty and PrivacyTransformerLarge Language ModelScore-based ModelText
🎯 What it does: Propose the SPECTRA method, which utilizes an LLM to generate multiple paraphrases and selects watermarked text with Min-K%++ scores similar to the original text, enabling subsequent detection models to determine if the data was used for training.
Perturbing Best Responses in Zero-Sum Games
Adam Dziwoki (Czech Technical University in Prague), Rostislav Horcik
OptimizationComputational EfficiencyReinforcement LearningTabular
🎯 What it does: The study uses perturbed best response (PBRO) in zero-sum games to improve the iterative complexity of Fictitious Play (FP) and Double Oracle (DO) algorithms, and proposes corresponding randomized versions: Stochastic Fictitious Play (SFP) and Stochastic Double Oracle (SDO).
Perturbing to Preserve: Defending Fragile Knowledge in Online Continual Learning
Dulan Zhou (National University of Defense Technology), Kele Xu (National University of Defense Technology)
Image
🎯 What it does: Propose the PDFK framework, combining exponential moving average (EMA) smoothing and structured perturbation consistency regularization to address knowledge fragility in online continual learning.
PET2Rep: Towards Vision-Language Model-Drived Automated Radiology Report Generation for Positron Emission Tomography
Yichi Zhang (Fudan University), Le Xue (Fudan University)
GenerationPrompt EngineeringVision Language ModelMultimodalityBiomedical DataPositron Emission TomographyBenchmark
🎯 What it does: Constructed the PET2Rep benchmark to evaluate the performance of visual-language models in PET image report generation tasks.
PFAvatar: Pose-Fusion 3D Personalized Avatar Reconstruction from Real-World Outfit-of-the-Day Photos
Dianbing Xi (Zhejiang University), Rui Wang (Zhejiang University)
GenerationPose EstimationSupervised Fine-TuningDiffusion modelNeural Radiance FieldImageMesh
🎯 What it does: Propose a two-stage method named PFAvatar that can rapidly generate high-quality, editable 3D human avatars from real-world 'outfit-of-the-day' (OOTD) photographs;
PGMamba: A Physical Model-Guided Global Mamba for Underwater Image Enhancement
Zijun Tan (Chongqing University), Fulin Luo (Chongqing University)
RestorationConvolutional Neural NetworkImagePhysics Related
🎯 What it does: Proposed a global enhancement model PGMamba that combines a physical model with Mamba for underwater image enhancement.
Phantom Menace: Exploring and Enhancing the Robustness of VLA Models Against Physical Sensor Attacks
Xuancun Lu (Zhejiang University), Wenyuan Xu (Zhejiang University)
Data SynthesisAdversarial AttackRobotic IntelligenceVision-Language-Action ModelImageVideoTextMultimodalityUltrasoundAudio
🎯 What it does: This paper systematically evaluates the robustness of Vision-Language-Action (VLA) models against physical sensor attacks (such as laser, light projection, EMI, ultrasound, etc.), and proposes an automated 'Real-Sim-Real' framework to simulate these attacks in simulation and verify them on real robots; subsequently, adversarial training is applied to enhance the model's defensive performance.
PharmaQA: Prompt-Based Molecular Representation Learning via Pharmacophore-Oriented Question Answering
Chengwei Ai (Central South University), Fei Guo (Central South University)
Representation LearningDrug DiscoveryGraph Neural NetworkTransformerPrompt EngineeringContrastive LearningTextGraph
🎯 What it does: Propose PharmaQA, a framework that learns molecular representations through pharmacophore-related question-answering, leveraging structured questions and descriptions to guide the model in capturing context-aware molecular semantics.
Phased One-Step Adversarial Equilibrium for Video Diffusion Models
Jiaxiang Cheng (Tencent Hunyuan), Qinglin Lu (Tencent Hunyuan)
GenerationKnowledge DistillationDiffusion modelScore-based ModelGenerative Adversarial NetworkVideoMultimodality
🎯 What it does: Proposed a two-stage single-step video generation distillation framework called V-PAE, achieving high-quality single-step generation for large-scale video diffusion models.
PHPFND: Detecting Fake News via Post-Hoc Processing of LLMs Hallucination
Jinke Ma (Heilongjiang University), Yong Liu (Heilongjiang University)
ClassificationAnomaly DetectionExplainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: Propose the PHPFND framework, integrating hallucination detection, correction, and feature enhancement of large language models (LLMs) into the fake news detection process, significantly improving discriminative performance through multi-round correction and hallucination-driven feature processing.
PhyPlan: Learning to Plan Tasks with Generalizable and Rapid Physical Reasoning for Embodied Manipulation
Ankit Kanwar (Indian Institute of Technology Delhi), Souvik Chakraborty (Indian Institute of Technology Delhi)
Robotic IntelligenceFlow-based ModelPhysics Related
🎯 What it does: Proposed the PhyPlan framework, combining GFlowNet with MCTS to achieve multi-step physical reasoning and tool use in 3D robot environments; rapidly predict dynamics using physics-informed neural networks and correct rewards with Gaussian Processes, enabling joint planning of discrete tool selection and continuous control.
Phys-Liquid: A Physics-Informed Dataset for Estimating 3D Geometry and Volume of Transparent Deformable Liquids
Ke Ma (Huazhong University of Science and Technology), Tian Xia (Huazhong University of Science and Technology)
SegmentationGenerationData SynthesisTransformerDiffusion modelImageMultimodalityMeshTime SeriesBenchmarkPhysics Related
🎯 What it does: Constructed the Phys-Liquid physical information dataset for estimating the 3D geometry and volume of transparent variable liquids, and proposed a four-stage reconstruction pipeline for validation
Physical-regularized Hierarchical Generative Model for Metallic Glass Structural Generation and Energy Prediction
Qiyuan Chen (University of Wisconsin - Madison), Bu Wang (Stanford University)
GenerationData SynthesisGraph Neural NetworkAuto EncoderGraphPhysics Related
🎯 What it does: A physics-regularized hierarchical graph variational autoencoder, GLASSVAE, was studied for structure generation and energy prediction in metallic glasses.
Physically-Based LiDAR Smoke Simulation for Robust 3D Object Detection
Shijun Zheng (Xiamen University), Cheng Wang (Xiamen University)
Data SynthesisAutonomous DrivingPoint CloudBenchmarkPhysics Related
🎯 What it does: This paper proposes a physics-based LiDAR fog simulation framework that uses Unity's 3D fluid dynamics to simulate smoke and combines it with a real LiDAR perception model, seamlessly integrating synthetic smoke point clouds into large driving datasets (e.g., Waymo) to enhance the robustness of 3D object detection in foggy environments.
Physically-Informed Flow Matching with Graph Neural Networks for Complex Fluid Dynamics
Xiaozhuang Song (Chinese University of Hong Kong), Tianshu Yu (Chinese University of Hong Kong)
GenerationGraph Neural NetworkFlow-based ModelGraphPhysics RelatedOrdinary Differential Equation
🎯 What it does: Propose a Physically-Informed Flow Matching Graph Networks (PIFM-GN) framework that directly samples fluid states satisfying incompressibility constraints from the target distribution, eliminating the need for time-step simulations in traditional CFD.
Physics-Aware Accelerated Unrolling Model for Sparse-View CT Reconstruction
Shaojie Guo (East China Normal University), Yan Wang (East China Normal University)
RestorationConvolutional Neural NetworkBiomedical DataComputed Tomography
🎯 What it does: Proposed a Physics-Aware Accelerated Iterative Model (PAUM) for sparse-view CT reconstruction.
Physics-Informed Approach for Exploratory Hamilton–Jacobi–Bellman Equations via Policy Iterations
Yeongjong Kim (Pohang University of Science and Technology), Yeoneung Kim (Seoul National University of Science and Technology)
Reinforcement LearningPhysics RelatedStochastic Differential Equation
🎯 What it does: Propose a mesh-free soft policy iteration framework based on physics-informed neural networks (PINN) for solving the Hamilton-Jacobi-Bellman equation in entropy-regularized stochastic control problems.
Physics-Informed Deformable Gaussian Splatting: Towards Unified Constitutive Laws for Time-Evolving Material Field
Haoqin Hong (University of Science and Technology of China), Jingrun Chen (University of Science and Technology of China)
GenerationGaussian SplattingOptical FlowVideoPhysics Related
🎯 What it does: Proposed the Physics-Informed Deformable Gaussian Expansion (PIDG) model, which treats 3D Gaussian particles as physical material points from a Lagrangian perspective, and reconstructs monocular dynamic scenes by jointly evolving time-varying material fields.
Physics-Informed Koopman Neural Estimation of the Heston Model from High-Frequency Observations
Qiuming Zhu (East China Normal University), Ziwei Zhou (Shanghai University of Finance and Economics)
Time SeriesFinance RelatedStochastic Differential Equation
🎯 What it does: Propose a framework called Koopman-PINN, which combines ART nonparametric volatility filtering, temporal matching initialization, and neural network learning of the Koopman operator to estimate five parameters of the Heston model using high-frequency price data.
Physics-Informed Multi-Task Learning for Battery State of Health Prediction with Uncertainty Quantification
Tianwen Zhu (Nanyang Technological University), Yonggang Wen (Wuhan University)
Auto EncoderTime SeriesPhysics Related
🎯 What it does: Propose a multi-task learning framework that jointly predicts battery SOH using physics-informed neural networks (PINN) and constructs an energy-based proxy metric to quantify prediction uncertainty through a deep autoencoder Gaussian Mixture Model (DAGMM).
PhysicsCorrect: A Training-Free Approach for Stable Neural PDE Simulations
Xinquan Huang (University of Pennsylvania), Paris Perdikaris (University of Pennsylvania)
Computational EfficiencyConvolutional Neural NetworkTransformerMeshTime SeriesPhysics Related
🎯 What it does: Proposes PhysicsCorrect, a physics-consistent correction framework that does not require retraining, to stabilize error accumulation in neural PDE solvers during long-time sequence simulations.
PhysPatch: A Physically Realizable and Transferable Adversarial Patch Attack for Multimodal Large Language Models-based Autonomous Driving Systems
Qi Guo (Xi'an Jiaotong University), Qing Guo (Hangzhou Dianzi University)
Autonomous DrivingAdversarial AttackTransformerLarge Language ModelVision Language ModelImageMultimodality
🎯 What it does: A physically realizable adversarial patch attack framework called PhysPatch is proposed for autonomous driving systems driven by multimodal large language models (MLLM), which exhibits good cross-model transferability.
Picking a Representative Set of Solutions in Multiobjective Optimization: Axioms, Algorithms, and Experiments
Niclas Boehmer (Hasso Plattner Institute, University of Potsdam), Maximilian T. Wittmann (Hasso Plattner Institute, University of Potsdam)
OptimizationBenchmark
🎯 What it does: The paper studies the Pareto pruning problem in multi-objective optimization, systematically evaluates existing representative quality metrics, and proposes a new metric called directed coverage, combining axiomatic analysis, complexity proof, and experimental evaluation.
Piercing the Fog: Disentangling Key Features for Vision Models in Multi-Degradation Scenarios
Siyu Chen, Fei Guo (School of Computer Science and Engineering Central South University)
RestorationSegmentationTransformerContrastive LearningImage
🎯 What it does: Propose a degradation decoupling model (DDM) for image processing in few-shot semantic segmentation tasks under various degradations (rain, fog, snow, motion blur, etc.).
PIF-Net: Ill-Posed Prior Guided Multispectral and Hyperspectral Image Fusion via Invertible Mamba and Fusion-Aware LoRA
Baisong Li (Jilin University), Haixiao Xu (Jilin University)
Super ResolutionImageMultimodality
🎯 What it does: Proposed PIF-Net, which leverages reversible Mamba and fusion-aware LoRA to perform lossless and interpretable fusion of multispectral and hyperspectral images, generating images with high spatial and hyperspectral resolutions.
PIMRL: Physics-Informed Multi-Scale Recurrent Learning for Burst-Sampled Spatiotemporal Dynamics
Han Wan (Renmin University of China), Hao Sun (Renmin University of China)
Recurrent Neural NetworkAuto EncoderTime SeriesSequentialPhysics Related
🎯 What it does: Proposes the Physics-Informed Multi-Scale Recurrent Learning (PIMRL) framework for predicting and simulating burst-sampled multi-scale spatiotemporal dynamics data;
PINet: Improving the Stability of Prototype Networks via Phantasia-Inspired Uncertain Representations
Ho Kyung Shin (Kyungpook National University), Woo-Jeoung Nam (Kyungpook National University)
ClassificationRecognitionExplainability and InterpretabilityRepresentation LearningConvolutional Neural NetworkImage
🎯 What it does: Propose Prototype in Imagery Network (PINet), which simulates visual mental imagery (Phantasia) by introducing blank inputs and using rough position guidance during feature extraction to enhance the interpretability and stability of prototype networks.
PINGS-X: Physics-Informed Normalized Gaussian Splatting with Axes Alignment for Efficient Super-Resolution of 4D Flow MRI
Sun Jo (Hanyang University), Je Hyeong Hong (Hanyang University)
Super ResolutionGaussian SplattingBiomedical DataMagnetic Resonance ImagingPhysics Related
🎯 What it does: Propose PINGS-X, a physics-informed explicit Gaussian splatting model for super-resolution reconstruction of 4D flow MRI.
PipeDiT: Accelerating Diffusion Transformers in Video Generation with Task Pipelining and Model Decoupling
Sijie Wang (Harbin Institute of Technology), Shaohuai Shi (Harbin Institute of Technology)
GenerationComputational EfficiencyTransformerDiffusion modelAuto EncoderVideo
🎯 What it does: This paper proposes the PipeDiT framework, which significantly accelerates DiT-based video generation by leveraging task pipelining and module decoupling.
PIPHEN: Physical Interaction Prediction with Hamiltonian Energy Networks
Kewei Chen (Chinese Academy of Sciences), Mingsheng Shang (University of Chinese Academy of Sciences)
OptimizationFederated LearningKnowledge DistillationRobotic IntelligenceGraph Neural NetworkTransformerWorld ModelPhysics Related
🎯 What it does: Propose a distributed multi-robot collaboration framework called PIPHEN, which generates low-dimensional physical representations through edge semantic distillation, and further reduces communication burden and decision latency by employing Hamiltonian energy networks for energy conservation control.
PITE: Multi-Prototype Alignment for Individual Treatment Effect Estimation
Fuyuan Cao (Shanxi University), Xiaoli Li (Singapore University of Technology and Design)
Domain AdaptationRepresentation LearningTabular
🎯 What it does: This paper proposes a multi-prototype alignment framework (PITE) for estimating individual treatment effects from observational data.
Pixel-level Quality Assessment for Oriented Object Detection
Yunhui Zhu (Nanjing Audit University), Buliao Huang (Jinling Institute of Technology)
Object DetectionImage
🎯 What it does: Proposes a Pixel-level Quality Assessment (PQA) framework, using pixel-level spatial consistency instead of traditional box-level IoU prediction to evaluate localization quality in directional object detection.
PKR-QA: A Benchmark for Procedural Knowledge Reasoning with Knowledge Module Learning
Thanh-Son Nguyen (Agency for Science, Technology and Research), Basura Fernando (Agency for Science, Technology and Research)
Explainability and InterpretabilityLarge Language ModelVision Language ModelVision-Language-Action ModelContrastive LearningVideoTextMultimodalityGraphBenchmark
🎯 What it does: Constructed the PKR-QA question-answering benchmark and the corresponding Program Knowledge Graph (PKG), and proposed the Knowledge Module Learning (KML) neuro-symbolic framework, which uses learnable knowledge modules to execute reasoning programs generated by LLMs, achieving interpretable reasoning on procedural tasks.
PLA-MGRA: Multi-Granularity and Relation-Aware Learning for Efficient and Generalizable Protein-Ligand Binding Affinity Prediction
Shunfan Li (China University of Geosciences), Xuesong Yan (China University of Geosciences)
Drug DiscoveryGraph Neural NetworkGraphBiomedical Data
🎯 What it does: Proposed the PLA-MGRA framework for protein-ligand affinity prediction.