AAAI 2024 Papers — Page 17
AAAI Conference on Artificial Intelligence · 2331 papers
Poincaré Differential Privacy for Hierarchy-Aware Graph Embedding
Yuecen Wei (Beihang University), Chunming Hu (Guangxi Normal University)
Safty and PrivacyRepresentation LearningGraph Neural NetworkGaussian SplattingGraph
🎯 What it does: This paper proposes the PoinDP framework, which implements differential privacy protection for hierarchical-aware graph embedding in hyperbolic (Poincaré) space.
Point Cloud Part Editing: Segmentation, Generation, Assembly, and Selection
Kaiyi Zhang (Fudan University), Cheng Jin (Fudan University)
SegmentationGenerationAuto EncoderGenerative Adversarial NetworkPoint Cloud
🎯 What it does: A point cloud component editing framework SGAS based on a four-stage process (segmentation, generation, assembly, selection) is proposed, which can achieve unsupervised component-aware point cloud generation through pruning.
Point Deformable Network with Enhanced Normal Embedding for Point Cloud Analysis
Xingyilang Yin (Xidian University), Xinbo Gao (Chongqing University of Posts and Telecommunications)
ClassificationSegmentationPoint Cloud
🎯 What it does: A point cloud analysis network PDNet based on MLP is proposed, with the core components being the Point Deformable Aggregation Module (PDAM) and Enhanced Normal Embedding (ENE);
Point Transformer with Federated Learning for Predicting Breast Cancer HER2 Status from Hematoxylin and Eosin-Stained Whole Slide Images
Bao Li (University of Science and Technology of China), Jie Tian (Institute of Automation, Chinese Academy of Sciences)
ClassificationFederated LearningTransformerImagePoint CloudBiomedical Data
🎯 What it does: Using point transformers combined with federated learning methods to predict HER2 status in HE-stained whole slide images (WSI);
Point-PEFT: Parameter-Efficient Fine-Tuning for 3D Pre-trained Models
Yiwen Tang (Northwestern Polytechnical University), Xuelong Li (Shanghai AI Laboratory)
ClassificationTransformerPrompt EngineeringPoint Cloud
🎯 What it does: A parameter-efficient fine-tuning framework for 3D point cloud pre-trained models, called Point-PEFT, is proposed.
Point-to-Spike Residual Learning for Energy-Efficient 3D Point Cloud Classification
Qiaoyun Wu (Anhui University), Changyin Sun (Anhui University)
ClassificationComputational EfficiencySpiking Neural NetworkPoint Cloud
🎯 What it does: This paper proposes a Point-to-Spike Residual Learning Network (P2SResLNet) for 3D point cloud classification that significantly reduces energy consumption.
Point2Real: Bridging the Gap between Point Cloud and Realistic Image for Open-World 3D Recognition
Hanxuan Li (Chinese Academy of Sciences), Xilin Chen (Chinese Academy of Sciences)
RecognitionDomain AdaptationPrompt EngineeringVision Language ModelPoint CloudMesh
🎯 What it does: By first converting point clouds into meshes, then performing realistic texture rendering and selecting the most recognizable viewpoints, the frozen CLIP model is utilized to achieve zero-shot and few-shot 3D recognition for unknown categories.
PointAttN: You Only Need Attention for Point Cloud Completion
Jun Wang (Zhejiang University of Technology), Chunhua Shen (Zhejiang University)
GenerationData SynthesisAutonomous DrivingTransformerPoint Cloud
🎯 What it does: A point cloud completion framework called PointAttN based on Transformer has been designed and implemented, utilizing cross-attention (GDP unit) and self-attention (SFA unit) to achieve generation from incomplete point clouds to complete point clouds without explicit local neighborhood partitioning.
PointPatchMix: Point Cloud Mixing with Patch Scoring
Yi Wang (Central South University), Pheng Ann Heng
ClassificationTransformerPoint Cloud
🎯 What it does: A patch-level mixed point cloud enhancement method called PointPatchMix is proposed, which uses a pre-trained teacher model to score the importance of patches and generate more reasonable mixed labels.
Polyper: Boundary Sensitive Polyp Segmentation
Hao Shao (Nankai University), Qibin Hou (Nankai University)
SegmentationConvolutional Neural NetworkTransformerBiomedical DataBenchmark
🎯 What it does: This paper proposes a multi-stage end-to-end multi-instance detection network called Polyper, based on morphological segmentation and internal multi-scale feature attention, for accurately segmenting the contours of colorectal polyps.
PORTAL: Automatic Curricula Generation for Multiagent Reinforcement Learning
Jizhou Wu (Tianjin University), Matthew E. Taylor (University of Alberta)
Reinforcement Learning
🎯 What it does: A multi-agent automatic curriculum learning framework named PORTAL is proposed, which utilizes dual indicators of task difficulty and task similarity to automatically select intermediate tasks, helping agents converge faster on extremely difficult collaborative tasks.
PosDiffNet: Positional Neural Diffusion for Point Cloud Registration in a Large Field of View with Perturbations
Rui She (Nanyang Technological University), Xingchao Jian (Nanyang Technological University)
Graph Neural NetworkTransformerDiffusion modelPoint CloudOrdinary Differential Equation
🎯 What it does: Utilizing Beltrami flow in graph neural partial differential equations to achieve dual embedding of point cloud features and positional information, and completing the correspondence at the window-patch-point three-level hierarchy through a neural ODE-driven Transformer, ultimately resulting in high-precision rigid registration outcomes.
PoseGen: Learning to Generate 3D Human Pose Dataset with NeRF
Mohsen Gholami (University of British Columbia), Z. Jane Wang (University of British Columbia)
GenerationData SynthesisPose EstimationNeural Radiance FieldGenerative Adversarial NetworkImage
🎯 What it does: Proposes an end-to-end PoseGen framework that utilizes NeRF to generate multi-view 3D human poses and corresponding images to enhance the generalization ability of pre-trained pose estimation models.
PPEA-Depth: Progressive Parameter-Efficient Adaptation for Self-Supervised Monocular Depth Estimation
Yue-Jiang Dong (Tsinghua University), Song-Hai Zhang
Depth EstimationDomain AdaptationKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: The PPEA-Depth framework is proposed, which achieves self-supervised monocular depth estimation by inserting encoder and decoder adapters into a pre-trained image model, implemented in two stages (first adapting to static scenes, then fine-tuning in dynamic scenes).
PPIDSG: A Privacy-Preserving Image Distribution Sharing Scheme with GAN in Federated Learning
Yuting Ma (University of Science and Technology of China), Xiaohua Xu (University of Science and Technology of China)
GenerationFederated LearningSafty and PrivacyAuto EncoderGenerative Adversarial NetworkImage
🎯 What it does: A privacy protection scheme PPIDSG is proposed in federated learning that does not upload classifier parameters. It utilizes block scrambling encryption, GAN image distribution sharing, and local classification training to defend against image reconstruction, label inference, and membership inference attacks.
PPO-Clip Attains Global Optimality: Towards Deeper Understandings of Clipping
Nai-Chieh Huang (National Yang Ming Chiao Tung University), I-Chen Wu (National Yang Ming Chiao Tung University)
OptimizationReinforcement Learning
🎯 What it does: This paper studies the global convergence of the PPO-Clip algorithm, establishing for the first time the global convergence results of the PPO-Clip variant in both tabular and neural network function approximation settings.
Practical Privacy-Preserving MLaaS: When Compressive Sensing Meets Generative Networks
Jia Wang (Shenzhen University), Jianqiang Li
ClassificationSafty and PrivacyConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: A lightweight privacy-preserving MLaaS framework is proposed, which combines compressed sensing, generative networks, and noise perturbation to achieve real-time inference on compressed data while ensuring the privacy of data and labels.
Predicting Real-World Penny Auction Durations by Integrating Game Theory and Machine Learning
Yujia Wang (Beijing Institute of Technology), Haoran Yu (Beijing Institute of Technology)
Mixture of ExpertsTabular
🎯 What it does: A three-stage framework (ADAPT) is proposed, which first predicts auction duration using game theory, then utilizes a multi-branch mixture density network combined with product description embeddings and game theory predictions, and finally outputs the actual auction duration distribution.
PrefAce: Face-Centric Pretraining with Self-Structure Aware Distillation
Siyuan Hu (Nanyang Technological University), Yew Soon Ong (Nanyang Technological University)
RecognitionKnowledge DistillationRepresentation LearningTransformerContrastive LearningVideo
🎯 What it does: This paper studies the unlabeled video face pre-training framework PrefAce, which utilizes multi-scale keypoint-guided self-distillation and instance-level updates of the FaceFeat Cache to learn general and transferable face representations.
PREFER: Prompt Ensemble Learning via Feedback-Reflect-Refine
Chenrui Zhang (Meituan Inc), Mingchen Cai (Meituan Inc)
ClassificationOptimizationTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: An automatic and jointly optimized Prompt integration framework called PREFER is proposed, which generates and optimizes Prompts through an iterative feedback-reflection-refinement loop, and enhances stability by combining forward and backward bidirectional bagging.
Preference Aware Dual Contrastive Learning for Item Cold-Start Recommendation
Wenbo Wang (Harbin Institute of Technology), Jian Guan (Harbin Engineering University)
Recommendation SystemContrastive LearningTabular
🎯 What it does: The PAD-CLRec model is proposed, which utilizes contrastive learning to align user preferences, content features, and collaborative features, addressing the cold start recommendation problem.
Preference Ranking Optimization for Human Alignment
Feifan Song (Peking University), Houfeng Wang (Peking University)
OptimizationReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningContrastive LearningText
🎯 What it does: This paper proposes Preference Ranking Optimization (PRO), a method that directly utilizes human preference rankings to supervise the fine-tuning of large language models, addressing the complexities and sample scarcity issues of RLHF.
Preparing Lessons for Progressive Training on Language Models
Yu Pan (Harbin Institute of Technology), Qun Liu (Huawei Technologies)
Computational EfficiencyTransformerLarge Language ModelText
🎯 What it does: The Apollo scheme is proposed to achieve progressive Transformer training from scratch; it significantly reduces training FLOPs and time by learning high-level features through low-level weight sharing, combined with low-value priority sampling (LVPS) and interpolation expansion.
PreRoutGNN for Timing Prediction with Order Preserving Partition: Global Circuit Pre-training, Local Delay Learning and Attentional Cell Modeling
Ruizhe Zhong (Shanghai Jiao Tong University), Junchi Yan (Shanghai Jiao Tong University)
Graph Neural NetworkAuto EncoderGraph
🎯 What it does: This paper studies a two-stage pre-routing timing prediction model, PreRoutGNN, which can accurately estimate timing metrics (slack, slope, net delay, cell delay) before routing, thereby accelerating the IC design process.
Primitive-Based 3D Human-Object Interaction Modelling and Programming
Siqi Liu (Shanghai Jiao Tong University), Cewu Lu (Shanghai Jiao Tong University)
Object DetectionPose EstimationOptimizationImage
🎯 What it does: This paper proposes a 3D human-object interaction modeling and programming based on superquadric primitives, and constructs the corresponding P3HAOI dataset to recover the 3D shapes of interacting humans and objects from a single RGB image.
Principal-Agent Reward Shaping in MDPs
Omer Ben-Porat (Technion Israel Institute of Technology), Boaz Taitler (Technion Israel Institute of Technology)
OptimizationReinforcement Learning from Human FeedbackReinforcement Learning
🎯 What it does: The principal-agent reward shaping problem is studied under a finite budget constraint, proposing polynomial approximation and optimal algorithms for stochastic trees and deterministic finite-horizon MDPs.
Principle Component Trees and Their Persistent Homology
Ben Kizaric (University of Wisconsin Madison), Daniel Pimentel-Alarcón (University of Wisconsin Madison)
Image
🎯 What it does: This paper studies a Principal Component Tree (PCT) structure designed to capture the subspace mixing characteristics of high-dimensional data.
Prior and Prediction Inverse Kernel Transformer for Single Image Defocus Deblurring
Peng Tang (Technical University of Munich), Tobias Lasser (Technical University of Munich)
RestorationTransformerImage
🎯 What it does: This paper proposes a single image defocus deblurring method called P2IKT, which utilizes prior knowledge and predicted inverse kernel transformation to achieve a divide-and-conquer deblurring approach.
Privacy Amplification by Iteration for ADMM with (Strongly) Convex Objective Functions
T-H. Hubert Chan (University of Hong Kong), Mengshi Zhao (University of Hong Kong)
OptimizationSafty and PrivacyTabular
🎯 What it does: A private version of the (strongly) convex optimization problem based on ADMM is proposed, which only adds Gaussian noise to the primal variable, and it is proven that the more iterations, the smaller the privacy leakage (i.e., privacy amplification).
Privileged Prior Information Distillation for Image Matting
Cheng Lyu (Beijing University of Posts and Telecommunications), Yong Tang (Xpeng)
SegmentationKnowledge DistillationImage
🎯 What it does: This paper proposes a 'Privileged Prior Information Distillation' (PPID) framework specifically for image matting, which transfers the environment-aware information from a trimap-based teacher model to a trimap-free student model to enhance the quality of image matting.
ProAgent: Building Proactive Cooperative Agents with Large Language Models
Ceyao Zhang (Chinese University of Hong Kong), Yaodong Yang (Peking University)
Large Language ModelChain-of-Thought
🎯 What it does: A proactive collaboration agent framework called ProAgent based on large language models is proposed, which can achieve zero-shot collaboration without training or fine-tuning by observing and inferring teammates' intentions.
Probabilistic Neural Circuits
Pedro Zuidberg Dos Martires (Örebro University)
ClassificationRecognitionConvolutional Neural NetworkImage
🎯 What it does: A probabilistic neural circuit (PNC) is proposed, which can be interpreted as a mixture of deep Bayesian networks, along with its hierarchical construction and executable queryable form.
Probabilistic Offline Policy Ranking with Approximate Bayesian Computation
Longchao Da (Arizona State University), Hua Wei (Arizona State University)
Reinforcement LearningAgentic AITabular
🎯 What it does: The Probabilistic Offline Policy Ranking (POPR) framework is proposed, which constructs the posterior performance distribution of candidate policies using expert data to achieve offline policy ranking; at the same time, the POPR-EABC (Energy-based Approximate Bayesian Computation) method is introduced to estimate the posterior distribution.
Probabilities of Causation with Nonbinary Treatment and Effect
Ang Li (Florida State University), Judea Pearl (University of California, Los Angeles)
TabularBiomedical Data
🎯 What it does: This paper derives the theoretical upper and lower bounds of any causal probability based on structural causal models (SCM) under non-binary treatment and effects (such as PNS, PN, PS, and more generally, multi-hypothesis causal probabilities);
Probability-Polarized Optimal Transport for Unsupervised Domain Adaptation
Yan Wang (Sun Yat-Sen University), Hong Yan (City University of Hong Kong)
Domain AdaptationImage
🎯 What it does: A framework for Probabilistic Polarization Optimal Transport (PPOT) is proposed for unsupervised domain adaptation, which explicitly distinguishes intra-class and inter-class transport through probabilistic polarization regularization, combined with dynamic thresholds to achieve a clearer transport plan.
ProCC: Progressive Cross-Primitive Compatibility for Open-World Compositional Zero-Shot Learning
Fushuo Huo (Hong Kong Polytechnic University), Xiaocheng Lu (Hong Kong University of Science and Technology)
ClassificationObject DetectionConvolutional Neural NetworkImage
🎯 What it does: ProCC network is proposed for open-world and partially supervised compositional zero-shot learning.
Procedural Level Generation with Diffusion Models from a Single Example
Shiqi Dai (Tsinghua University), Zhi Wang (Tsinghua University)
GenerationConvolutional Neural NetworkDiffusion modelMultimodality
🎯 What it does: A method for unconditional procedural level generation is proposed, training a latent diffusion model on a single example to achieve style-consistent generation of levels of any size.
Progressive Distillation Based on Masked Generation Feature Method for Knowledge Graph Completion
Cunhang Fan (Anhui University), Zhao Lv (Tsinghua University)
CompressionKnowledge DistillationTransformerLarge Language ModelGraph
🎯 What it does: A progressive distillation method based on mask-generated features (PMD) is proposed to compress the parameter size of pre-trained language models in knowledge graph completion tasks while maintaining or even improving performance.
Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation
Jingxuan He (Zhejiang Lab), Mingli Song (Zhejiang University)
SegmentationKnowledge DistillationTransformerImage
🎯 What it does: This paper proposes a single-stage weakly supervised semantic segmentation framework that enhances the features of uncertain and confident regions in images through adaptive masking and self-distillation techniques.
Progressive High-Frequency Reconstruction for Pan-Sharpening with Implicit Neural Representation
Ge Meng (Xiamen University), Yue Huang (Xiamen University)
RestorationSuper ResolutionImage
🎯 What it does: A novel super-resolution fusion network based on implicit feature fusion (PIF-Net) is proposed, which achieves high-quality pyramid sharpening of multispectral images by gradually injecting multi-scale high-frequency signals.
Progressive Painterly Image Harmonization from Low-Level Styles to High-Level Styles
Li Niu (Shanghai Jiao Tong University), Liqing Zhang (Shanghai Jiao Tong University)
Image TranslationImage HarmonizationConvolutional Neural NetworkRecurrent Neural NetworkAuto EncoderImage
🎯 What it does: This paper proposes an evolutionary multi-stage painting style image harmonization network called ProPIH, which can gradually transition from low-level styles (color, simple textures) to high-level styles (complex textures), achieving a natural fusion of the foreground and the painting style background.
Progressive Poisoned Data Isolation for Training-Time Backdoor Defense
Yiming Chen (University of Macau), Jiantao Zhou (University of Macau)
ClassificationAnomaly DetectionConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: A backdoor defense framework called PIPD is proposed, which utilizes progressive data poisoning isolation and selective training to build a clean model.
Progressive Text-to-Image Diffusion with Soft Latent Direction
YuTeng Ye, Wei Yang (Huazhong University of Science and Technology)
GenerationData SynthesisTransformerLarge Language ModelPrompt EngineeringDiffusion modelImageText
🎯 What it does: This paper proposes an evolutionary text-to-image generation framework called SRF based on large language models, which can progressively synthesize, edit, and erase multi-entity images while ensuring spatial relationships.
Progressively Knowledge Distillation via Re-parameterizing Diffusion Reverse Process
Xufeng Yao (Chinese University of Hong Kong), Bei Yu (Chinese University of Hong Kong)
ClassificationKnowledge DistillationConvolutional Neural NetworkTransformerDiffusion modelImage
🎯 What it does: This paper proposes a method for knowledge distillation through a reparameterized diffusion inverse process, using multi-step diffusion decomposition to gradually approximate the student distribution to the teacher distribution.
Project-Fair and Truthful Mechanisms for Budget Aggregation
Rupert Freeman (University of Virginia), Ulrike Schmidt-Kraepelin (TU Eindhoven)
🎯 What it does: A new movable phantom mechanism, the Ladder mechanism, is proposed and analyzed in the context of the allocatable budget aggregation problem, proving that it has optimal or approximately optimal theoretical bounds in terms of project fairness and ℓ₁ distance to the mean.
Prompt to Transfer: Sim-to-Real Transfer for Traffic Signal Control with Prompt Learning
Longchao Da (Arizona State University), Hua Wei (Arizona State University)
Autonomous DrivingOptimizationGraph Neural NetworkTransformerLarge Language ModelReinforcement LearningPrompt EngineeringTabularTime SeriesChain-of-Thought
🎯 What it does: This paper proposes PromptGAT, a grounded action transformation method that utilizes large language models (LLM) for sim-to-real transfer in traffic signal control (TSC) tasks, enabling better modeling of road dynamics without the need to collect large amounts of real data.
Prompt-Based Distribution Alignment for Unsupervised Domain Adaptation
Shuanghao Bai (Xi'an Jiaotong University), Badong Chen (Xi'an Jiaotong University)
Domain AdaptationTransformerPrompt EngineeringVision Language ModelImageMultimodality
🎯 What it does: A Prompt-based Distribution Alignment (PDA) method is proposed, which utilizes two branches (the base branch and the alignment branch) for prompt learning in visual-language models, achieving unsupervised domain adaptation.
Prompting Multi-Modal Image Segmentation with Semantic Grouping
Qibin He (University of Chinese Academy of Sciences)
SegmentationTransformerPrompt EngineeringImageMultimodality
🎯 What it does: A parameter-efficient visual prompt tuning framework named GoPT is proposed, which utilizes semantic grouping to complete multimodal image segmentation tasks on a frozen pre-trained RGB model.
Prompting Segmentation with Sound Is Generalizable Audio-Visual Source Localizer
Yaoting Wang (Renmin University of China), Xi Li (Zhejiang University)
SegmentationContrastive LearningMultimodalityAudio
🎯 What it does: By constructing Semantic-Aware Audio Prompts (SAP) and Correlation Adapters (ColA), audio information is injected as prompts into the visual foundation model SAM, enabling audio-driven semantic segmentation.
PromptMRG: Diagnosis-Driven Prompts for Medical Report Generation
Haibo Jin (Hong Kong University of Science and Technology), Hao Chen (Hong Kong University of Science and Technology)
GenerationRetrievalTransformerPrompt EngineeringTextBiomedical DataElectronic Health RecordsRetrieval-Augmented Generation
🎯 What it does: A medical report generation framework based on diagnostic-driven prompts (PromptMRG) is designed to convert disease classification results into token prompts to guide text generation.
Propagation Tree Is Not Deep: Adaptive Graph Contrastive Learning Approach for Rumor Detection
Chaoqun Cui (Beijing Jiaotong University), Caiyan Jia (Beijing Jiaotong University)
ClassificationGraph Neural NetworkContrastive LearningGraph
🎯 What it does: An adaptive graph contrastive learning framework RAGCL is proposed, targeting the wide structural characteristics of rumor propagation trees, utilizing node centrality to guide data augmentation to highlight subtrees with high information content.
Proportional Aggregation of Preferences for Sequential Decision Making
Nikhil Chandak (International Institute of Information Technology Hyderabad), Dominik Peters (National Centre for Scientific Research)
TabularSequential
🎯 What it does: Under the given voter preference conditions, a proportional representation rule for sequential decision-making is proposed, and strong EJR, strong PJR, and other fairness axioms are defined;
Proportional Representation in Metric Spaces and Low-Distortion Committee Selection
Yusuf Kalayci (University of Southern California), Vikram Kher (Yale University)
🎯 What it does: A new fairness definition for representative sets in metric spaces is proposed, and it is proven that the EXPANDING APPROVALS RULE can achieve approximate proportional representation, core fairness, and proportional fair clustering within a resource augmentation framework using only ordinal information, with a constant factor guarantee.
Prot2Text: Multimodal Protein’s Function Generation with GNNs and Transformers
Hadi Abdine (Ecole Polytechnique), Michalis Vazirgiannis (Ecole Polytechnique)
GenerationProtein Structure PredictionGraph Neural NetworkTransformerLarge Language ModelMultimodality
🎯 What it does: The Prot2Text model is proposed, which transforms the protein function prediction task into free text generation, utilizing multimodal information (protein sequences, 3D structures, text annotations) to generate detailed functional descriptions.
Protect Your Score: Contact-Tracing with Differential Privacy Guarantees
Rob Romijnders (University of Amsterdam), Max Welling (University of Amsterdam)
Safty and PrivacyTabular
🎯 What it does: Designed and evaluated a decentralized contact tracing algorithm with differential privacy (DPFN) that significantly reduces epidemic peaks without disclosing individual risk scores.
Protein 3D Graph Structure Learning for Robust Structure-Based Protein Property Prediction
Yufei Huang (Zhejiang University), Stan Z. Li (Westlake University)
Protein Structure PredictionGraph Neural NetworkContrastive LearningGraph
🎯 What it does: This paper addresses the issue of embedding bias in the prediction accuracy of protein structures predicted using tools like AlphaFold2. It proposes the SAO (Structure Embedding Alignment Optimization) framework, which enhances the robustness of property predictions by aligning predicted structures with experimental structures in the embedding space.
Provably Convergent Federated Trilevel Learning
Yang Jiao (Tongji University), Jianwei Huang (Chinese University of Hong Kong)
Domain AdaptationOptimizationFederated LearningHyperparameter SearchImageTabular
🎯 What it does: An Asynchronous Federated Three-Layer Optimization (AFTO) algorithm is proposed for efficiently solving three-layer optimization problems in a distributed environment.
Provably Powerful Graph Neural Networks for Directed Multigraphs
Béni Egressy (ETH Zurich), Kubilay Atasu (IBM Research Europe)
Graph Neural NetworkGraphFinance Related
🎯 What it does: This paper transforms the standard information propagation graph neural network by incorporating three modifications: reverse message passing, port numbering, and ego IDs, making it provably capable of detecting any directed multigraph subgraph pattern.
ProxyDet: Synthesizing Proxy Novel Classes via Classwise Mixup for Open-Vocabulary Object Detection
Joonhyun Jeong (NAVER Cloud), Heesu Kim (Korea Advanced Institute of Science and Technology)
Object DetectionVision Language ModelImage
🎯 What it does: By using linear combinations of base classes to generate proxy new categories during the training phase, and training the object detector with proxy loss, the generalization ability of open vocabulary object detection is enhanced.
Proxyformer: Nyström-Based Linear Transformer with Trainable Proxy Tokens
Sangho Lee (Sungkyunkwan University), Dongkun Shin (Sungkyunkwan University)
Computational EfficiencyTransformerContrastive LearningTextBenchmark
🎯 What it does: A new linear Transformer called Proxyformer has been developed, utilizing trainable proxy tokens as Nyström sampling points to achieve self-attention linearization, thereby improving the efficiency of long sequence processing.
PRP Rebooted: Advancing the State of the Art in FOND Planning
Christian Muise (Queen's University), J. Christopher Beck (University of Toronto)
OptimizationBenchmark
🎯 What it does: This paper presents PR2, a new FOND planner that employs a dual structure of repeated planning and controller/reachable states to address strong cyclic planning problems.
PSC-CPI: Multi-Scale Protein Sequence-Structure Contrasting for Efficient and Generalizable Compound-Protein Interaction Prediction
Lirong Wu (Westlake University), Stan Z. Li (Westlake University)
Drug DiscoveryRecurrent Neural NetworkGraph Neural NetworkContrastive LearningGraphBiomedical Data
🎯 What it does: This paper proposes a multi-scale protein sequence-structure contrastive learning framework (PSC-CPI) for predicting the interaction patterns and affinities between compounds and proteins, addressing the issues of modality missing and domain transfer.
Pseudo-Label Calibration Semi-supervised Multi-Modal Entity Alignment
Luyao Wang (Renmin University of China), Biao Qin (Renmin University of China)
Representation LearningGraph Neural NetworkContrastive LearningMultimodality
🎯 What it does: A semi-supervised pseudo-label calibration multimodal entity alignment framework PCMEA is proposed, which utilizes feature extraction and fusion from four modalities: visual, structural, relational, and attribute, combining pseudo-label calibration and momentum contrast learning to enhance entity alignment performance.
PTMQ: Post-training Multi-Bit Quantization of Neural Networks
Ke Xu (Anhui University), Xingyi Zhang (Anhui University)
CompressionOptimizationKnowledge DistillationConvolutional Neural NetworkTransformerImage
🎯 What it does: This paper proposes a post-training multi-bit quantization framework (PTMQ) that enables real-time switching between different bit widths with just one calibration, supporting both uniform and mixed precision.
PTUS: Photo-Realistic Talking Upper-Body Synthesis via 3D-Aware Motion Decomposition Warping
Luoyang Lin (Sun Yat-sen University), Xiaochun Cao (Sun Yat-sen University)
GenerationData SynthesisPose EstimationOptical FlowImageVideo
🎯 What it does: This paper proposes a method called PTUS for 'talking upper-body synthesis' that can simultaneously generate realistic upper body and facial animations from source images and driving videos.
Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention
Ying-Yu Chen (National Yang Ming Chiao Tung University), Ming-Ching Chang (University at Albany - State University of New York)
ClassificationMeta LearningConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: This paper enhances the fine-tuning effect of Few-Shot Learning (FSL) by introducing a feature enhancement module and a cross-scale attention mechanism on a pre-trained model.
PVALane: Prior-Guided 3D Lane Detection with View-Agnostic Feature Alignment
Zewen Zheng (Guangdong University of Technology), Xiaochen Yuan (Macao Polytechnic University)
Object DetectionAutonomous DrivingConvolutional Neural NetworkTransformerImage
🎯 What it does: Developed a 3D lane detection framework PVALane based on front-view priors, achieving alignment of sparse prior anchor points with view-invariant features.
QAGait: Revisit Gait Recognition from a Quality Perspective
Zengbin Wang (Beijing University of Posts and Telecommunications), Shibiao Xu (Beijing University of Posts and Telecommunications)
RecognitionImageVideo
🎯 What it does: This paper proposes the QAGait framework, which first conducts quality assessment of gait profiles (maximum connected area, template matching), followed by tilt-aware alignment and data augmentation, and finally introduces quality-aware QACE and QATriplet losses during the feature learning phase to achieve full-process control of gait recognition quality.
QCS-SGM+: Improved Quantized Compressed Sensing with Score-Based Generative Models
Xiangming Meng (Zhejiang University), Yoshiyuki Kabashima (University of Tokyo)
RestorationCompressionDiffusion modelScore-based ModelImage
🎯 What it does: This paper upgrades the existing quantized compressed sensing algorithm QCS-SGM to QCS-SGM+, achieving effective recovery of general (non-orthogonal) measurement matrices.
QDETRv: Query-Guided DETR for One-Shot Object Localization in Videos
Yogesh Kumar (Indian Institute of Technology Jodhpur), Roshni Ramnani (Accenture Labs)
Object DetectionObject TrackingTransformerVideo
🎯 What it does: This paper studies one-shot video object localization based on query images and proposes the QDETRv model.
QI-IRA: Quantum-Inspired Interactive Ranking Aggregation for Person Re-identification
Chunyu Hu (Wuhan University), Hao Huang (Wuhan University)
RecognitionRetrievalImageVideo
🎯 What it does: A quantum-inspired interactive ranking aggregation (QI-IRA) method is proposed for multi-method ranking aggregation in person re-identification tasks.
QLABGrad: A Hyperparameter-Free and Convergence-Guaranteed Scheme for Deep Learning
Minghan Fu (University of Saskatchewan), Fang-Xiang Wu (University of Saskatchewan)
OptimizationHyperparameter SearchConvolutional Neural NetworkImage
🎯 What it does: A hyperparameter-free learning rate adaptation scheme called QLABGrad is proposed, which automatically determines the optimal learning rate at each step of gradient descent using quadratic loss approximation and requires only one additional forward pass.
QPEN: Quantum Projection and Quantum Entanglement Enhanced Network for Cross-Lingual Aspect-Based Sentiment Analysis
Xingqiang Zhao (Sun Yat-sen University), Kunxun Qi (Sun Yat-sen University)
ClassificationRecognitionTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: In cross-lingual aspect-based sentiment analysis (cross-lingual ABSA), the QPEN model is proposed, which enhances multilingual BERT using quantum projection and quantum entanglement modules to improve multilingual sentiment recognition performance.
Quad Bayer Joint Demosaicing and Denoising Based on Dual Encoder Network with Joint Residual Learning
Bolun Zheng (Hangzhou Dianzi University), Chenggang Yan (Hangzhou Dianzi University)
RestorationConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: An end-to-end dual encoder network (DRNet) is proposed for the joint denoising and demosaicing of Quad Bayer CFA images.
Quality-Diversity Generative Sampling for Learning with Synthetic Data
Allen Chang (University of Southern California), Stefanos Nikolaidis (University of Southern California)
RecognitionGenerationData SynthesisPrompt EngineeringGenerative Adversarial NetworkContrastive LearningImage
🎯 What it does: A generative sampling framework QDGS based on quality diversity optimization is proposed, which uses language prompts to guide the generator to uniformly cover the multi-dimensional attribute space in the latent space, generating balanced synthetic data.
Quantifying and Analyzing Entity-Level Memorization in Large Language Models
Zhenhong Zhou (Beijing University of Posts and Telecommunications), Sen Su (Beijing University of Posts and Telecommunications)
Safty and PrivacyTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Proposes a definition of memory at the entity level and an efficient extraction method, assessing the risk of memory leakage at the entity level in large language models.
Quantum Interference Model for Semantic Biases of Glosses in Word Sense Disambiguation
Junwei Zhang (Hangzhou Institute of Medicine), Chang Liu (Tianjin Normal University)
ClassificationRecognitionTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposes a multi-view word sense representation based on quantum superposition states and uses a quantum interference model to calculate the probability of word sense recognition.
Quantum-Inspired Neural Network with Runge-Kutta Method
Zipeng Fan (Tianjin University), Hui Gao (Tianjin University)
ClassificationRecognitionComputational EfficiencyRecurrent Neural NetworkTextStochastic Differential EquationOrdinary Differential Equation
🎯 What it does: This paper proposes embedding a quantum-similar high-order Runge-Kutta (QRK) method into a Quantum-Inspired Neural Network (QINN) to enhance its training efficiency and performance.
QuerySum: A Multi-Document Query-Focused Summarization Dataset Augmented with Similar Query Clusters
Yushan Liu (Fudan University), Ruifeng Yuan (Hong Kong Polytechnic University)
GenerationRetrievalTransformerSupervised Fine-TuningTextBenchmark
🎯 What it does: Created the QuerySum multi-document query-focused summarization dataset, which includes 27,041 non-factual (What/How/Why) question-answer samples, and provides similar query clusters and corresponding summaries for each query.
Question Calibration and Multi-Hop Modeling for Temporal Question Answering
Chao Xue (Beihang University), Jing Zhang (Zhejiang University)
Graph Neural NetworkTransformerGraphTime Series
🎯 What it does: This paper proposes a model QC-MHM for temporal knowledge graph question answering, which mainly improves answer reasoning accuracy through question calibration and multi-hop modeling.
R3CD: Scene Graph to Image Generation with Relation-Aware Compositional Contrastive Control Diffusion
Jinxiu Liu (South China University of Technology), Qi Liu (South China University of Technology)
GenerationData SynthesisGraph Neural NetworkTransformerDiffusion modelContrastive LearningImageGraph
🎯 What it does: This study explores how to utilize large-scale diffusion models combined with abstract relationships in scene graphs to generate images.
Racing Control Variable Genetic Programming for Symbolic Regression
Nan Jiang (Purdue University), Yexiang Xue (Purdue University)
OptimizationTabularPhysics Related
🎯 What it does: This paper presents Racing-CVGP, a control variable genetic program that conducts parallel competition and early stopping in experimental scheduling during the symbolic regression process.
RadarMOSEVE: A Spatial-Temporal Transformer Network for Radar-Only Moving Object Segmentation and Ego-Velocity Estimation
Changsong Pang (Northwestern Polytechnical University), Yuwei Cheng (Tsinghua University)
SegmentationAutonomous DrivingTransformerPoint Cloud
🎯 What it does: This paper proposes a Transformer network that utilizes millimeter-wave radar point clouds to simultaneously achieve moving object segmentation (MOS) and ego-vehicle velocity estimation (EVE).
RadOcc: Learning Cross-Modality Occupancy Knowledge through Rendering Assisted Distillation
Haiming Zhang (Chinese University of Hong Kong Shenzhen), Zhen Li (Chinese University of Hong Kong Shenzhen)
SegmentationAutonomous DrivingKnowledge DistillationMultimodalityPoint Cloud
🎯 What it does: Utilizing a multi-modal teacher model, cross-modal knowledge distillation is achieved for a single-modal visual student model through differentiable voxel rendering, enhancing 3D occupancy prediction.
Rating-Based Reinforcement Learning
Devin White (University of Texas), Yongcan Cao (University of Texas)
Reinforcement Learning from Human FeedbackReinforcement LearningTabular
🎯 What it does: A reinforcement learning framework based on multi-class scoring (RbRL) is proposed, which learns the reward function by having humans provide absolute rank scores for a single trajectory.
Reachability of Fair Allocations via Sequential Exchanges
Ayumi Igarashi (University of Tokyo), Sheung Man Yuen (National University of Singapore)
🎯 What it does: This paper studies the reachability problem of EF1 (envy-freeness up to one item) allocation in fair distribution, defines the EF1 exchange graph, and explores its connectivity and the existence of optimal paths.
READ-PVLA: Recurrent Adapter with Partial Video-Language Alignment for Parameter-Efficient Transfer Learning in Low-Resource Video-Language Modeling
Thong Nguyen (National University of Singapore), Anh Tuan Luu (Nanyang Technological University)
Recurrent Neural NetworkTransformerVideoText
🎯 What it does: The READ-PVLA framework is proposed, which uses a lightweight recursive adapter and partial video-language alignment objectives to perform parameter-efficient fine-tuning of pre-trained Transformers for low-resource video-language tasks.
Real3D: The Curious Case of Neural Scene Degeneration
Dengsheng Chen (Meituan), Enhua Wu (University of Macau)
GenerationData SynthesisOptimizationDiffusion modelScore-based ModelNeural Radiance FieldGenerative Adversarial NetworkImageText
🎯 What it does: A text-prompt-based 3D scene generation model called Real3D is proposed, which generates higher quality 3D objects by simultaneously using implicit rendering (NeRF) and explicit rendering (DMTET);
Recall-Oriented Continual Learning with Generative Adversarial Meta-Model
Haneol Kang (Inha University), Dong-Wan Choi (Inha University)
Knowledge DistillationMeta LearningGenerative Adversarial NetworkImage
🎯 What it does: A memory-oriented continual learning framework is proposed, which uses a working memory network to learn new tasks and recalls old knowledge at the parameter level through a Generative Adversarial Meta Model (GAMM), addressing the stability-plasticity dilemma.
Recasting Regional Lighting for Shadow Removal
Yuhao Liu (City University of Hong Kong), Rynson W.H. Lau (City University of Hong Kong)
RestorationDiffusion modelImage
🎯 What it does: A shadow-aware decomposition network and a bidirectional correction network are proposed, which first restore the local illumination of the shadow region and then guide texture recovery with the restored illumination, achieving high-quality shadow removal.
Recognizing Ultra-High-Speed Moving Objects with Bio-Inspired Spike Camera
Junwei Zhao (Peking University), Tiejun Huang (Peking University)
RecognitionObject DetectionConvolutional Neural NetworkImageVideo
🎯 What it does: A theoretical analysis of motion blur caused by ultra-high-speed motion is proposed, and a robust peak representation method is designed based on spatiotemporal context learning, along with the construction of the first real ultra-high-speed peak recognition dataset UHSR.
Reconciling Predictive and Statistical Parity: A Causal Approach
Drago Plecko (Columbia University), Elias Bareinboim (Columbia University)
Tabular
🎯 What it does: This paper unifies the conflict between Statistical Parity (SP) and Predictive Parity (PP) through causal decomposition methods and proposes a spectrum of 'Business Necessity' (BN), providing methods for judgment and implementation.
Rectangle Search: An Anytime Beam Search
Sofia Lemons (University of New Hampshire), Carlos Linares Lopez (Universidad Carlos III de Madrid)
Optimization
🎯 What it does: A new anytime heuristic search algorithm called Rectangle Search is proposed, which is based on beam search with increasing width/depth exploration, aimed at solving deep local minimum problems.
Recurrent Graph Neural Networks and Their Connections to Bisimulation and Logic
Maximilian Pflueger (University of Oxford), Egor V. Kostylev (University of Oslo)
Graph Neural NetworkGraph
🎯 What it does: This paper proposes and systematically studies two types of recursive graph neural networks, RecGNN and GSGNN, establishing their strict expressive power hierarchy relative to ordinary GNNs, and provides a semantic and logical representation centered around bisimulation.
Recurrent Partial Kernel Network for Efficient Optical Flow Estimation
Henrique Morimitsu (University of Science and Technology Beijing), Xu-Cheng Yin (Tsinghua University)
Computational EfficiencyConvolutional Neural NetworkRecurrent Neural NetworkOptical FlowImageVideo
🎯 What it does: This paper proposes RPKNet, an efficient optical flow estimation model that utilizes spatial recursive encoders, significantly reducing the number of parameters and computational load while maintaining accuracy.
RedCore: Relative Advantage Aware Cross-Modal Representation Learning for Missing Modalities with Imbalanced Missing Rates
Jun Sun (Institute of Artificial Intelligence Zhejiang Lab), Taihao Li (Institute of Artificial Intelligence Zhejiang Lab)
Representation LearningAuto EncoderVideoTextMultimodalityAudio
🎯 What it does: This study addresses the issue of missing modalities and the imbalance of different missing rates in multimodal learning, proposing the RedCore model to achieve cross-modal representation learning and adaptively adjust the supervisory weights of each modality.
Redefining ABA+ Semantics via Abstract Set-to-Set Attacks
Yannis Dimopoulos (University of Cyprus), Stefan Woltran (Vienna University of Technology)
🎯 What it does: The researchers proposed Hyper Argumentation Frameworks (HYPAFs) and proved that they can serve as an abstract graphical representation of Assumption-Based Argumentation with Preferences (ABA+). Subsequently, they redefined the complete and grounded semantics (Θ-com and Θ-grd) based on HYPAFs, ensuring that ABA+ guarantees at least one extension in all frameworks, and analyzed its complexity.
Reducing Spatial Fitting Error in Distillation of Denoising Diffusion Models
Shengzhe Zhou (Zhejiang University), Lingyun Sun (Alibaba Group)
GenerationKnowledge DistillationDiffusion modelImage
🎯 What it does: A diffusion model distillation framework called SFERD is proposed, which is based on attention guidance and semantic gradient prediction, to reduce the spatial fitting error between the teacher and student models, thereby achieving high-quality image generation in very few steps (even a single step).
Referred by Multi-Modality: A Unified Temporal Transformer for Video Object Segmentation
Shilin Yan (Fudan University), Peng Gao (China Telecom Corporation Ltd.)
Object DetectionSegmentationTransformerVideoTextMultimodalityAudio
🎯 What it does: This paper proposes MUTR, a unified multimodal temporal Transformer for video object segmentation based on language or audio instructions.
Refined Characterizations of Approval-Based Committee Scoring Rules
Chris Dong (Technical University of Munich), Patrick Lederer (Technical University of Munich)
🎯 What it does: This study provides a complete axiomatic characterization of two important scoring rules in the election of the Approval Base Committee (ABC) - the Thiele rule and the BSAV rule;