AAAI 2026 Papers — Page 31
AAAI Conference on Artificial Intelligence · 4149 papers
Real-time 3D Object Detection with Inference-Aligned Learning
Chenyu Zhao (Wuhan University), Nan Xue (Ecole Polytechnique Federale De Lausanne)
Object DetectionKnowledge DistillationConvolutional Neural NetworkPoint Cloud
🎯 What it does: Propose the SR3D framework to achieve real-time indoor point cloud 3D object detection, bridging the training-inference gap through spatial-prioritized label assignment and self-distillation mechanisms.
Real-Time Path Planning for UAVs in Windy Environments Without Computational Fluid Dynamics
Abhudaya Shrivastava (Temple University), Zoran Obradovic (Temple University)
OptimizationComputational EfficiencyRobotic IntelligenceGraph Neural NetworkPoint CloudMesh
🎯 What it does: Proposes a Graphlets-based Zero-Shot Planning (GZS) framework for drones to real-time perceive and avoid aerodynamic risks in wind-perturbed environments, entirely without relying on CFD or any training process.
ReAlign: Text-to-Motion Generation via Step-Aware Reward-Guided Alignment
Wanjiang Weng (Southeast University), Hongsong Wang (Southeast University)
GenerationTransformerDiffusion modelScore-based ModelContrastive LearningMultimodalityStochastic Differential Equation
🎯 What it does: Proposes ReAlign, a plug-and-play method that employs reward-guided sampling in diffusion-based text-to-motion generation.
Realism Control One-step Diffusion for Real-world Image Super Resolution
Zongliang Wu (Zhejiang University), Xin Yuan (Westlake University)
Super ResolutionPrompt EngineeringDiffusion modelScore-based ModelImage
🎯 What it does: Proposed the RCOD framework, enabling single-step diffusion models to flexibly control the trade-off between fidelity and realism in real image super-resolution tasks.
Realistic Face Reconstruction from Facial Embeddings via Diffusion Models
Dong Han (Huawei Technologies), Joachim Denzler (Friedrich Schiller University)
GenerationSafty and PrivacyDiffusion modelImage
🎯 What it does: Propose a privacy-preserving facial reconstruction framework called FEM, which maps arbitrary facial embeddings to a pre-trained IPA-FaceID diffusion model to generate high-resolution, realistic facial images, and verifies its attack effectiveness in traditional FR and PPFR systems.
Reality vs Counterfactual: Multi-World Contrastive Reinforcement Learning for Enhancing MLLM’s Theory of Mind in Egocentric Videos
Guiyang Hou (Zhejiang University), Weiming Lu (Zhejiang University)
Large Language ModelReinforcement LearningVision Language ModelContrastive LearningVideoMultimodality
🎯 What it does: Propose Multi-World Contrastive Reinforcement Learning (MWCRL), which encourages multimodal large language models to infer user goals, beliefs, and next actions in first-person videos by contrasting real and shuffled human action sequences.
RealRep: Generalized SDR-to-HDR Conversion via Attribute-Disentangled Representation Learning
Li Xu (Xidian University), Yu-Wing Tai (Dartmouth)
RestorationConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: Proposed the RealRep framework, achieving inverse tone mapping from SDR to HDR.
RealUHR: Harnessing Patch-Cascade Flows for Photorealistic Ultra-High-Resolution Synthesis
Yongsheng Yu (University of Rochester), Jiebo Luo (Adobe Research)
GenerationSuper ResolutionDiffusion modelFlow-based ModelRectified FlowImage
🎯 What it does: Propose the RealUHR framework, which generates realistic images at the 4K level using Patch-Cascade Flow Matching, and supports zero-shot generation repair and creative upsampling.
RealWebAssist: A Benchmark for Long-Horizon Web Assistance with Real-World Users
Suyu Ye (Johns Hopkins University), Tianmin Shu (Johns Hopkins University)
Supervised Fine-TuningVision Language ModelImageTextMultimodalityBenchmarkAudio
🎯 What it does: Proposes RealWebAssist, a new benchmark to evaluate AI assistants' instruction-following capabilities in real user long-sequence web interactions.
REAP: Enhancing RAG with Recursive Evaluation and Adaptive Planning for Multi-Hop Question Answering
Yijie Zhu (Jiangnan University), Ning Wang (Jiangnan University)
TransformerTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose the REAP framework, combining Sub-Task Planner and Fact Extractor to iteratively enhance RAG's reasoning and fact extraction in multi-hop question answering.
ReaSon: Reinforced Causal Search with Information Bottleneck for Video Understanding
Yuan Zhou (Nanjing University of Information Science and Technology), Haoran Duan (Nanjing University of Information Science and Technology)
Reinforcement LearningVision Language ModelVideo
🎯 What it does: Proposed and implemented the ReaSon framework, which efficiently selects keyframes in videos through reinforcement learning and causal information bottleneck
Reason2Attack: Jailbreaking Text-to-Image Models via LLM Reasoning
Chenyu Zhang, Anan Liu (School Of Electrical And Information Engineering Tianjin University)
Adversarial AttackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningDiffusion modelImageTextChain-of-Thought
🎯 What it does: This paper proposes a jailbreak method based on LLM reasoning called Reason2Attack (R2A), which enhances the efficiency of exploiting security vulnerabilities in text-to-image models by generating and optimizing stealthy and effective adversarial prompts through framework semantics and two-phase training.
ReasonAct: Progressive Training for Fine-Grained Video Reasoning in Small Models
Jiaxin Liu (University of Illinois Urbana-Champaign), Zhaolu Kang (Peking University)
RecognitionSupervised Fine-TuningReinforcement LearningVideoChain-of-Thought
🎯 What it does: Enhancing video reasoning capabilities of small-scale multimodal models through three-stage progressive training.
Reasoning About the Unsaid: Misinformation Detection with Omission-Aware Graph Inference
Zhengjia Wang (Institute of Computing Technology Chinese Academy of Sciences), Juan Cao (Institute of Computing Technology Chinese Academy of Sciences)
ClassificationGraph Neural NetworkTransformerLarge Language ModelTextGraph
🎯 What it does: Propose the OMIGRAPH framework, which constructs an omission-aware graph, performs fine-grained sentence-level splitting of target news and its contextual news, infers cross-source omission relationships via LLM, and extracts omission patterns using attention-based message passing and global aggregation in graph neural networks to enhance fake news detection;
Reasoning or Memorization? Unreliable Results of Reinforcement Learning Due to Data Contamination
Mingqi Wu (Fudan University), Qi Zhang (Fudan University)
TransformerLarge Language ModelReinforcement LearningTextBenchmark
🎯 What it does: Investigated why reinforcement learning (RLVR) achieves 'pseudo reward' improvements in mathematical reasoning tasks, confirming that the primary cause is data leakage; subsequently constructed a leakage-free random arithmetic dataset called RandomCalculation, and compared the performance of Qwen and Llama series models under different reward signals.
Reasoning Shapes Alignment: Investigating Cultural Alignment in Large Reasoning Models with Cultural Norms
Yuhang Wang (Beijing Jiaotong University), Jitao Sang (Beijing Jiaotong University)
Data-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningTextBenchmarkChain-of-Thought
🎯 What it does: Propose a cultural alignment framework (CNCA) based on cultural norms, which enhances the model's reflection of different national cultural values by automatically mining cultural norms from limited questionnaire data and using them for context alignment and fine-tuning in large-scale inference models.
Reasoning Transfer for an Extremely Low-Resource and Endangered Language: Bridging Languages Through Sample-Efficient Language Understanding
Khanh-Tung Tran (University College Cork), Hoang D. Nguyen (University College Cork)
Representation LearningData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextChain-of-Thought
🎯 What it does: This paper proposes an English-Pivoted Chain-of-Thought (CoT) training method, which enhances the mathematical reasoning ability of extremely low-resource languages (e.g., Irish) by enforcing the reasoning steps to remain in English internally within the model.
Reasoning via Implicit Self-supervised Emergence for Instruction Segmentation
Qing Zhou (Northwestern Polytechnical University), Qi Wang (Northwest Institute of Nuclear Technology)
SegmentationTransformerLarge Language ModelReinforcement LearningContrastive LearningImageTextMultimodalityChain-of-Thought
🎯 What it does: Propose a mask-free self-supervised reinforcement learning framework called RISE to achieve instruction-driven image segmentation, which can spontaneously generate chain-of-thought reasoning from pure semantic rewards;
Reasoning with Exploration: An Entropy Perspective
Daixuan Cheng (Renmin University of China), Furu Wei (Microsoft Research)
TransformerLarge Language ModelReinforcement LearningText
🎯 What it does: In large language models trained with reinforcement learning, this paper proposes adding a clipped and gradient-separated entropy term to the advantage function by analyzing the association between entropy values and exploratory reasoning behaviors to guide deeper reasoning chains;
ReBoot: Encrypted Training of Deep Neural Networks with CKKS Bootstrapping
Alberto Pirillo (Politecnico di Milano), Luca Colombo (Politecnico di Milano)
Safty and PrivacyImageTabular
🎯 What it does: Propose the ReBoot framework to realize fully encrypted, non-interactive multi-layer perceptron (MLP) training.
ReCAD: Reinforcement Learning Enhanced Parametric CAD Model Generation with Vision-Language Models
Jiahao Li (Fudan University), Xiangdong Zhou (Fudan University)
GenerationTransformerSupervised Fine-TuningReinforcement LearningVision Language ModelImageTextMesh
🎯 What it does: Proposes the ReCAD framework, which jointly generates editable, parameterized high-precision CAD models using reinforcement learning and pre-trained vision-language models;
ReCast: Reliability-aware Codebook-assisted Lightweight Time Series Forecasting
Xiang Ma (Shandong University), Caiming Zhang (Shandong University)
Computational EfficiencyTime SeriesBenchmark
🎯 What it does: Proposed a lightweight time series forecasting framework ReCast, which achieves prediction by leveraging codeword quantization of local shapes and a dual-path structure.
RecCocktail: A Generalizable and Efficient Framework for LLM-Based Recommendation
Min Hou (Hefei University Of Technology), Meng Wang (Shanghai Jiao Tong University)
Recommendation SystemLarge Language ModelSupervised Fine-TuningTextTabular
🎯 What it does: Propose the RecCocktail framework, integrating global and domain-specific LLM recommendation paradigms through LoRA weight mixing to achieve model generality and domain specialization;
ReCode: Updating Code API Knowledge with Reinforcement Learning
Haoze Wu (Zhejiang University), Ningyu Zhang (Nanjing University)
AI Code AssistantReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningPrompt EngineeringTextBenchmark
🎯 What it does: This paper proposes the ReCode framework, which utilizes reinforcement learning fine-tuning to enable large language models to migrate code when facing updates to external library APIs.
Reconcile Gradient Modulation for Harmony Multimodal Learning
Xiyuan Gao (Tianjin University), Pengfei Zhu (Tianjin University)
OptimizationRepresentation LearningConvolutional Neural NetworkTransformerImageVideoMultimodalityAudio
🎯 What it does: Proposes a unified Reconcile Gradient Modulation (RGM) framework that utilizes adaptive gradient amplitude and direction modulation to simultaneously address modality imbalance and conflicts in multi-modal learning.
Reconfiguring Proportional Committees
Chris Dong (University of Potsdam), Warut Suksompong (National University of Singapore)
OptimizationComputational Efficiency
🎯 What it does: This paper investigates whether, given two committees satisfying proportional representation (JR or EJR) in an approval-based multi-choice committee, there exists a transformation path through replacing a single candidate such that every intermediate committee maintains proportional representation at each step.
Reconstruction Attack-Resistant Inference Paradigm for LLM Cloud Services
Zipeng Ye (Harbin Institute of Technology), Yubo Tang (Harbin Institute of Technology)
Safty and PrivacyAdversarial AttackTransformerLarge Language ModelText
🎯 What it does: Designed a distributed privacy-preserving inference paradigm, applying direction-preserving random scaling and compensation mechanisms in the shallow layers of the Transformer in large language models (LLMs) to prevent user-uploaded data from being reconstructed or inverted.
Reconstruction Using the Invisible: Intuition from NIR and Metadata for Enhanced 3D Gaussian Splatting
Gyusam Chang (Korea University), M. Khalid Jawed (University of California Los Angeles)
GenerationTransformerGaussian SplattingImageTextMultimodalityPoint CloudAgriculture Related
🎯 What it does: Propose the NIRPlant multimodal agricultural dataset and design the NIRSplat framework to achieve the integration of NIR, RGB, and text information in 3D Gaussian Splatting, aiming to enhance crop 3D reconstruction under sparse viewpoints.
ReconVLA: Reconstructive Vision-Language-Action Model as Effective Robot Perceiver
Wenxuan Song (Hong Kong University of Science and Technology Guangzhou), Haoang Li (Hong Kong University of Science and Technology Guangzhou)
Robotic IntelligenceTransformerLarge Language ModelVision-Language-Action ModelDiffusion modelAuto EncoderMultimodality
🎯 What it does: Designed and implemented the ReconVLA model, which implicitly guides visual attention through the reconstruction of the target grasping area, thereby enhancing the robot's localization accuracy and operational precision under multimodal inputs.
RECoRD: A Multi-Agent LLM Framework for Reverse Engineering Codebase to Relational Diagram
Yuan Xue (Ohio State University), Hoiyi Ng
Explainability and InterpretabilityAI Code AssistantTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextGraphBenchmark
🎯 What it does: Construct the RECoRD multi-agent framework, utilizing LLM reverse parsing of code to generate causal graphs.
Recovering Coherent Affective Patterns: Addressing Modality Missing in Multimodal Sentiment Analysis
Huiting Huang (Xi'an Jiaotong University), Mengling Feng (National University of Singapore)
ClassificationRestorationExplainability and InterpretabilityTransformerGenerative Adversarial NetworkMultimodality
🎯 What it does: Propose a two-stage multimodal sentiment analysis framework RECAP, addressing the common modality missing problem in real scenarios. First, a generative model is used to recover missing modalities, then self-supervised mutual information decomposition ensures semantic completeness, followed by information-guided ranking attention for modality fusion and sentiment prediction.
Rectification Reimagined: A Unified Mamba Model for Image Correction and Rectangling with Prompts
Linwei Qiu (Beihang University), Fengying Xie (Chinese University of Hong Kong)
RestorationPrompt EngineeringMixture of ExpertsOptical FlowImage
🎯 What it does: Proposed a unified image correction and rectangularization framework called UniRect, which can complete four common camera distortion correction tasks (portrait correction, wide-angle rectangularization, stitching rectangularization, and rotation correction) within the same model.
Rectified Noise: A Generative Model Using Positive-incentive Noise
Zhenyu Gu (China Telecom), Hongyuan Zhang (China Telecom)
GenerationTransformerFlow-based ModelRectified FlowImage
🎯 What it does: Introduce π-noise (forward incentive noise) into existing Rectified Flow (RF) generative models, enhancing image generation quality by injecting learned noise into the velocity field of pre-trained RF models;
Rectify Evaluation Preference: Improving LLMs’ Critique on Math Reasoning via Perplexity-aware Reinforcement Learning
Changyuan Tian (Aerospace Information Research Institute, Chinese Academy of Sciences), Guozhi Cas (Meituan)
TransformerLarge Language ModelReinforcement LearningTextBenchmark
🎯 What it does: Construct the OPS (One-to-Many Problem Solution) benchmark, revealing that LLMs tend to rate solutions with low perplexity as correct, and propose a perplexity-aware reinforcement learning method to correct evaluation biases and enhance critical thinking in multi-step mathematical reasoning.
RecToM: A Benchmark for Evaluating Machine Theory of Mind in LLM-based Conversational Recommender Systems
Mengfan Li (Huazhong University of Science and Technology), Yang Deng (Singapore Management University)
Recommendation SystemTransformerLarge Language ModelTextBenchmarkChain-of-Thought
🎯 What it does: Propose the RECTOM benchmark to evaluate the theory of mind (ToM) capabilities of large language models in conversational recommendation systems;
Reducing Goal State Divergence with Environment Design
Kelsey Sikes (Colorado State University), Sarath Sreedharan (Colorado State University)
Robotic IntelligenceReinforcement Learning from Human FeedbackBenchmark
🎯 What it does: Propose the Goal State Divergence (GSD) metric to measure the discrepancy between the robot's final state after completing a human-specified task and the human's expected state, and define the Human-Robot Goal State Alignment Design (HRGAD) problem, which aims to find the minimal environmental modifications to reduce GSD.
Reducing the Scope of Language Models
David Yunis (Toyota Technological Institute at Chicago), Danish Contractor (IBM Research AI)
Domain AdaptationTransformerSupervised Fine-TuningText
🎯 What it does: Investigated and evaluated multiple methods to restrict large language models to a specified domain (i.e., 'scoping'), making them generate answers only for relevant queries and reject irrelevant ones.
Redundancy-optimized Multi-head Attention Networks for Multi-view Multi-label Feature Selection
Yuzhou Liu (Jilin University), Wanfu Gao (Jilin University)
OptimizationRepresentation LearningTransformerImageTabular
🎯 What it does: Propose the RMAN-MMFS method, which employs a redundancy-optimized multi-head attention network for multi-view multi-label feature selection.
Redundant Queries in DETR-Based 3D Detection Methods: Unnecessary and Prunable
Lizhen Xu (Xi'an Jiaotong University), Jianru Xue (Xi'an Jiaotong University)
Object DetectionComputational EfficiencyTransformerImagePoint Cloud
🎯 What it does: Propose the GPQ (Gradually Pruning Queries) method, which gradually prunes redundant queries in the DETR-based 3D detection model to reduce computational and memory costs.
RefAdGen: High-Fidelity Advertising Image Generation
Yiyun Chen (Hong Kong University of Science and Technology (Guangzhou)), Weikai Yang (Hong Kong University of Science and Technology (Guangzhou))
GenerationData SynthesisConvolutional Neural NetworkDiffusion modelImage
🎯 What it does: Proposed a high-fidelity advertising image generation framework called RefAdGen that does not require fine-tuning for each product individually.
Reference Recommendation Based Membership Inference Attack Against Hybrid-Based Recommender Systems
Xiaoxiao Chi, Wanchun Dou (Nanjing University)
Recommendation SystemSafty and PrivacyAdversarial AttackTabular
🎯 What it does: A membership inference attack (MIA) based on reference recommendations is proposed for hybrid recommendation systems, which utilizes a relative membership metric to distinguish whether user data was used for model training.
RefiDiff: Progressive Refinement Diffusion for Efficient Missing Data Imputation
Md Atik Ahamed (University of Kentucky), Qiang Cheng (University of Kentucky)
Computational EfficiencyData-Centric LearningDiffusion modelTabularBenchmarkStochastic Differential Equation
🎯 What it does: Propose a hybrid predictive and generative missing value imputation method called RefiDiff, adopting a progressive refinement strategy in pre- and post-stages;
Refine-IQA: Multi-Stage Reinforcement Finetuning for Perceptual Image Quality Assessment
Ziheng Jia (Shanghai Jiao Tong University), Xiongkuo Min (Shanghai Jiao Tong University)
Large Language ModelReinforcement LearningVision Language ModelImageMultimodality
🎯 What it does: Propose a multi-stage reinforcement learning fine-tuning framework called Refine-IQA to enhance the perception and scoring capabilities of multimodal large models in image quality assessment (IQA).
ReFINE: A Reward-Based Framework for Interpretable and Nuanced Evaluation of Radiology Report Generation
Yunyi Liu (University of Sydney), Luping Zhou (University of Adelaide)
Explainability and InterpretabilityData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextBiomedical DataMagnetic Resonance Imaging
🎯 What it does: Proposed a reward-based evaluation framework called ReFINE for interpretable and fine-grained quality assessment of radiology report generation.
Refine3D: Scene-Adaptive Reference Point Refinement for Sparse 3D Object Detection
Fan Li (Hikvision Research Institute), Yi Niu (Hikvision Research Institute)
Object DetectionAutonomous DrivingTransformerPoint Cloud
🎯 What it does: Propose Refine3D, which enhances the localization accuracy and robustness of sparse query-based 3D detection through scene-adaptive reference point refinement (SAR-Head) and global distribution supervision (RPD-Loss).
Refinement Contrastive Learning of Cell–Gene Associations for Unsupervised Cell Type Identification
Liang Peng (Shantou University), Hau-San Wong (Shantou University)
ClassificationRepresentation LearningGraph Neural NetworkContrastive LearningGraphBiomedical Data
🎯 What it does: Developed an unsupervised cell type identification framework named scRCL, which utilizes cell-gene associations to improve cell embedding representations.
RefineVAD: Semantic-Guided Feature Recalibration for Weakly Supervised Video Anomaly Detection
Junhee Lee (Kyung Hee University), MyeongAh Cho (Kyung Hee University)
Anomaly DetectionTransformerVision Language ModelVideo
🎯 What it does: Proposes the RefineVAD framework, which simultaneously leverages motion-aware temporal attention and category-oriented feature reconstruction in weakly supervised video anomaly detection to achieve more precise localization of abnormal events.
Reflect Then Learn: Active Prompting for Information Extraction Guided by Introspective Confusion
Dong Zhao (Nanjing University of Aeronautics and Astronautics), Sheng-Jun Huang (Nanjing University of Aeronautics and Astronautics)
TransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Propose an active prompting framework APIE based on LLMs, which actively selects the most informative examples by evaluating model uncertainty through 'introspective confusion,' thereby improving the accuracy and robustness of information extraction with few samples.
ReflexDiffusion: Reflection-Enhanced Trajectory Planning for High-lateral-acceleration Scenarios in Autonomous Driving
Xuemei Yao (National University of Defense Technology), Kewei Yang (National University of Defense Technology)
Autonomous DrivingDiffusion model
🎯 What it does: Proposes improving diffusion model-based trajectory planning through a reflection mechanism during the inference phase, specifically enhancing vehicle driving safety in high lateral acceleration scenarios.
RefleXNet: Targeted Self-Reflection for Accurate Chest X-ray Reporting
Xin Mei (Northwestern Polytechnical University), Erik Cambria (Northwestern Polytechnical University)
ClassificationObject DetectionAnomaly DetectionConvolutional Neural NetworkGraph Neural NetworkTransformerLarge Language ModelImageTextBiomedical Data
🎯 What it does: Proposed RefleXNet, which integrates multi-scale visual features, anatomical relationship graph models, and target self-reflection learning for accurate generation of chest X-ray reports.
REFO: Reinforced Evolutionary Faithfulness Optimization for Large Language Models
Yi Wang (Hong Kong University of Science and Technology), Sihong Xie (Hong Kong University of Science and Technology)
RetrievalOptimizationTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextRetrieval-Augmented Generation
🎯 What it does: Proposed the REFO framework, which leverages a self-evolving iterative process to automatically generate high-quality data and employs DPO for iterative training, significantly enhancing the faithfulness of retrieval-augmented generation (RAG) systems.
RefRea: Reference-Guided Reasoning with Meta-Cognition for Accurate Language Model Agents
Yuxiang Mai (University of Chinese Academy of Sciences), Kaiqi Huang (University of Chinese Academy of Sciences)
Reinforcement Learning from Human FeedbackLarge Language ModelReinforcement LearningAgentic AITextChain-of-Thought
🎯 What it does: Introduce a dual-path framework in large language model (LLM) agents, leveraging a reference model to calibrate the reasoning model's behavior, and generating verifiable thinking summaries through a metacognitive module to enhance reasoning consistency and decision accuracy.
RefSTAR: Blind Face Image Restoration with Reference Selection, Transfer, and Reconstruction
Zhicun Yin (Harbin Institute of Technology), Wangmeng Zuo (Nanyang Technological University)
RestorationConvolutional Neural NetworkDiffusion modelImage
🎯 What it does: Proposed the RefSTAR framework for blind face image restoration tasks, significantly enhancing facial identity preservation and texture detail recovery through three steps: reference image selection, transfer, and reconstruction.
Region-Point Joint Representation for Effective Trajectory Similarity Learning
Hao Long (University of Electronic Science and Technology of China), Shuo Shang (University of Electronic Science and Technology of China)
RetrievalRepresentation LearningConvolutional Neural NetworkGraph Neural NetworkTransformerMixture of ExpertsContrastive LearningGraphTime SeriesStochastic Differential Equation
🎯 What it does: Proposes a trajectory similarity learning framework called RePo that integrates regional and point-level features for more accurate trajectory similarity computation.
RegionMarker: A Region-Triggered Semantic Watermarking Framework for Embedding-as-a-Service Copyright Protection
Shufan Yang (Nanjing University), Qing Gu (Nanjing University)
Safty and PrivacyRepresentation LearningData-Centric LearningText
🎯 What it does: Propose RegionMarker, a watermarking framework based on semantic region triggers, to protect Embedding-as-a-Service against copyright infringement under model extraction attacks.
RegionRAG: Region-level Retrieval-Augmented Generation for Visual Document Understanding
Yinglu Li (University of Science and Technology of China), Hongtao Xie (University of Science and Technology of China)
RetrievalLarge Language ModelVision Language ModelContrastive LearningImageTextMultimodalityRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose the RegionRAG framework, shifting retrieval granularity from document-level to region-level, leveraging region-based retrieval to provide precise context for large language models, thereby enhancing visual document understanding and question-answering performance.
Regression over Classification: Assessing Image Aesthetics via Multimodal Large Language Models
Xingyuan Ma (Beijing University of Posts and Telecommunications), Huadong Ma (Beijing University of Posts and Telecommunications)
TransformerLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: Propose a new method, ROC4MLLM, to improve score prediction of multimodal large language models in image aesthetics assessment (IAA);
Regressor-guided Diffusion Model for De Novo Peptide Sequencing with Explicit Mass Control
Shaorong Chen (Zhejiang University), Jun Xia (Hong Kong University Of Science And Technology (Guangzhou))
GenerationTransformerDiffusion modelBiomedical DataBenchmark
🎯 What it does: Propose a de novo peptide sequencing method called DiffuNovo based on diffusion models, which can explicitly control the theoretical mass of predicted peptides to be consistent with the experimental precursor mass during inference.
Regular Games -- an Automata-Based General Game Playing Language
Radosław Miernik (University of Wrocław), Wojciech Pawlik (University of Wrocław)
OptimizationComputational EfficiencyBenchmark
🎯 What it does: Proposed a new general-purpose game play system called Regular Games (RG), which implements rule descriptions through a low-level finite automaton language and provides high-level languages and tool ecosystems, making game design both concise and efficient.
Reimagining Anomalies: What If Anomalies Were Normal?
Philipp Liznerski (RPTU University Kaiserslautern-Landau), Marius Kloft (RPTU University Kaiserslautern-Landau)
Anomaly DetectionExplainability and InterpretabilityDiffusion modelGenerative Adversarial NetworkImage
🎯 What it does: Propose a method using adversarial generative networks and diffusion models to generate diverse counterfactual examples to explain the discriminative criteria of image anomaly detectors.
REINA: Regularized Entropy Information-Based Loss for Efficient Simultaneous Speech Translation
Nameer Hirschkind (Roblox), Mahesh Kumar Nandwana (Roblox)
Computational EfficiencyTransformerSupervised Fine-TuningTextAudio
🎯 What it does: Proposed a new loss function REINA to efficiently convert non-streaming speech translation models into simultaneous speech translation (SimulST) models, and trained a large-scale system on multilingual open-source data.
Reinforce Trustworthiness in Multimodal Emotional Support System
Huy M. Le (Mohamed bin Zayed University of Artificial Intelligence), Binh T. Nguyen (Singapore Management University)
ClassificationRecognitionReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningVision Language ModelVideoTextMultimodalityAudio
🎯 What it does: Proposed and implemented the MultiMood framework, which integrates multimodal features from video, audio, and text to predict emotional support elements, generates responses aligned with professional psychological therapy standards using a large language model (LLM), and enhances response credibility through reinforcement learning.
Reinforced Rate Control for Neural Video Compression via Inter-Frame Rate–Distortion Awareness
Wuyang Cong (Nanjing University), Zhan Ma (Nanjing University)
CompressionConvolutional Neural NetworkReinforcement LearningVideo
🎯 What it does: This paper proposes a reinforcement learning-based neural video compression rate control framework, treating the encoding decision for each frame as a constrained Markov decision process, jointly considering information such as reference frames, current frames, and network bandwidth, and directly outputting the Lagrange multiplier and downsampling factor for each frame;
Reinforcement Learning Enhanced Muti-hop Reasoning for Temporal Knowledge Question Answering
Wuzhenghong Wen (Beihang University), Minlong Peng (Beihang University)
TransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningPrompt EngineeringContrastive LearningTextGraph
🎯 What it does: Proposed a Multi-hop Reasoning Enhancement (MRE) framework to improve the reasoning quality in Temporal Knowledge Graph Question Answering (TKGQA).
Reinforcement Learning with Fuzzy Human Attention-Guided Graph for Heterogeneous Multiagent Systems
Dingbang Liu (University of Wollongong), Wen Gu (University of Wollongong)
Reinforcement Learning from Human FeedbackGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: In heterogeneous multi-agent reinforcement learning, utilizing fuzzy human attention-guided graphs to improve collaboration and graph structure learning.
Rejoining Precious Artifacts: Efficiently Bone Stick Rejoining Based Massive Fragment Images by Contour, Script, and Texture
Xingyi Wang (Sichuan University), Jian Peng (Sichuan University)
ClassificationRecognitionConvolutional Neural NetworkGraph Neural NetworkImage
🎯 What it does: This paper proposes a lightweight visual graph neural network, RejoinViG, which can directly determine the feasibility of joining bone pen fragments and predict the joining direction using fragment images without requiring manual annotations.
RelaCtrl: Relevance-Guided Efficient Control for Diffusion Transformers
Ke Cao (University of Science and Technology of China), Jie Zhang (University of Science and Technology of China)
GenerationTransformerDiffusion modelImage
🎯 What it does: Propose a controllable generation framework called RelaCtrl based on Diffusion Transformer, which dynamically determines the insertion position, scale, and capability of control layers by leveraging the relevance of control information, and designs a lightweight Two-Dimensional Shuffle Mixer (TDSM) to replace traditional self-attention and FFN, achieving efficient integration of control modules.
Relation-R1: Progressively Cognitive Chain-of-Thought Guided Reinforcement Learning for Unified Relation Comprehension
Lin Li (Hong Kong University of Science and Technology), Long Chen (Hong Kong University of Science and Technology)
RecognitionTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningVision Language ModelMultimodalityChain-of-Thought
🎯 What it does: Propose a unified relation understanding framework named Relation‑R1, combining supervised fine-tuning guided by cognitive chain-of-thought (CoT) and group relative strategy optimization, achieving reasoning and localization of binary and N-ary relations in images.
Relational Verification for Cost-Aware Quantum Program Optimization
Ziming Zhao (Zhejiang University), Jianwei Yin (Zhejiang University)
OptimizationBenchmarkPhysics Related
🎯 What it does: Proposes RelOpt, a quantum program optimizer based on relational verification, which achieves semantically safe quantum circuit optimization by leveraging verified rewrite rules and a multi-objective cost model.
Relative Advantage Debiasing for Watch-Time Prediction in Short-Video Recommendation
Emily Liu (ByteDance), Yang Song (ByteDance)
Recommendation SystemVideo
🎯 What it does: Propose a de-biasing framework for watch time based on Relative Advantage (RAD), mapping original watch time to quantile labels at both video and user levels to more accurately capture user interests.
Reliability-Guaranteed and Reward-Seeking Sequence Modeling for Model-Based Offline Reinforcement Learning
Shenghong He, Xuetao Ding (Sun Yat Sen University)
TransformerReinforcement LearningAuto EncoderSequentialBenchmark
🎯 What it does: Propose an offline reinforcement learning algorithm RT based on Transformer, which utilizes reliability estimation and high-reward conditions to generate reliable and high-reward trajectories, thereby enhancing policy learning in model-based offline RL.
Reliable-View 2D-3D Key-Part Aligned Transformer with Reinforced Masking for 3D Point Cloud Understanding
Xianglong Jin (Northwestern Polytechnical University), Feiping Nie (Northwestern Polytechnical University)
ClassificationSegmentationTransformerReinforcement LearningAuto EncoderContrastive LearningPoint CloudBenchmark
🎯 What it does: Train 3D point cloud representations using self-supervised multi-view 2D-3D alignment and reinforced masking Masked Autoencoder, integrating multi-view images to enhance point cloud understanding.
Relink: Constructing Query-Driven Evidence Graph On-the-Fly for GraphRAG
Manzong Huang (Hefei University of Technology), Xindong Wu (College of William and Mary)
RetrievalRepresentation LearningGraph Neural NetworkTransformerLarge Language ModelContrastive LearningTextGraphRetrieval-Augmented Generation
🎯 What it does: Built a dynamic, query-driven evidence graph framework called Relink to instantly complete missing knowledge and filter noisy facts in open-domain question answering, enhancing the accuracy of multi-hop reasoning.
ReLUPruner: Rethinking ReLU Importance with Taylor Expansion for Efficient Private Inference
Zhenpeng Li (Wuhan University), Jeff Z. Pan (Wuhan University)
ClassificationComputational EfficiencyKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: Proposed a ReLU importance evaluation method called ReLUPruner based on second-order Taylor expansion, adopting hierarchical importance metrics and progressive sparsification strategies, ultimately significantly reducing the number of ReLUs in private inference while maintaining accuracy.
Remember Me: Bridging the Long-Range Gap in LVLMs with Three-Step Inference-Only Decay Resilience Strategies
Peng Gao (Hong Kong Baptist University), Hui Zhang (University of Wollongong)
Computational EfficiencyVision Language ModelMultimodality
🎯 What it does: Propose a three-step decay tolerance strategy (T-DRS) during the reasoning phase, utilizing semantic-driven, distance control, and remote re-enhancement mechanisms to address the long-range attention decay issue in large vision-language models based on RoPE.
REMISVFU: Vertical Federated Unlearning via Representation Misdirection for Intermediate Output Feature
Wenhan Wu (Wuhan University), Dazhao Cheng (Wuhan University)
Federated LearningSafty and PrivacyImage
🎯 What it does: Propose a fast client-level no-learning framework REMISVFU for vertical federated learning, which eliminates contributions by introducing representation shift on the intermediate features of the forgetting party.
Remodeling Semantic Relationships in Vision-Language Fine-Tuning
Xiangyang Wu, Zhenwei Shi (Nanyang Technological University)
Computational EfficiencyRepresentation LearningSupervised Fine-TuningVision Language ModelMultimodality
🎯 What it does: Propose a parameter-efficient visual-language fine-tuning framework named LSRM, which achieves more precise cross-modal alignment and fusion by leveraging multi-layer visual features, semantic relation projection, and inheritable cross-attention.
RemoteReasoner: Towards Unifying Geospatial Reasoning Workflow
Liang Yao (Hohai University), Pai Peng (Hohai University)
Object DetectionSegmentationTransformerLarge Language ModelReinforcement LearningVision Language ModelImageMultimodalityBenchmark
🎯 What it does: Proposed RemoteReasoner, a unified geospatial reasoning workflow capable of performing pixel-level, region-level, and contour-level reasoning tasks within a single model;
Removing Box-Free Watermarks for Image-to-Image Models via Query-Based Reverse Engineering
Haonan An (City University of Hong Kong), Yuguang Fang (City University of Hong Kong)
Image TranslationAdversarial AttackDiffusion modelImageText
🎯 What it does: Proposes two query-based reverse engineering attacks to remove embedded frameless watermarks from black-box image-to-image models.
RENEW: Risk- and Energy-Aware Navigation in Dynamic Waterways
Mingi Jeong (Virginia Tech), Alberto Quattrini Li (Dartmouth)
OptimizationPhysics Related
🎯 What it does: Designed a risk and energy consumption joint planner RENEW for Autonomous Surface Vehicles (ASVs), integrating dynamic current environments and safety constraints to generate safe and energy-optimal paths.
Renormalization Group Guided Tensor Network Structure Search
Maolin Wang (City University of Hong Kong), Xiangyu Zhao (National University of Defense Technology)
CompressionOptimizationImageVideoPhysics Related
🎯 What it does: Propose a tensor network structure search framework RGTN based on renormalization group (RG) flow, achieving continuous evolution of network topology through learnable edge gates, automatically discovering efficient tensor decomposition structures.
Rep Deep & Machine Learning: Exemplar-Free Continual Video Action Recognition via Slow-Fast Collaborative Learning
Xueyi Zhang (National University of Defense Technology), Huiping Zhuang (National University of Defense Technology)
RecognitionKnowledge DistillationConvolutional Neural NetworkVideo
🎯 What it does: Studied sample-free continual video action recognition and proposed the slow-fast collaborative learning framework SFCL;
Repetition Makes Perfect: Recurrent Graph Neural Networks Match Message Passing Limit
Eran Rosenbluth (RWTH Aachen University), Martin Grohe (RWTH Aachen University)
Representation LearningRecurrent Neural NetworkGraph Neural NetworkGraph
🎯 What it does: Prove that the recursive graph neural network (R-GNN) with repeated summation aggregation can equivalently simulate any message passing algorithm satisfying color refinement invariance under finite precision parameters, thereby achieving the upper bound of expressive power for graph neural networks
Res-Bench: Benchmarking the Robustness of Multimodal Large Language Models to Dynamic Resolution Input
Chenxu Li (University of Science and Technology of China), Xiang Wang (University of Science and Technology of China)
TransformerLarge Language ModelSupervised Fine-TuningMultimodalityBenchmark
🎯 What it does: Constructed the ResBench benchmark to systematically evaluate the robustness of multimodal large language models (MLLMs) across different image resolutions, and proposed novel continuous error metrics such as ACE/RCE;
Resilient UAV Swarm with Fast Connectivity Recovery and Extensive Coverage
Yabin Peng (Southeast University), Shaoxun Liu (Purple Mountain Laboratories)
Robotic IntelligenceGraph Neural NetworkGraph
🎯 What it does: Propose an R2C framework that utilizes buffer virtual forces and multi-component graph convolution to achieve self-healing in drone swarms, rapidly restoring connectivity and expanding coverage.
ResMAS: Resilience Optimization in LLM-based Multi-agent Systems
Zhilun Zhou (Tsinghua University), Fengli Xu (Huawei)
OptimizationGraph Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningAgentic AIPrompt EngineeringTextGraph
🎯 What it does: Propose ResMAS—a two-phase framework that first uses a reward model and RL to train LLMs to automatically generate robust topologies, then optimizes prompts based on the topology to enhance the robustness of large language model multi-agent systems under random failures.
Resource Efficient Sleep Staging via Multi-Level Masking and Prompt Learning
Lejun Ai (South China University of Technology), Rui Wang (Huazhong University of Technology)
ClassificationComputational EfficiencyConvolutional Neural NetworkRecurrent Neural NetworkTransformerPrompt EngineeringTime SeriesBiomedical Data
🎯 What it does: Studied the sleep staging task in resource-constrained environments, proposing a Mask-Aware Sleep Staging (MASS) framework based on multi-level masking and global prompt learning, which can achieve high-precision sleep stage classification using only about 10% of 30-second EEG signals.
Response Attack: Exploiting Contextual Priming to Jailbreak Large Language Models
Miao Ziqi (Shanghai Artificial Intelligence Laboratory), Jing Shao (Shanghai Artificial Intelligence Laboratory)
Safty and PrivacyAdversarial AttackTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: This paper proposes a new attack framework called Response Attack (RA), which utilizes a mildly harmful response generated in the conversation as a contextual preface to induce large language models to generate non-compliant content.
ResProto-FD: Visual-Language Residual Prototype Sets for Generalized Face Forgery Detection
Jiuyao Jing (Xidian University), Chunlei Peng (Xidian University)
ClassificationAnomaly DetectionVision Language ModelImageTextMultimodality
🎯 What it does: The ResProto-FD framework, which leverages visual-language residual learning and gradient-aware prototypes, is proposed to enhance the generalization performance of facial forgery detection.
RESTL: Reinforcement Learning Guided by Multi-Aspect Rewards for Signal Temporal Logic Transformation
Yue Fang (Peking University), Naijun Zhan (JD.com)
GenerationReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningText
🎯 What it does: Proposed RESTL, a reinforcement learning-based natural language to Signal Temporal Logic (STL) conversion framework that employs a multi-dimensional reward model for fine-grained supervision;
Retaliatory Attacks Against Federated Unlearning via Data Leakage
Xinyi Sheng (University of Sydney), Sen Fu (University of Sydney)
Federated LearningSafty and PrivacyImageTabular
🎯 What it does: This paper proposes novel retaliatory attacks against Federated Unlearning (FU), including Anti-Unlearning Attack (AUA) and Discrimination-Unlearning Attack (DUA).
Rethink Representation Learning for Questionnaire Data
Guanhua Ye (Beijing University of Posts and Telecommunications), Yawen Li (Beijing University of Posts and Telecommunications)
ClassificationRepresentation LearningTransformerLarge Language ModelTabularRetrieval-Augmented Generation
🎯 What it does: Propose the SemantiQ framework, which transforms questionnaire questions, options, and external knowledge into semantic sentences through retrieval-augmented generation (RAG) and large language models, then embeds them into a unified semantic space. The framework improves the representation and prediction performance of questionnaire data using three-stage training and training during testing (TTT).
Rethinking Bias in Generative Data Augmentation for Medical AI: A Frequency Recalibration Method
Chi Liu (City University of Macau), Wanlei Zhou (City University of Macau)
ClassificationTransformerAuto EncoderImageMagnetic Resonance Imaging
🎯 What it does: Investigated the bias issue in medical image generation data augmentation (GDA), proposing a post-processing scheme called FreRec, which includes statistical high-frequency replacement and reconstructing high-frequency mapping to calibrate the frequency distribution of synthesized images, significantly enhancing downstream classification performance.
Rethinking Crystal Symmetry Prediction: A Decoupled Perspective
Liheng Yu (University of Science and Technology of China), Pengkun Wang (University of Science and Technology of China)
ClassificationConvolutional Neural NetworkTransformerPhysics Related
🎯 What it does: Built the XRDecoupler framework, adopting a decoupled perspective to address sub-attribute confusion (SPC) in PXRD prediction.
Rethinking Deep Alignment Through the Lens of Incomplete Safety Learning
Thong Bach (Deakin University), Truyen Tran (Deakin University)
Safty and PrivacyKnowledge DistillationTransformerLarge Language ModelText
🎯 What it does: Proposed and implemented a training-time safety alignment completion framework based on mechanism analysis, capable of fully covering safety alignment for all token positions in large language models;
Rethinking Direct Preference Optimization in Diffusion Models
Junyong Kang (Korea Advanced Institute of Science and Technology), Hyunjung Shim (Sejong University)
GenerationOptimizationReinforcement LearningDiffusion modelImageTextMultimodality
🎯 What it does: Improve Diffusion-DPO by proposing stable reference model updates and temporal-aware optimization schemes to enhance the alignment performance of text-to-image models.
Rethinking Explanation Evaluation Under the Retraining Scheme
Yi Cai (Freie Universität Berlin), Gerhard Wunder (Freie Universität Berlin)
ClassificationExplainability and InterpretabilityComputational EfficiencyConvolutional Neural NetworkTransformerSupervised Fine-TuningImage
🎯 What it does: Proposed and improved a retraining-based feature importance explanation evaluation method, identified and corrected the Sign problem, and introduced efficient evaluation schemes such as KAFT-C.
Rethinking Flow and Diffusion Bridge Models for Speech Enhancement
Dahan Wang (Nanjing University), Jing Lu (Nanjing University)
RestorationDiffusion modelFlow-based ModelAudio
🎯 What it does: This paper unifies the theoretical frameworks of flow models and diffusion bridge models, and proposes an improved bridge model;
Rethinking Irregular Time Series Forecasting: A Simple Yet Effective Baseline
Xvyuan Liu (East China Normal University), Bin Yang (East China Normal University)
Computational EfficiencyRepresentation LearningTime SeriesBiomedical Data
🎯 What it does: Proposes the APN framework, which adaptively generates high-quality patch representations through the Time-Aware Patch Aggregation (TAPA) module, and combines query aggregation with a shallow MLP to achieve efficient prediction for irregular multivariate time series.
Rethinking Label Consistency of In-Context Learning: An Implicit Transductive Label Propagation Perspective
Haoyang Chen (Beihang University), Junfan Chen (Beihang University)
ClassificationData SynthesisLarge Language ModelText
🎯 What it does: This paper reconsiders ICL as transductive learning, proposes a label propagation framework based on Bayesian transduction, and designs the TopK-SD data synthesis method to enhance the consistency of demonstration labels.