AAAI 2025 Papers — Page 23
AAAI Conference on Artificial Intelligence · 3028 papers
Prompt Tuning In a Compact Attribute Space
Shiyu Hou (Beijing Institute of Technology), Guoren Wang (Beijing Institute of Technology)
ClassificationRecognitionExplainability and InterpretabilityTransformerLarge Language ModelPrompt EngineeringVision Language ModelImageMultimodality
🎯 What it does: A new prompt tuning method called PTinCAS is proposed, aimed at improving the performance of visual-language models through a compact attribute space for prompt tuning.
Prompt-based Unifying Inference Attack on Graph Neural Networks
Yuecen Wei (Beihang University), Chunming Hu (Guangxi Normal University)
Adversarial AttackGraph Neural NetworkPrompt EngineeringContrastive LearningGraph
🎯 What it does: A unified prompt-based reasoning attack framework named ProIA has been designed and implemented for attribute inference and membership inference attacks on Graph Neural Networks (GNNs); it retains graph topology information through pre-training, generates attack samples using prompts, and incorporates a disentanglement mechanism in downstream tasks to enhance attack effectiveness.
Prompt-SID: Learning Structural Representation Prompt via Latent Diffusion for Single Image Denoising
Huaqiu Li (Tsinghua University), Haoqian Wang (Tsinghua University)
RestorationTransformerPrompt EngineeringDiffusion modelImage
🎯 What it does: A self-supervised single image denoising framework based on prompt learning, Prompt-SID, is proposed, which utilizes structural representation generated diffusion (RG-Diff) and a Structural Attention Module (SAM) to preserve details and eliminate structural loss caused by sampling.
Promptable Anomaly Segmentation with SAM Through Self-Perception Tuning
Hui-Yue Yang (Tsinghua University), Guiguang Ding (Tsinghua University)
SegmentationAnomaly DetectionSupervised Fine-TuningPrompt EngineeringImage
🎯 What it does: A Self-Perception Tuning (SPT) method is proposed, which enhances the perception and generalization capabilities of SAM in industrial defect segmentation tasks through two steps: self-sketch tuning and a visual relationship perception adapter.
Promptable Representation Distribution Learning and Data Augmentation for Gigapixel Histopathology WSI Analysis
Kunming Tang (Beihang University), Yushan Zheng (Beihang University)
ClassificationRepresentation LearningTransformerContrastive LearningBiomedical Data
🎯 What it does: This paper proposes the Promptable Representation Distribution Learning (PRDL) framework, which combines prompt-based representation distribution estimation with Promptable Representation Sampling (PRS) to achieve feature space data augmentation for pathological whole-slide images (WSI) and subsequently perform WSI classification.
PromptDet: A Lightweight 3D Object Detection Framework with LiDAR Prompts
Kun Guo (University of Science and Technology of China), Qiang Ling (University of Science and Technology of China)
Object DetectionAutonomous DrivingPrompt EngineeringMultimodalityPoint Cloud
🎯 What it does: The PromptDet framework is proposed for multi-camera 3D detection, which injects LiDAR point clouds as prompts into a baseline camera detector using a small number of parameters, achieving lightweight multimodal fusion.
PromptHaze: Prompting Real-world Dehazing via Depth Anything Model
Tian Ye (Hong Kong University of Science and Technology), Lei Zhu (Hong Kong University of Science and Technology)
RestorationDepth EstimationPrompt EngineeringImage
🎯 What it does: By utilizing depth prompts generated by the Depth Anything model and employing a Prompt-by-Prompt strategy for iterative prompt updates, controllable real-world image dehazing is achieved.
Proof Simulation via Round-based Strategy Extraction for QBF
Leroy Chew (TU Wien)
🎯 What it does: This paper formalizes the extraction of round strategies from QBF proof systems and proves that eFrege + ∀ red can polynomially simulate LD-Q(D rrs)-Res, thereby solving an open problem in proof complexity.
Proportional Representation in Practice: Quantifying Proportionality in Ordinal Elections
Tuva Bardal (University of Warwick), Jannik Peters (National University of Singapore)
Tabular
🎯 What it does: A quantifiable measure of proportionality is proposed, and the proportional representation of multi-vote methods is evaluated using real Scottish local election data.
Proportionally Fair Makespan Approximation
Michal Feldman (Tel Aviv University), Tomasz Ponitka (University of Illinois at Urbana Champaign)
Optimization
🎯 What it does: In the unrelated machine scheduling problem, a payable proportional fairness mechanism is proposed and analyzed, proving that it can achieve a 3/2 approximation on optimal completion time (normalized instances can achieve optimal).
Proportionally Fair Matching via Randomized Rounding
Sharmila Duppala (University of Maryland), Aravind Srinivasan (University of Maryland)
OptimizationGraph Neural NetworkGraph
🎯 What it does: The paper proposes an approximate algorithm for the Proportional Fair Matching problem on edge-colored graphs, which achieves approximate fairness for each color edge through randomized sampling while maximizing the matching weight.
ProPose: Probabilistic 3D Human Pose Estimation with Instance-Level Distribution and Normalizing Flow
Jumin Han (Korea University), Seong-Whan Lee (Korea University)
Pose EstimationFlow-based ModelAuto EncoderImage
🎯 What it does: ProPose framework is proposed, implementing multi-hypothesis inference from 2D pose to 3D pose using instance-level Gaussian distribution and regularized flow.
PROSAC: Provably Safe Certification for Machine Learning Models under Adversarial Attacks
Chen Feng (University College London), Miguel Rodrigues (University College London)
ClassificationSafty and PrivacyAdversarial AttackHyperparameter SearchTransformerImage
🎯 What it does: This paper proposes the PROSAC framework, which uses hypothesis testing and Bayesian optimization for distribution-independent security certification of machine learning models under adversarial attacks.
ProsodyFM: Unsupervised Phrasing and Intonation Control for Intelligible Speech Synthesis
Xiangheng He (Imperial College London), Björn Schuller (Technical University of Munich)
GenerationData SynthesisTransformerFlow-based ModelAudio
🎯 What it does: An unsupervised ProsodyFM speech synthesis model is proposed, which can accurately control phrase breaks and terminal pitch, thereby enhancing intelligibility.
ProsodyTalker: 3D Visual Speech Animation via Prosody Decomposition
Zonglin Li (Harbin Institute of Technology), Shengping Zhang (Harbin Institute of Technology)
GenerationData SynthesisPose EstimationTransformerDiffusion modelVideoMultimodality
🎯 What it does: Proposes the ProsodyTalker framework, which utilizes speech prosody decomposition to simultaneously generate lip movements and head motions, resulting in realistic 3D audiovisual animations.
ProtCLIP: Function-Informed Protein Multi-Modal Learning
Hanjing Zhou (Zhejiang University), Zheng Wang (Alibaba Cloud Computing)
Protein Structure PredictionTransformerContrastive LearningTextMultimodality
🎯 What it does: Constructed the ProtCLIP multimodal protein-text pre-training framework, combined with the ProtAnno dataset for large-scale training.
Protecting Model Adaptation from Trojans in the Unlabeled Data
Lijun Sheng (University of Science and Technology of China), Tieniu Tan (Chinese Academy of Sciences)
ClassificationDomain AdaptationAdversarial AttackImage
🎯 What it does: This study investigates the use of unlabelled target data to implant backdoor attacks during the model adaptation process and its defense methods.
ProtoArgNet: Interpretable Image Classification with Super-Prototypes and Argumentation
Hamed Ayoobi (Imperial College London), Francesca Toni (Cardiff University)
ClassificationExplainability and InterpretabilityConvolutional Neural NetworkImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: This paper proposes ProtoArgNet, an interpretable image classification model based on super-prototypes and argumentation, which can capture the spatial relationships of classes within a single super-prototype and provide support/attack explanations.
ProtoCar: Learning 3D Vehicle Prototypes from Single-View and Unconstrained Driving Scene Images
Hongyuan Liu (University of Science and Technology Beijing), Huimin Ma (University of Science and Technology Beijing)
GenerationAutonomous DrivingGaussian SplattingImagePoint Cloud
🎯 What it does: A single-view vehicle 3D reconstruction framework called ProtoCar is designed based on 3D Gaussian splatting, capable of generating high-quality, real-time renderable 3D models of vehicles from a single view image of real driving scenes that may contain occlusions.
ProtoOcc: Accurate, Efficient 3D Occupancy Prediction Using Dual Branch Encoder-Prototype Query Decoder
Jungho Kim (Seoul National University), Jun Won Choi (Hanyang University)
SegmentationAutonomous DrivingConvolutional Neural NetworkTransformerPoint Cloud
🎯 What it does: ProtoOcc is a 3D occupancy prediction model based on a dual-branch encoder and a prototype query decoder.
Prototype-Guided Multimodal Relation Extraction based on Entity Attributes
Zefan Zhang (Jilin University), Tian Bai (Jilin University)
ClassificationRecognitionRepresentation LearningTransformerLarge Language ModelImageTextMultimodality
🎯 What it does: A prototype-guided multimodal relation extraction framework based on entity attributes (PG-MRE) is proposed, which generates entity explanations through a large language model, utilizes attribute prototype modules and relation prototype modules to extract and aggregate attribute information, and generates compact and discriminative multimodal representations for predicting relationships between entities.
Prototypical Calibrating Ambiguous Samples for Micro-Action Recognition
Kun Li (Hefei University of Technology), Meng Wang (Zhejiang University)
RecognitionContrastive LearningVideoMultimodality
🎯 What it does: This paper proposes the Prototypical Calibrating Ambiguous Network (PCAN), which identifies and calibrates ambiguous samples in micro-action recognition through hierarchical prototype learning.
Prototypical Replay with Old-class Focusing Knowledge Distillation for Incremental Named Entity Recognition
Zesheng Liu (Renmin University of China), Hong Chen (Renmin University of China)
RecognitionKnowledge DistillationTransformerSupervised Fine-TuningText
🎯 What it does: A method for incremental named entity recognition based on Prototypical Replay and Old-Class-Focused Knowledge Distillation (POF) is proposed, which alleviates catastrophic forgetting by saving prototypes of old entity types and periodically replaying them.
Provable Discriminative Hyperspherical Embedding for Out-of-Distribution Detection
Zhipeng Zou (Nanjing University of Science and Technology), Chen Gong (Shanghai Jiao Tong University)
ClassificationAnomaly DetectionConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: The DHE (Provable Discriminative Hyperspherical Embedding) framework is proposed, which first maximizes the global class prototype distance through angular spread loss, and then uses prototype-enhanced contrastive (PEC) loss to bring ID samples closer to their corresponding prototypes, thereby obtaining a more discriminative embedding space for OOD detection.
Provably Secure Image Robust Steganography via Cross-modal Error Correction
Yuang Qi (University of Science and Technology of China), Weiming Zhang (University of Science and Technology of China)
GenerationCompressionSafty and PrivacyConvolutional Neural NetworkLarge Language ModelImageText
🎯 What it does: A provably secure and robust image steganography system CMSTEG is designed, utilizing the autoregressive (AR) image generation model LlamaGen and VQ tokenizer to embed secret information into the generated image token sequences. Robust extraction against lossy processing in social networks (such as JPEG compression) is achieved through discrete token optimization and cross-modal error correction (image to text).
Pruning Large Language Models with Semi-Structural Adaptive Sparse Training
Weiyu Huang (Tsinghua University), Jianfei Chen (Tsinghua University)
CompressionComputational EfficiencyKnowledge DistillationTransformerLarge Language ModelText
🎯 What it does: The Adaptive Sparse Trainer (AST) framework is proposed, achieving efficient compression of large language models through semi-structured sparse training and knowledge distillation.
PScalpel: A Machine Learning-based Guider for Protein Phase-Separating Behaviour Alteration
Jia Wang (Shenzhen University), Jianqiang Li (Sun Yat-sen University)
OptimizationProtein Structure PredictionGraph Neural NetworkContrastive LearningBiomedical Data
🎯 What it does: PScalpel is proposed, a tool that utilizes machine learning to assess and recommend changes in protein phase transition behavior.
Pseudo Informative Episode Construction for Few-Shot Class-Incremental Learning
Chaofan Chen (Institute of Automation, Chinese Academy of Sciences), Changsheng Xu (Institute of Automation, Chinese Academy of Sciences)
ClassificationMeta LearningConvolutional Neural NetworkImage
🎯 What it does: The Pseudo Informative Episode Construction (PIEC) framework is proposed, which generates pseudo new categories through distribution-level mixing on the base class Gaussian distribution, and selects information-rich pseudo new categories based on new-new diversity and base-new diversity criteria to construct pseudo incremental tasks for training.
PSMGD: Periodic Stochastic Multi-Gradient Descent for Fast Multi-Objective Optimization
Mingjing Xu (Rochester Institute of Technology), Haibo Yang (Rochester Institute of Technology)
OptimizationGraph Neural NetworkTabular
🎯 What it does: A periodic updating dynamic weight multi-objective optimization algorithm PSMGD is proposed, which utilizes gradient fusion to quickly solve multi-objective problems.
PSReg: Prior-guided Sparse Mixture of Experts for Point Cloud Registration
Xiaoshui Huang (Shanghai Jiao Tong University), Yuming Fang
RecognitionTransformerMixture of ExpertsPoint Cloud
🎯 What it does: A point cloud registration method called PSReg based on Prior-guided SMoE is proposed.
Public Opinion Field Effect and Hawkes Process Join Hands for Information Popularity Prediction
Junliang Li (Tianjin University), Hong Gao (Zhejiang Normal University)
Representation LearningRecurrent Neural NetworkGraph Neural NetworkTransformerTime SeriesSequential
🎯 What it does: A prediction model for information heat based on the Hawkes process that integrates public opinion field effects is proposed.
Pushing the Limits of BFP on Narrow Precision LLM Inference
Hui Wang (Southeast University), Zhe Jiang (Southeast University)
Computational EfficiencyTransformerLarge Language ModelImageText
🎯 What it does: This paper proposes the DBFP format and the DB-Attn framework, achieving hardware-software co-acceleration for nonlinear operations such as Softmax in the attention layer of LLMs;
Putting People in LLMs’ Shoes: Generating Better Answers via Question Rewriter
Junhao Chen (Osaka University), Yuta Nakashima (Osaka University)
GenerationOptimizationTransformerLarge Language ModelReinforcement LearningPrompt EngineeringText
🎯 What it does: Improving the answer quality of black-box LLMs in long-text question-answering tasks through question rewriting, addressing the issue of imprecise answers caused by vague user questions.
PVTree: Realistic and Controllable Palm Vein Generation for Recognition Tasks
Sheng Shang (Hefei University of Technology), Wei Jia (Hefei University of Technology)
RecognitionGenerationData SynthesisConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: Proposes the PVTree framework, which first generates a 3D vascular tree using an improved CCO algorithm, then projects it into multi-view vascular patterns, and subsequently converts it into realistic finger vein images through the improved PCE-Palm, achieving identity consistency and diversity;
Q-MAML: Quantum Model-Agnostic Meta-Learning for Variational Quantum Algorithms
Junyong Lee (Yonsei University), Shiho Kim (Yonsei University)
OptimizationMeta LearningGraphPhysics Related
🎯 What it does: A Q-MAML framework based on MAML is proposed, which provides adaptive initialization for parameterized quantum circuits using a classical neural network Learner, significantly accelerating the convergence of VQE;
QCS:Feature Refining from Quadruplet Cross Similarity for Facial Expression Recognition
Chengpeng Wang (Wisesoft Inc.), Xuebin Lv (Sichuan University)
RecognitionKnowledge DistillationTransformerContrastive LearningImage
🎯 What it does: A four-branch symmetric network QCS is proposed, which refines facial expression features through cross-similarity attention, achieving suppression of redundant features.
QiMeng-GEMM: Automatically Generating High-Performance Matrix Multiplication Code by Exploiting Large Language Models
Qirui Zhou (Institute of Computing Technology, Chinese Academy of Sciences), Yunji Chen (Institute of Computing Technology, Chinese Academy of Sciences)
OptimizationAI Code AssistantTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Designed and implemented a prompt mechanism based on large language models, QiMeng-GEMM, to automatically generate high-performance matrix multiplication (GEMM) code.
QiMLP: Quantum-inspired Multilayer Perceptron with Strong Correlation Mining and Parameter Compression
Junwei Zhang (Hangzhou Institute of Medicine, Chinese Academy of Sciences), Xiaolin Li (Hangzhou Institute of Medicine, Chinese Academy of Sciences)
ClassificationCompressionImageText
🎯 What it does: A QiMLP based on quantum many-body systems is constructed, which reconstructs layer weights and biases through the Kronecker product to achieve feature strong correlation mining and parameter compression, and its effectiveness is validated in image, text sentiment, and text classification tasks.
QJL: 1-Bit Quantized JL Transform for KV Cache Quantization with Zero Overhead
Amir Zandieh (Google Research), Insu Han (KAIST)
CompressionOptimizationLarge Language ModelText
🎯 What it does: A 1-bit quantization method QJL for LLM KV caching is proposed, providing an unbiased estimator without additional storage overhead.
QORT-Former: Query-optimized Real-time Transformer for Understanding Two Hands Manipulating Objects
Elkhan Ismayilzada (Michigan State University), Seungryul Baek (Ulsan National Institute of Science and Technology)
Pose EstimationTransformerImage
🎯 What it does: This paper proposes QORT-Former, a real-time framework based on Transformer for estimating the 3D poses of two hands and objects from a single RGB image.
Qsco: A Quantum Scoring Module for Open-Set Supervised Anomaly Detection
Yifeng Peng (Stevens Institute of Technology), Ying Wang (Stevens Institute of Technology)
Anomaly DetectionBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: A quantum scoring module Qsco is proposed and implemented, embedding variational quantum circuits into traditional deep models to enhance the performance of open-set supervised anomaly detection.
Qua2SeDiMo: Quantifiable Quantization Sensitivity of Diffusion Models
Keith G. Mills (University of Alberta), Di Niu (Huawei Technologies)
GenerationData SynthesisOptimizationGraph Neural NetworkDiffusion modelImageText
🎯 What it does: A mixed-precision post-training quantization framework named Qua SeDiMo 2 is proposed, which achieves weight quantization below 4 bits for different layers, operations, and block structures in diffusion models without using a calibration dataset, and provides an interpretable sensitivity analysis.
Quality over Quantity: Boosting Data Efficiency Through Ensembled Multimodal Data Curation
Jinda Xu (Shanghai Jiao Tong University), Qinya Li (Shanghai Jiao Tong University)
OptimizationData-Centric LearningImageTextMultimodality
🎯 What it does: The EcoDatum framework is proposed for quality-guided deduplication, single/multi-modal filtering, and weakly supervised integration in multi-stage cleaning of large-scale web-crawled image-text data.
Quantified Linear and Polynomial Arithmetic Satisfiability via Template-based Skolemization
Krishnendu Chatterjee (Institute of Science and Technology Austria), Đorđe Žikelić (Vienna University of Technology)
Benchmark
🎯 What it does: A method for determining the satisfiability of quantified linear and polynomial arithmetic based on templated Skolemization is proposed, which can efficiently eliminate quantifiers and generate witnesses.
Quantitative Predictive Monitoring and Control for Safe Human-Machine Interaction
Shuyang Dong (University of Virginia), Lu Feng (University of Virginia)
Autonomous DrivingSafty and PrivacyRecurrent Neural NetworkTime SeriesBiomedical Data
🎯 What it does: This paper proposes a method for uncertainty quantification monitoring and adaptive control based on Signal Temporal Logic (STL) to enhance the safety of human-computer interaction.
Quantum Best Arm Identification with Quantum Oracles
Xuchuang Wang (University of Massachusetts), Don Towsley (University of Massachusetts)
Physics Related
🎯 What it does: This paper studies the best arm identification (BAI) problem under weak quantum oracles and m-constrained quantum oracles, and proposes two quantum algorithms: Q-Elim and Q-Part.
Quantum-inspired Non-homologous Representation Constraint Mechanism for Long-tail Senses of Word Sense Disambiguation
Junwei Zhang (Hangzhou Institute of Medicine, Chinese Academy of Sciences), Xiaolin Li (Hangzhou Institute of Medicine, Chinese Academy of Sciences)
RecognitionTransformerLarge Language ModelText
🎯 What it does: A long-tail word sense disambiguation (WSD) system called QiWSD is proposed and implemented, which is based on a quantum-inspired heterogeneous representation constraint mechanism, combining traditional high-frequency sense recognition with quantum-constrained low-frequency sense recognition.
QuARF: Quality-Adaptive Receptive Fields for Degraded Image Perception
Fei Gao (Xidian University), Nannan Wang (Xidian University)
RestorationSegmentationGenerationConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: A pluggable quality adaptive receptive field module QUARF is proposed, which can automatically select multi-scale convolution kernels based on the degradation level of the input image, thereby enhancing the perception and generation effects for low-quality images.
Queries, Representation & Detection: The Next 100 Model Fingerprinting Schemes
Augustin Godinot (University of Rennes), Gilles Tredan
ClassificationObject DetectionAnomaly DetectionImageBenchmark
🎯 What it does: The study investigates model fingerprinting techniques to detect cases of model theft and proposes a simple AKH baseline and QuRD framework to systematize fingerprint design.
Query Quantized Neural SLAM
Sijia Jiang (Wayne State University), Zhizhong Han (Wayne State University)
OptimizationRobotic IntelligenceSimultaneous Localization and MappingImage
🎯 What it does: This paper proposes a Query Quantized Neural SLAM, which discretizes queries such as coordinates, position encoding, geometric features, and TSDF priors into quantized codes. It uses quantized queries to quickly approximate neural implicit surfaces (SDF and color) and achieves real-time SLAM through joint optimization of camera pose, SDF, color function, and codebook.
Query-centric Audio-Visual Cognition Network for Moment Retrieval, Segmentation and Step-Captioning
Yunbin Tu (University of Chinese Academy of Sciences), Qingming Huang (University of Chinese Academy of Sciences)
SegmentationRetrievalTransformerContrastive LearningVideoMultimodalityAudio
🎯 What it does: A query-centered audio-visual cognitive network QUAG is proposed for instant retrieval, segmentation, and step description in videos.
Query-efficient Attack for Black-box Image Inpainting Forensics via Reinforcement Learning
Xianbo Mo (Shenzhen MSU-BIT University), Jiwu Huang (Shenzhen University)
RestorationAdversarial AttackConvolutional Neural NetworkReinforcement LearningImage
🎯 What it does: This paper proposes a reinforcement learning-based query-efficient black-box image inpainting forensic attack framework (RLGC), which can effectively attack black-box forensic models with very few queries.
Queryable Prototype Multiple Instance Learning with Vision-Language Models for Incremental Whole Slide Image Classification
Jiaxiang Gou (University of Electronic Science and Technology of China), Mao Ye (University of Electronic Science and Technology of China)
ClassificationTransformerVision Language ModelImageBiomedical Data
🎯 What it does: A queryable prototype multi-instance learning framework based on a visual-language model (QPMIL-VL) is proposed, achieving bufferless incremental whole slide image classification.
Quickest Detection of Adversarial Attacks Against Correlated Equilibria
Kiarash Kazari (KTH Royal Institute of Technology), György Dán
Anomaly DetectionOptimizationAdversarial AttackGraph
🎯 What it does: The study utilizes rapid change detection to timely identify and suppress situations where public signals are attacked under Cooperative Equilibrium (CE), and proposes a detection strategy based on generalized CUSUM.
R-DTI: Drug Target Interaction Prediction Based on Second-Order Relevance Exploration
Yang Hua (Jiangnan University), Xiaojun Wu (Jiangnan University)
Drug DiscoveryGraph Neural NetworkBiomedical Data
🎯 What it does: This paper proposes a drug-target interaction prediction framework R-DTI based on second-order correlation exploration, combining protein structure and drug bimodal features, utilizing Riemannian space learning to enhance representation and make predictions based on second-order correlation.
R^2-Art: Category-Level Articulation Pose Estimation from Single RGB Image via Cascade Render Strategy
Li Zhang (Hefei Institute of Physical Science, Chinese Academy of Sciences), Liu Liu (Hefei University of Technology)
Pose EstimationConvolutional Neural NetworkImage
🎯 What it does: This study proposes an R-ART2 framework that achieves category-level decoupled object 6D pose estimation using a single RGB image, completely without the need for depth information.
RA-GAR: A Richly Annotated Benchmark for Gait Attribute Recognition
Chenye Wang (Beijing Normal University), Yongzhen Huang (Beijing Normal University)
RecognitionTransformerContrastive LearningVideoBenchmark
🎯 What it does: A large-scale and attribute-rich gait attribute dataset RA-GAR (533 individuals, 120k+ sequences, 15 attributes) was collected and constructed, and a two-stage CLIP-GAR method was proposed for gait attribute recognition.
RA-SGG: Retrieval-Augmented Scene Graph Generation Framework via Multi-Prototype Learning
Kanghoon Yoon (Korea Advanced Institute of Science and Technology), Chanyoung Park (Korea University)
GenerationRetrievalGraph Neural NetworkGraphRetrieval-Augmented Generation
🎯 What it does: This paper proposes the Retrieval-Augmented Scene Graph Generation (RA-SGG) framework, which transforms the scene graph generation task into a partially labeled multi-label classification problem and enhances labels by retrieving potential fine-grained predicates.
RaCMC: Residual-Aware Compensation Network with Multi-Granularity Constraints for Fake News Detection
Xinquan Yu (Sun Yat-sen University), Jiantao Zhou (University of Macau)
ClassificationAnomaly DetectionConvolutional Neural NetworkTransformerLarge Language ModelContrastive LearningTextMultimodality
🎯 What it does: This paper proposes RaCMC, which utilizes a residual-aware compensation network and multi-granularity constraints to achieve sufficient interaction and fusion of cross-modal features for detecting fake news.
RaDIO: Real-Time Hallucination Detection with Contextual Index Optimized Query Formulation for Dynamic Retrieval Augmented Generation
Jia Zhu (Zhejiang Normal University), Pasquale De Meo (University of Messina)
GenerationRetrievalTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: This paper presents RaDIO, a dynamic retrieval-augmented generation framework that determines when to retrieve and optimizes queries using multi-head attention through real-time hallucination detection.
Radiology Report Generation via Multi-objective Preference Optimization
Ting Xiao (East China University of Science and Technology), Chenjia Bai (China Telecom)
GenerationOptimizationTransformerReinforcement LearningImageText
🎯 What it does: A multi-objective preference optimization (MPO) framework is proposed to generate radiology reports based on the preferences of different radiologists.
RAGG: Retrieval-Augmented Grasp Generation Model
Zhenhua Tang (Hefei University of Technology), Richang Hong (Hefei University of Technology)
GenerationRobotic IntelligenceTransformerDiffusion modelPoint CloudRetrieval-Augmented Generation
🎯 What it does: This paper proposes a retrieval-augmented grasp generation model (RAGG), which includes a retrieval-guided diffusion model (ReDim) and a structurally stable improvement network (PRN) for intent-based 3D hand grasp generation.
RAP-SR: RestorAtion Prior Enhancement in Diffusion Models for Realistic Image Super-Resolution
Jiangang Wang (Sun Yat-sen University), Wenqi Ren (Fuzhou University)
RestorationSuper ResolutionDiffusion modelImage
🎯 What it does: The RAP-SR method is proposed, which achieves higher quality reconstruction of real image super-resolution by enhancing the recovery prior of the pre-trained diffusion model.
Rapid Learning in Constrained Minimax Games with Negative Momentum
Zijian Fang (Sun Yat-sen University), Chaohao Hu (Sun Yat-sen University)
OptimizationReinforcement LearningTabular
🎯 What it does: This paper studies and implements the application of Negative Momentum in constrained minimax games, proposing a negative momentum update framework that can seamlessly integrate with classical algorithms (OMD, FTRL, RM+, etc.), and extends it to generalized limit games and extensive form games (EFG).
Rare Event Detection in Imbalanced Multi-Class Datasets Using an Optimal MIP-Based Ensemble Weighting Approach
Georgios Tertytchny (KIOS Research and Innovation Center of Excellence), Maria K. Michael (KIOS Research and Innovation Center of Excellence)
Anomaly DetectionOptimizationTabular
🎯 What it does: This paper proposes a weighted voting ensemble method based on Mixed Integer Programming (MIP) combined with elastic net regularization for rare event detection in critical cyber-physical systems, aimed at selecting a predetermined number of classifiers and assigning class-level weights in imbalanced multi-class datasets.
RAT: Adversarial Attacks on Deep Reinforcement Agents for Targeted Behaviors
Fengshuo Bai (Shanghai Jiao Tong University), Yaodong Yang (Peking University)
Adversarial AttackRobotic IntelligenceReinforcement LearningSequential
🎯 What it does: This paper proposes a general attack framework named RAT, which can achieve precise targeted attacks on the behavior of deep reinforcement learning (DRL) agents by making subtle perturbations to their observations without changing the target rewards.
RATT: A Thought Structure for Coherent and Correct LLM Reasoning
Jinghan Zhang (Portland State University), Kunpeng Liu (University of Kansas)
GenerationRetrievalAI Code AssistantTransformerLarge Language ModelTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: A method for LLM reasoning called RATT is proposed and implemented, which integrates retrieval enhancement and tree of thought structure to improve factual accuracy and logical coherence.
RAZOR: Sharpening Knowledge by Cutting Bias with Unsupervised Text Rewriting
Shuo Yang (Technical University of Munich), Gjergji Kasneci (Technical University of Munich)
TransformerLarge Language ModelText
🎯 What it does: The RAZOR method is proposed, utilizing large language models for unsupervised text rewriting to eliminate shortcut biases in the dataset and enhance the model's generalization ability.
RCTrans: Radar-Camera Transformer via Radar Densifier and Sequential Decoder for 3D Object Detection
Yiheng Li (Chinese Academy of Sciences), Zhen Lei (Chinese Academy of Sciences)
Object DetectionAutonomous DrivingTransformerMultimodalityPoint Cloud
🎯 What it does: A query-based radar-camera Transformer framework RCTrans is proposed, which addresses the issue of sparse noise in radar point clouds through a radar dense encoder and a stepwise pruning decoder, achieving 3D object detection.
RDPI: A Refine Diffusion Probability Generation Method for Spatiotemporal Data Imputation
Zijin Liu (Beihang University), You Song (Beihang University)
Graph Neural NetworkDiffusion modelTime Series
🎯 What it does: A two-stage refined diffusion probability imputation method (RDPI) is proposed, which first uses a deterministic model to roughly impute missing values, and then refines it through a conditional diffusion model with the residuals as the target, significantly improving the quality of spatiotemporal data imputation.
Re-Attentional Controllable Video Diffusion Editing
Yuanzhi Wang (Nanjing University of Science and Technology), Antoni B. Chan (City University of Hong Kong)
GenerationData SynthesisDiffusion modelVideoText
🎯 What it does: This paper proposes a video diffusion editing framework based on re-attention, ReAtCo, which can precisely control the spatial position and quantity of foreground objects in a video through text prompts without training the model, while keeping the background unchanged.
Re2LLM: Reflective Reinforcement Large Language Model for Session-based Recommendation
Ziyan Wang (Nanyang Technological University), Jie Zhang (Nanyang Technological University)
Recommendation SystemTransformerLarge Language ModelReinforcement LearningPrompt EngineeringTextChain-of-Thought
🎯 What it does: The Re2LLM method is proposed, which generates professional knowledge prompts through self-reflection of LLM and uses a lightweight retriever to select prompts via reinforcement learning to enhance conversational recommendation performance.
ReactGPT: Understanding of Chemical Reactions via In-Context Tuning
Zhe Chen (East China Normal University), Man Lan (East China Normal University)
GenerationRetrievalTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextRetrieval-Augmented Generation
🎯 What it does: The ReactGPT framework is proposed to align chemical reactions with natural language text and defines a new task for generating reaction descriptions.
Read, Watch and Scream! Sound Generation from Text and Video
Yujin Jeong (NAVER AI Lab), Jiyoung Lee (NAVER AI Lab)
GenerationData SynthesisDiffusion modelVideoTextMultimodalityAudio
🎯 What it does: This paper proposes ReWaS, an audio generation framework that combines video and text control, utilizing the energy information from video as temporal control to generate audio that is synchronized with the video and aligns with the text description using a pre-trained text-to-audio model.
Ready for You When You Are Back: Content-Driven Session-Based Recommendation for Continuity of Experience
Brijraj Singh (Sony Research India), Niranjan Pedanekar (Sony Research India)
Recommendation SystemRecurrent Neural NetworkLarge Language ModelTextSequential
🎯 What it does: A conversation partitioning method based on content homogeneity is proposed to enhance the continuous experience and accuracy of conversational recommendation systems.
Real-Time Calibration Model for Low-Cost Sensor in Fine-Grained Time Series
Seokho Ahn (Inha University), Young-Duk Seo (Inha University)
TransformerTime Series
🎯 What it does: This paper proposes a real-time calibration model for low-cost sensors, TESLA, aimed at achieving a balance of high accuracy, high speed, and low energy consumption;
Real-Time Neural Denoising with Render-Aware Knowledge Distillation
Mengxun Kong (Nanjing University), Yanwen Guo (Nanjing University)
RestorationKnowledge DistillationConvolutional Neural NetworkGenerative Adversarial NetworkImageVideo
🎯 What it does: A denoising framework for real-time Monte Carlo rendering has been designed and implemented—Render-Aware Knowledge Distillation (RAKD), which compresses a large teacher network into a lightweight student network while maintaining high-quality denoising effects.
Real-Time Recurrent Reinforcement Learning
Julian Lemmel (Vienna University of Technology), Radu Grosu (Vienna University of Technology)
Meta LearningRecurrent Neural NetworkReinforcement LearningSequential
🎯 What it does: A real-time recursive reinforcement learning framework (RTRRL) based on biological interpretability is proposed, which combines the Meta-RL structure of the basal ganglia, TD(λ) reinforcement learning, and RFLO/RTRL that can compute gradients online, to solve partially observable Markov decision process (POMDP) tasks.
RealisHuman: A Two-Stage Approach for Refining Malformed Human Parts in Generated Images
Benzhi Wang (Chinese Academy of Sciences), Zhen Lei (Chinese Academy of Sciences)
RestorationGenerationPose EstimationDiffusion modelImage
🎯 What it does: In the human body images generated by diffusion models, a two-stage local reconstruction and seamless fusion is performed on deformed local areas such as hands and faces.
RealisID: Scale-Robust and Fine-Controllable Identity Customization via Local and Global Complementation
Zhaoyang Sun (Wuhan University of Technology), Shengwu Xiong (Interdisciplinary Artificial Intelligence Research Institute)
Image TranslationGenerationPose EstimationDiffusion modelImage
🎯 What it does: This paper proposes the RealistID framework, which utilizes local and global branches to customize identity in Stable Diffusion, achieving fine-grained control over facial details, posture, expressions, and facial positioning, while maintaining high identity fidelity even with small face sizes.
Realistic Noise Synthesis with Diffusion Models
Qi Wu (Megvii Technology Inc), Shuaicheng Liu (University of Electronic Science and Technology of China)
RestorationGenerationData SynthesisDiffusion modelImage
🎯 What it does: This paper proposes a real noise synthesis method based on diffusion models, RNSD, which generates RGB noise that conforms to different camera settings through conditional encoding, and utilizes the generated noise to enhance the performance of image denoising models.
RealPortrait: Realistic Portrait Animation with Diffusion Transformers
Zejun Yang (Huawei), Zhisheng Wang (Tencent)
Image TranslationGenerationTransformerDiffusion modelImageVideo
🎯 What it does: Proposes the RealPortrait framework, which uses a Diffusion Transformer to transfer facial expressions and poses from driving videos to a single portrait photo, generating realistic dynamic videos.
Reasoning About Actual Causes in Nondeterministic Domains
Shakil M. Khan (University of Regina), Maryam Rostamigiv (York University)
🎯 What it does: The concepts of 'Possibly Causes' and 'Certainly Causes' are proposed and formalized in the domain of non-determinism, and the scenario operator is extended to reason about these causal relationships within the framework of situation calculus.
Reasoning over Uncertain Text by Generative Large Language Models
Aliakbar Nafar (Michigan State University), Parisa Kordjamshidi (Michigan State University)
TransformerLarge Language ModelPrompt EngineeringTextChain-of-Thought
🎯 What it does: This study investigates the challenges of LLMs in text reasoning that includes probabilistic information, proposes a new dataset called BLInD, and designs various prompting and symbolic methods to enhance reasoning capabilities.
Rebalancing Multi-Label Class-Incremental Learning
Kaile Du (Southeast University), Guangcan Liu (Southeast University)
ClassificationKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: This paper addresses the issue of positive and negative sample imbalance in multi-label incremental learning and proposes the RebLL framework to achieve positive-negative balance.
Reconsidering Feature Structure Information and Latent Space Alignment in Partial Multi-label Feature Selection
Hanlin Pan (Jilin University), Wanfu Gao
Tabular
🎯 What it does: A partial multi-label feature selection method based on latent space alignment (PMLFSLA) is proposed, which utilizes feature space information for denoising and denoising of labels.
Reconstruction Target Matters in Masked Image Modeling for Cross-Domain Few-Shot Learning
Ran Ma (Huazhong University of Science and Technology), Ruixuan Li (Huazhong University of Science and Technology)
Domain AdaptationRepresentation LearningTransformerContrastive LearningImage
🎯 What it does: This study investigates the impact of the reconstruction objective of Masked Image Modeling (MIM) on transfer performance in Cross-Domain Few-Shot Learning (CDFSL) and proposes Domain-agnostic Masked Image Modeling (DAMIM), which reconstructs by aggregating multi-layer features and introduces a lightweight decoder to balance the removal of low-level domain information and the preservation of global structure.
Recording for Eyes, Not Echoing to Ears: Contextualized Spoken-to-Written Conversion of ASR Transcripts
Jiaqing Liu (Alibaba Group), Wen Wang (Alibaba Group)
GenerationData SynthesisTransformerLarge Language ModelPrompt EngineeringTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: This paper proposes the context-aware spoken-to-written (CoS2W) task, aimed at correcting recognition errors and grammatical mistakes in ASR transcriptions, as well as rewriting colloquial text into a formal written style. It also constructs a cross-lingual (Chinese-English) and cross-scenario (meetings, podcasts, lectures) SWAB dataset for training and evaluation with human-corrected targets.
Recoverable Compression: A Multimodal Vision Token Recovery Mechanism Guided by Text Information
Yi Chen (University of Chinese Academy of Sciences), Cheng-Lin Liu (University of Chinese Academy of Sciences)
CompressionComputational EfficiencyTransformerLarge Language ModelVision Language ModelTextMultimodality
🎯 What it does: A training-independent multimodal visual token recovery mechanism is proposed, which dynamically filters and recovers important visual tokens using textual information, ultimately compressing the number of visual tokens to about 10% of the original without significant performance loss.
Recoverable Facial Identity Protection via Adaptive Makeup Transfer Adversarial Attacks
Xiyao Liu (Central South University), Hui Fang (Loughborough University)
RecognitionGenerationAdversarial AttackGenerative Adversarial NetworkImage
🎯 What it does: A recoverable makeup transfer generative adversarial network (RMT-GAN) has been developed, which generates adversarial facial images that can deceive unauthorized facial recognition systems while being recoverable by authorized systems.
Rectified Lagrangian for Out-of-Distribution Detection in Modern Hopfield Networks
Ryo Moriai (Denso IT Laboratory), Ikuro Sato (Institute of Science)
Anomaly DetectionImageOrdinary Differential Equation
🎯 What it does: The Rectified Lagrangian (RecLag) is proposed to explicitly generate OOD point attractors in modern Hopfield networks for OOD detection.
Recursive Aggregates as Intensional Functions in Answer Set Programming: Semantics and Strong Equivalence
Jorge Fandinno (University of Nebraska Omaha), Zachary Hansen (University of Nebraska Omaha)
🎯 What it does: This paper proposes a semantic framework for transforming ASP programs with recursive aggregates into first-order logic sentences containing intensional functions, and uses this framework to study the strong equivalence of programs; by constructing transformations τ and γ, the problem of strong equivalence determination is reduced to classical first-order logic reasoning.
Reducing AUV Energy Consumption Through Dynamic Sensor Directions Switching via Deep Reinforcement Learning
Jiawei Liu (Jilin University), Lu Jiang (Dalian Maritime University)
OptimizationRobotic IntelligenceReinforcement Learning
🎯 What it does: This paper proposes a dynamic switching method for AUV sensor direction and radius based on deep reinforcement learning (RECS). By dividing the environment into grids, assigning weights to edges, and only perceiving high-weight edges, it reduces sensor energy consumption.
Reducing Divergence in Batch Normalization for Domain Adaptation
Ellen Yi-Ge (Carnegie Mellon University), Zhenghan Chen (Microsoft)
Domain AdaptationConvolutional Neural NetworkImage
🎯 What it does: To address the estimation bias problem of Batch Normalization (BN) during training and inference in Unsupervised Domain Adaptation (UDA), we propose the Refined Batch Normalization (RBN) module, which replaces the BN layers in the bottleneck of residual networks with RBNBlock to reduce the accumulation of estimation bias caused by the stacking of BN layers, thereby improving cross-domain transfer performance.
Reducing Leximin Fairness to Utilitarian Optimization
Eden Hartman (Bar-Ilan University), Erel Segal-Halevi (Ariel University)
Optimization
🎯 What it does: This paper proposes a general reduction scheme: given any usable utilitarian optimizer, one can obtain the leximin optimal (or approximate) probability distribution in the expected sense.
RefDetector: A Simple Yet Effective Matching-based Method for Referring Expression Comprehension
Yabing Wang (Xi'an Jiaotong University), Le Wang (Xi'an Jiaotong University)
RecognitionObject DetectionTransformerContrastive LearningImageMultimodality
🎯 What it does: This paper proposes a concise and efficient Referring Expression Comprehension framework called RefDetector based on matching.
ReFF: Reinforcing Format Faithfulness in Language Models Across Varied Tasks
Jiashu Yao (Beijing Institute of Technology), Yuhang Guo (Beijing Institute of Technology)
Reinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextBenchmark
🎯 What it does: The FORMATBENCH benchmark is proposed to comprehensively evaluate the format fidelity of LLMs, and the REFF method is utilized to enhance the format adherence capability of LLMs within a reinforcement learning framework using a decidable format checker, while maintaining or improving overall quality.
Refine then Classify: Robust Graph Neural Networks with Reliable Neighborhood Contrastive Refinement
Shuman Zhuang (Fuzhou University), Ximeng Liu (Hong Kong Baptist University)
ClassificationAdversarial AttackGraph Neural NetworkContrastive LearningGraph
🎯 What it does: This paper proposes a new robustness framework for graph neural networks called GRANCE, which enhances resistance to structural attacks by performing reliable contrastive learning on the local neighborhood of the graph.
Region-aware Difference Distilling with Attribute-guided Contrastive Regularization for Change Captioning
Rong Li (Hangzhou Dianzi University), Chenggang Yan (Hangzhou Dianzi University)
GenerationKnowledge DistillationTransformerContrastive LearningImageText
🎯 What it does: This paper studies a change description model based on region-aware differential distillation networks and attribute-guided contrastive regularization, aiming to capture changes in image pairs more precisely and generate natural language descriptions.
Region-Based Optimization in Continual Learning for Audio Deepfake Detection
Yujie Chen (Anhui University), Xiaohui Zhang (Institute of Automation Chinese Academy of Sciences)
Anomaly DetectionOptimizationBenchmarkAudio
🎯 What it does: A continuous learning method called Region-Based Optimization (RegO) is proposed for audio deepfake detection.