AAAI 2026 Papers — Page 32
AAAI Conference on Artificial Intelligence · 4149 papers
Rethinking Long-tailed Dataset Distillation: A Uni-Level Framework with Unbiased Recovery and Relabeling
Xiao Cui (University of Science and Technology of China), Houqiang Li (University of Science and Technology of China)
ClassificationKnowledge DistillationData-Centric LearningImageBenchmark
🎯 What it does: Achieve dataset distillation on long-tailed datasets through a single-layer optimization framework for unbiased retrieval and soft re-labeling;
Rethinking Membership Inference Attacks for CLIP
Lluis Gomez (Universitat Autonoma De Barcelona)
Safty and PrivacyRepresentation LearningAdversarial AttackTransformerVision Language ModelContrastive LearningMultimodalityBenchmark
🎯 What it does: Redesign and evaluate membership inference attacks on the CLIP model, constructing a strict same-distribution benchmark to reveal the misleading nature of previous cross-dataset evaluations.
Rethinking Multi-Instance Learning Through Graph-Driven Fusion: A Dual-Path Approach to Adaptive Representation
Yu-Xuan Zhang (Southwest Jiaotong University), Mingxing Zhang (Southwest Jiaotong University)
ClassificationRecommendation SystemGraph Neural NetworkImageTextBiomedical Data
🎯 What it does: This paper proposes a graph-driven multi-instance learning framework called GDF-MIL, which employs an adaptive dual-path fusion strategy to enhance weakly supervised classification performance.
Rethinking Multimodal Point Cloud Completion: A Completion-by-Correction Perspective
Wang Luo (Sun Yat-sen University), Guocong Quan (Sun Yat-sen University)
GenerationGraph Neural NetworkTransformerImageMultimodalityPoint Cloud
🎯 What it does: Generate complete 3D priors from single-view RGB images and perform correction in the feature space, followed by a three-stage process of dual-modal feature encoding, seed generation, and hierarchical refinement to achieve point cloud completion.
Rethinking Open-world Prompt Tuning: A Systematic Framework for Evaluation and Optimization
Mengwei Li (University of Science and Technology of China), Yixin Zhang (Institute of Artificial Intelligence, Hefei Comprehensive National Science Center)
ClassificationOptimizationTransformerLarge Language ModelPrompt EngineeringVision Language ModelImageVideo
🎯 What it does: This paper addresses the open-world prompt tuning (OPT) task by proposing a more comprehensive evaluation framework, a training-free OOD detection method ERF, and a plug-and-play Gated Dual-Merging (GDM) strategy to enhance the recognition performance of base classes and new classes.
Rethinking Progression of Memory State in Robotic Manipulation: An Object-Centric Perspective
Nhat Chung (FPT Software AI Center), Ngan Le
Robotic IntelligenceTransformerLarge Language ModelVideoSequentialBenchmark
🎯 What it does: This paper proposes LIBERO-Mem, a benchmark for object-level non-Markovian robot manipulation, along with the Embodied-SlotSSM framework based on Slot Attention and State Space Model, addressing challenges of long-term memory and subgoal alignment.
Rethinking Rainy 3D Scene Reconstruction via Perspective Transforming and Brightness Tuning
Qianfeng Yang (Dalian Polytechnic University), Jiyu Jin (Dalian Polytechnic University)
RestorationGenerationData SynthesisRecurrent Neural NetworkGaussian SplattingImage
🎯 What it does: This study constructs a new 3D rainy scene reconstruction dataset called OmniRain3D and proposes an end-to-end REVR-GSNet framework, which can simultaneously achieve brightness recovery, rain removal, and 3D Gaussian Splatting reconstruction under rainy images;
Rethinking Saliency Maps: A Cognitive Human Aligned Taxonomy and Evaluation Framework for Explanations
Yehonatan Elisha (Tel Aviv University), Noam Koenigstein (Tel Aviv University)
Explainability and InterpretabilityConvolutional Neural NetworkTransformerImage
🎯 What it does: Proposed the Reference-Frame × Granularity (RFxG) two-axis framework, redefining the reference frame (point-to-point vs. contrastive) and granularity (fine-grained vs. group-level) of saliency maps, and introduced four new contrastive and group-level faithfulness metrics based on this framework.
Rethinking Surgical Smoke: A Smoke-Type-Aware Laparoscopic Video Desmoking Method and Dataset
Qifan Liang (Wuhan University), Bin Mei (Zhongnan Hospital)
RestorationData SynthesisConvolutional Neural NetworkTransformerVideo
🎯 What it does: This paper proposes a smoke-type-aware laparoscopic video de-smoking network called STANet, which can specifically remove smoke based on two distinct motion patterns: diffusion smoke and ambient smoke.
Rethinking Target Label Conditioning in Adversarial Attacks: A 2D Tensor-Guided Generative Approach
Hangyu Liu (Zhejiang University), Donglin Wang (Westlake University)
Adversarial AttackConvolutional Neural NetworkTransformerDiffusion modelGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes a generative adversarial attack framework called TGAF, based on a two-dimensional semantic tensor, to achieve multi-target transferable adversarial attacks.
Rethinking the Dark Knowledge and Kullback-Leibler Divergence Loss in Knowledge Distillation Under Capacity Mismatching
Yingchao Wang (Beijing Institute of Technology), Weilun Fei (Beijing Institute of Technology)
ClassificationKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: Proposed a knowledge distillation method called RCKD based on a relative confidence matrix to address the shortcomings of traditional KL distillation in terms of class ranking and gradient competition;
Rethinking the Reliability of Multi-agent System: A Perspective from Byzantine Fault Tolerance
Lifan Zheng (Zhejiang University), Yu Tian (Tsinghua University)
TransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: The study investigates the reliability of LLM-driven multi-agent systems (MAS) from a Byzantine fault tolerance perspective, and proposes a confidence-based weighted Byzantine fault tolerance consensus mechanism called CP-WBFT, enhancing system robustness in extreme Byzantine environments.
Rethinking the Sampling Criteria in Reinforcement Learning for LLM Reasoning: A Competence-Difficulty Alignment Perspective
Deyang Kong (Peking University), Wei Ye (Peking University)
AI Code AssistantTransformerLarge Language ModelReinforcement LearningTextBenchmark
🎯 What it does: Proposes the CDAS sampling framework based on matching model capability with problem difficulty, improving sample efficiency and performance in RL training.
Rethinking the Spatio-Temporal Alignment of End-to-End 3D Perception
Xiaoyu Li (Harbin Institute of Technology), Lining Sun (Harbin Institute of Technology)
Object DetectionObject TrackingAutonomous DrivingMultimodalityPoint Cloud
🎯 What it does: Proposed the HAT module, improving spatiotemporal alignment in end-to-end 3D perception using a multi-hypothesis motion model and adaptive decoder.
Rethinking Video-Language Model from the Language Input Perspective
Xiang Fang (Huazhong University of Science and Technology), Daizong Liu (Wuhan University)
RetrievalLarge Language ModelVision Language ModelContrastive LearningVideoTextMultimodality
🎯 What it does: This paper proposes a pluggable text enhancement framework, significantly improving the robustness and performance of video language models in tasks such as video sentence localization, video question answering, and video retrieval through the generation of positive and negative text variants, attribute-level text reasoning, and self-weighted cross-modal bridging loss.
Rethinking Visual Token Reduction in LVLMs Under Cross-Modal Misalignment
Rui Xu, Bo Du (Wuhan University)
Computational EfficiencyRepresentation LearningTransformerImageVideoTextMultimodalityBenchmark
🎯 What it does: Propose VisionDrop, an untrained vision-only attention-based token sparsification framework that progressively prunes and fuses residual information in visual encoders and LLMs stage-by-stage.
RetouchAgent: Towards Interactive and Explainable Image Retouching with MLLM Agents
Shuo Zhang (Xi'an Jiaotong University), Xinyu Yang (Xi'an Jiaotong University)
TransformerLarge Language ModelAgentic AIPrompt EngineeringVision Language ModelImageTextMultimodalityRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose the RetouchAgent framework, which utilizes multimodal large language models to collaboratively achieve interactive and interpretable image retouching;
ReTrack: Evidence-Driven Dual-Stream Directional Anchor Calibration Network for Composed Video Retrieval
Zixu Li, Meng Liu (Shandong University)
RetrievalTransformerVision Language ModelVideoTextMultimodality
🎯 What it does: Proposed the ReTrack framework to address the directional bias problem of composite features in compositional video retrieval (CVR), achieving more accurate multimodal retrieval through semantic contribution decoupling, geometric calibration, and reliable evidence alignment;
Retrieval-driven Reasoning for Deliberative Visual Classification
Jianye Xie (China University of Petroleum East China), Xiaokang Zhou (University of Massachusetts Boston)
ClassificationRetrievalExplainability and InterpretabilityLarge Language ModelImageTextMultimodalityRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose a retrieval-driven reasoning model (RdR) that first constructs a retrieval database containing visual and contextual descriptions, then performs multi-step reasoning to complete visual classification tasks.
Retriever Encoder Selection Matters for In-Context Learning-based Medical Segmentation
Fan Wang (Shandong University), Yilong Yin (Shandong University)
SegmentationTransformerBiomedical Data
🎯 What it does: Propose the IRES framework, which dynamically selects retrieval encoders (RE) for each query to enhance the performance of medical image segmentation based on context learning.
RetroLM: Retrieval-Augmented KVs for Long-Context Processing
Kun Luo (Key Laboratory of Cognition and Decision Intelligence for Complex Systems, Institute of Automation, Chinese Academy of Sciences), Kang Liu (Key Laboratory of Cognition and Decision Intelligence for Complex Systems, Institute of Automation, Chinese Academy of Sciences)
RetrievalComputational EfficiencyTransformerLarge Language ModelContrastive LearningTextRetrieval-Augmented Generation
🎯 What it does: RetroLM achieves retrieval-augmented long-context reasoning by implementing retrieval enhancement at the KV cache level of LLMs, splitting long texts into pages and retrieving critical pages on demand.
RetrySQL: Text-to-SQL Training with Retry Data for Self-Correcting Query Generation
Alicja Rączkowska (Allegro), Paweł Olszewski (Allegro)
AI Code AssistantTransformerLarge Language ModelSupervised Fine-TuningTextBenchmarkChain-of-Thought
🎯 What it does: Proposes RetrySQL, a training paradigm that utilizes retry data for self-correction during the text-to-SQL generation process. After training, the model can identify and correct erroneous reasoning steps during generation, thereby producing more accurate SQL queries.
Revealing POMDPs: Qualitative and Quantitative Analysis for Parity Objectives
Ali Asadi (Institute of Science and Technology Austria), Raimundo Saona (London School of Ecnonomics and Political Science)
Reinforcement Learning
🎯 What it does: The study reveals the quantitative and qualitative analysis issues of revealing POMDP under parity objectives, and provides an EXPTIME decidable algorithm;
Revealing the Invisible: Latent Structure Modeling for Semantically Consistent Cloud Removal
Jingwei Xin (Xidian University), Nannan Wang (Xidian University)
RestorationTransformerAuto EncoderOptical FlowImage
🎯 What it does: Propose the VISER-CR framework, which redefines cloud removal as a structure-guided completion task, achieving high-quality recovery of cloud-occluded regions through a visible-guided mask and semantic refinement module.
Revisiting (Un)Fairness in Recourse by Minimizing Worst-Case Social Burden
Ainhize Barrainkua (Basque Center for Applied Mathematics), Novi Quadrianto (Basque Center for Applied Mathematics)
OptimizationExplainability and InterpretabilityTabular
🎯 What it does: Proposes a unified algorithmic fairness explanation framework and minimizes social burden via the MISOB method without using sensitive attributes;
Revisiting Attention in the Dark for Low-Light Person Re-Identiffcation
Xiang Guo, Mei Wang (Wuhan University)
RetrievalKnowledge DistillationTransformerImage
🎯 What it does: This paper proposes a Norm-Ratio Attention and Semantic Recovery Distillation Network, addressing the modeling and recovery of missing semantic information in pedestrian re-identification under extremely low illumination conditions.
Revisiting Conjunctive Query Entailment for S
Yazmín Ibáñez-García (Cardiff University), Filip Murlak (University of Warsaw)
🎯 What it does: This paper studies the answer reasoning problem for UCQ (union of conjunctive queries) in description logic S (ALC with transitive roles), proving that the problem is 2EXPTIME-complete when the query contains at least two transitive roles; if only one transitive role is allowed or the query is rooted (connected and non-Boolean), it is CONEXPTIME-complete.
Revisiting Contrastive Learning in Collaborative Filtering via Parallel Graph Filters
Fang Kai (Macau University of Science and Technology), Yiwen Zhang (Anhui University)
Recommendation SystemGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Propose the LightCSCF method, which suppresses structurally similar negative samples by using boundary-constrained cosine similarity in contrastive learning, addressing the GCN over-smoothing and contrastive learning gradient saturation issues, thereby improving recommendation effectiveness.
Revisiting Cross-Architecture Distillation: Adaptive Dual-Teacher Transfer for Lightweight Video Models
Ying Peng (South China University of Technology), Runhao Zeng (Shenzhen MSU-BIT University)
RecognitionComputational EfficiencyKnowledge DistillationRepresentation LearningConvolutional Neural NetworkTransformerVideo
🎯 What it does: This study proposes a dual-teacher knowledge distillation framework that leverages a heterogeneous Vision Transformer (ViT) teacher and a homogeneous lightweight CNN teacher to jointly guide a lightweight CNN student network, enhancing video action recognition performance.
Revisiting Differentiable Structure Learning: Inconsistency of L1 Penalty and Beyond
Kaifeng Jin (University of Illinois Urbana-Champaign), Biwei Huang (University of California San Diego)
OptimizationRepresentation LearningGraph
🎯 What it does: Investigated the inconsistency problem of L1 penalty in differentiable structure learning, and proposed the CALM method based on L0 penalty, hard DAG constraints, and moral graphs.
Revisiting Downsampling in Semantic Segmentation: Fighting Aliasing with Dynamic Gaussian and Gabor Frequency Filters
Yu Bing Luo (North University of China), Jianghui Cai (North University of China)
SegmentationImage
🎯 What it does: This paper conducts a systematic analysis of phase distortion caused by downsampling in semantic segmentation, proposing a frequency-aware filter (Freq Aware Filter), including dynamic Gaussian kernel (DGK) and learnable Gabor frequency selector (LFS), to eliminate aliasing-induced boundary blurriness and texture loss.
Revisiting Fairness-aware Interactive Recommendation: Item Lifecycle as a Control Knob
Yun Lu (Chongqing Institute of Green and Intelligent Technology, Chinese Academy of Sciences), Mingsheng Shang (Chongqing Institute of Green and Intelligent Technology, Chinese Academy of Sciences)
Recommendation SystemTransformerReinforcement LearningVideo
🎯 What it does: This paper proposes using the lifecycle of short video content (rapid growth, stable period, decline period) as a control mechanism to improve fairness and accuracy in interactive recommendations;
Revisiting MLLM Based Image Quality Assessment: Errors and Remedy
Zhenchen Tang (New Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences), Jing Dong (New Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences)
Large Language ModelSupervised Fine-TuningVision Language ModelImage
🎯 What it does: Propose the Q-Scorer framework to address conversion errors and semantic confusion caused by discrete tokens in MLLMs during image quality assessment.
Revisiting Network Inertia: Dynamic Inertia Inhibition Coupled Multidimensional Periodicity for Infrared and Visible Image Fusion
Yufeng Chen (Southwest University of Science and Technology), Xingfeng Li (Southwest University of Science and Technology)
Image HarmonizationConvolutional Neural NetworkTransformerImageMultimodality
🎯 What it does: Proposed a lightweight infrared-visible image fusion method called AIDFusion, aiming to address the network laziness problem and enhance fusion performance.
Revisiting the Canonicalization for Fast and Accurate Crystal Tensor Property Prediction
Haowei Hua (Hong Kong Polytechnic University), Pan Zhou
Graph Neural NetworkGraphPhysics Related
🎯 What it does: This paper proposes a general framework called GoeCTP based on continuous diagonalization (polar decomposition) for predicting O(3) pseudo-symmetry in crystal tensor properties, and presents it as a plugin that can be used in conjunction with any scalar property prediction network.
Revisiting the Data Sampling in Multimodal Post-training from a Difficulty-Distinguish View
Jianyu Qi (Central South University), Rongchang Zhao (Central South University)
Data-Centric LearningSupervised Fine-TuningReinforcement LearningVision Language ModelMultimodalityBenchmark
🎯 What it does: Propose two difficulty-aware sampling strategies (PISM and CMAB), and construct a multi-modal post-training framework based on difficulty hierarchy, demonstrating that training with GRPO alone can surpass the traditional SFT+GRPO scheme.
Revitalizing Canonical Pre-Alignment for Irregular Multivariate Time Series Forecasting
Ziyu Zhou (Hong Kong University of Science and Technology), Yuxuan Liang (Hong Kong University of Science and Technology)
Computational EfficiencyConvolutional Neural NetworkTransformerTime SeriesBiomedical Data
🎯 What it does: Proposes KAFNet, an irregular multivariate time series (IMTS) prediction model that leverages Canonical Pre-Alignment (CPA), integrating pre-convolution, temporal kernel aggregation, and frequency-domain linear attention modules.
Reward Model Evaluation via Automatically-Ranked Policy Alignment
Aoran Wang (Nanjing University), Zongzhang Zhang (Nanjing University)
Reinforcement Learning from Human FeedbackReinforcement Learning
🎯 What it does: This paper proposes a metric called PPAC for evaluating reward models without requiring reinforcement learning training, and automatically generates policy rankings through expert-guided policy improvement (DGPI).
Reward Redistribution via Gaussian Process Likelihood Estimation
Minheng Xiao (Ohio State University), Xian Yu (Ohio State University)
Reinforcement Learning
🎯 What it does: Propose a Gaussian Process-based Likelihood Reward Redistribution framework (GP-LRR), which generates dense reward signals by maximizing the likelihood of the entire trajectory through a leave-one-out strategy, and combines it with Soft Actor-Critic (SAC) to significantly improve sample efficiency and final performance in environments with sparse terminal rewards.
RFF-TTA: Physical Information-Aware Prototype for Temporally Varying RF Fingerprinting Online Test-Time-Adaptation
Taotao Li (Zhejiang University of Technology), Zhen Hong (Zhejiang University of Technology)
ClassificationDomain AdaptationConvolutional Neural NetworkContrastive LearningTime SeriesPhysics Related
🎯 What it does: Propose a physics-informed prototype-driven test-time adaptation (RFF-TTA) method for online adaptation to time-varying distribution shift in radio frequency spectrum fingerprint (RFF) identification, enhancing the model's domain generalization capability.
RFI: Rectified Flow Intervention for Mitigating Object Hallucination in Large Vision-Language Models
Junyu Cheng (Xiamen University), Shuangyin Li (South China Normal University)
Object DetectionVision Language ModelRectified FlowImageMultimodality
🎯 What it does: Propose RFI (Rectified Flow Intervention), a lightweight method that dynamically predicts hidden space intervention vectors in large vision-language models to suppress object hallucinations.
RFKG-CoT: Relation-Driven Adaptive Hop-count Selection and Few-Shot Path Guidance for Knowledge-Aware QA
Chao Zhang (Soochow University), Guodong Zhou (Soochow University)
Graph Neural NetworkLarge Language ModelPrompt EngineeringGraphBenchmarkChain-of-Thought
🎯 What it does: Propose the RFKG-CoT framework, combining knowledge graph reasoning with large language models to address the fixed hop-count and insufficient path utilization issues in traditional KG-CoT.
RflyPano: A Panoramic Benchmark for Ultra-low Altitude UAV Localization Powered by RflySim
Dun Dai (Beihang University), Quan Quan (Central South University)
Data SynthesisPose EstimationRetrievalAutonomous DrivingConvolutional Neural NetworkTransformerSimultaneous Localization and MappingImageBenchmark
🎯 What it does: This paper constructs the first panoramic visual localization benchmark for ultra-low altitude (<120 m) drones - the RflyPano dataset, and verifies two categories of localization methods (image retrieval and pose regression) on it.
RFNNS: Robust Fixed Neural Network Steganography with Universal Text-to-Image Models
Yu Cheng (East China Normal University), Xinpeng Zhang (Fudan University)
GenerationSafty and PrivacyDiffusion modelImage
🎯 What it does: Proposed a robust steganography method RFNNS based on a fixed neural network and a general text-to-image model, which embeds robust perturbations only in texture-complex regions of the cover image to achieve steganography and recovery.
RGMP: Recurrent Geometric-prior Multimodal Policy for Generalizable Humanoid Robot Manipulation
Xuetao Li (Wuhan University), Miao Li (Wuhan University)
Robotic IntelligenceRecurrent Neural NetworkVision-Language-Action ModelMultimodalityAudio
🎯 What it does: Proposed the RGMP framework, achieving a geometry-prior multi-modal policy based on voice commands, which includes a Geometric-prior Skill Selector and an Adaptive Recursive Gaussian Network, completing the full process from semantic parsing to trajectory generation;
RI-Loss: A Learnable Residual-Informed Loss for Time Series Forecasting
Jieting Wang (Shanxi University), Furong Peng (Shanxi University)
TransformerTime Series
🎯 What it does: Propose a loss function called RI-Loss based on the residual-random noise mutual information, utilizing the Hilbert-Schmidt Independence Criterion (HSIC) to model the statistical independence between prediction residuals and noise, thereby simultaneously suppressing observational noise and capturing long-term dependencies in time series.
Richer Representations for Neural Algorithmic Reasoning via Auxiliary Reconstruction
Jiafu Huang, Irene Zheng (East China Normal University)
Representation LearningGraph Neural NetworkAuto EncoderGraphBenchmark
🎯 What it does: This paper introduces an auxiliary reconstruction task into the traditional encoder-processor-decoder architecture and designs a stronger GNN+gated encoder along with a hint-level masking reconstruction strategy to enhance representation learning and inference performance of neural algorithmic reasoning models.
RICo: Refined In-Context Contribution for Automatic Instruction-Tuning Data Selection
Yixin Yang (Peking University), Zhifang Sui (Peking University)
Data-Centric LearningLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: Propose the RICO (Refined In-Context Contribution for Instruction-Tuning Data Selection) method, which uses gradient-free in-context learning to evaluate the fine-grained contribution of individual training samples to LLM instruction tuning, and builds a lightweight sample selection model based on this.
RiemanLine: Riemannian Manifold Representation of 3D Lines for Factor Graph Optimization
Yan Li, Gim Hee Lee (University Of Edinburgh)
Optimization
🎯 What it does: Proposed a Riemannian manifold-based 3D line minimal parameterization (RiemanLine) that simultaneously handles single lines and parallel line sets, significantly reducing parameter dimensionality;
Riemannian Manifold Learning for Stackelberg Games with Neural Flow Representations
Larkin Liu (Technische Universitat Munchen), Jalal Etesami (Hugging Face)
OptimizationRepresentation LearningReinforcement LearningFlow-based Model
🎯 What it does: In the online learning framework of Stackelberg games, the joint action space is mapped to a spherical Riemannian manifold (Stackelberg manifold), leveraging neural normalizing flows to achieve reversible mapping, thereby transforming the original nonlinear decision problem into a linear bandit algorithm solvable on the manifold.
Right Branches Matter in Failure-based Variable Ordering Heuristics
Yang Zhang (Northeast Normal University), Hongbo Li (Northeast Normal University)
OptimizationBenchmark
🎯 What it does: Improved and experimented with a variable ordering heuristic based on failure rate and failure length using right branch failure information
Ripple Shapley: Data Influence Attribution in One Federated Training Run
Dewen Zeng (University of South China), Zhiyong Xu (Huazhong University of Science and Technology)
Federated LearningExplainability and InterpretabilityImageBenchmark
🎯 What it does: Proposed the Ripple Shapley framework, which real-time evaluates sample contributions in a single federated training process, decomposing them into drop and ripple components, and achieving efficient propagation through low-rank Jacobian chains.
RIS-LAD: A Benchmark and Model for Referring Image Segmentation in Low-Altitude Drone Imagery
Kai Ye (Xiamen University), Liujuan Cao (Xiamen University)
SegmentationTransformerLarge Language ModelVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: This paper proposes the Referring Image Segmentation (RIS) task for low-altitude drone images, constructs the RIS-LAD benchmark dataset, and designs the SAARN model.
Risk-Sensitive Exponential Actor Critic
Alonso Granados (University of Arizona), Jason Pacheco (University of Arizona)
Reinforcement Learning
🎯 What it does: Proposed the risk-sensitive exponential actor-critic (rsEAC) algorithm, combining a new gradient theorem and numerically stable exponential value function estimation to achieve deep reinforcement learning with entropy risk measures.
RL-U2Net: A Dual-Branch UNet with Reinforcement Learning-Assisted Multimodal Feature Fusion for Accurate 3D Whole-Heart Segmentation
Jierui Qu (National University of Singapore), Jianchun Zhao (Xi'an Jiaotong University)
SegmentationConvolutional Neural NetworkTransformerReinforcement LearningContrastive LearningMultimodalityBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: Propose a dual-branch U-Net combined with reinforcement learning (PPO) to achieve cross-modal feature alignment and fusion between CT and MRI, thereby enabling precise 3D whole-heart segmentation.
RLKD: Distilling LLMs’ Reasoning via Reinforcement Learning
Shicheng Xu (State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences), Xueqi Cheng (Huawei Inc.)
Knowledge DistillationTransformerSupervised Fine-TuningReinforcement LearningText
🎯 What it does: This paper proposes a reinforcement learning-based knowledge distillation framework, RLKD, which utilizes a Generative Structural Reward Model (GSRM) to transfer the teacher LLM's implicit multi-branch reasoning structure to the student LLM, addressing the issue that traditional SFT can only replicate surface-level reasoning paths.
RLMR: Reinforcement Learning with Mixed Rewards for Creative Writing
JianXing Liao (Tencent Hunyuan Team), Runzhi Shi (Peking University)
GenerationOptimizationReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningTextBenchmark
🎯 What it does: Propose a dynamic hybrid reward reinforcement learning framework RLMR to balance subjective quality and objective constraint satisfaction in creative writing.
RLSLM: A Hybrid Framework Combining Reinforcement Learning and a Rule-based Social Locomotion Model for Socially-aware Navigation
Yitian Kou (East China Normal University), Shu-Guang Kuai (East China Normal University)
Explainability and InterpretabilityReinforcement Learning
🎯 What it does: Propose the RLSLM hybrid framework, which integrates a psychology-based social comfort scene model into reinforcement learning rewards, achieving interpretable and learnable social navigation strategies;
RMAdapter: Reconstruction-based Multi-Modal Adapter for Vision-Language Models
Xiang Lin (Beihang University), Di Huang (National Computer Network Emergency Response Technical Team Coordination Center Of China)
ClassificationDomain AdaptationRepresentation LearningTransformerSupervised Fine-TuningVision Language ModelAuto EncoderImageTextMultimodality
🎯 What it does: Proposes RMAdapter, a dual-branch reconstructive multimodal adapter for efficiently fine-tuning pre-trained Vision-Language Models (VLMs) in few-shot scenarios.
RMFAT: Recurrent Multi-scale Feature Atmospheric Turbulence Mitigator
Zhiming Liu (University of Bristol), Nantheera Anantrasirichai
RestorationRecurrent Neural NetworkTransformerOptical FlowImageVideo
🎯 What it does: Proposed a lightweight recursive multi-scale feature atmospheric turbulence suppressor (RMFAT), which performs two-frame recursive restoration using only the current frame and the previous frame;
RMLer: Synthesizing Novel Objects Across Diverse Categories via Reinforcement Mixing Learning
Jun Li (Nanjing University of Science and Technology), Jian Yang (Nanjing University)
GenerationData SynthesisReinforcement LearningDiffusion modelImage
🎯 What it does: Generate novel and balanced object images by dynamically mixing cross-category concepts through the reinforcement learning framework RMLer.
rMMEA: Robust Multi-Modal Entity Alignment with Missing and Noise Visual Modality
Lingbing Guo (Tianjin University), Xin Wang (Tianjin University)
Knowledge DistillationRepresentation LearningGraph Neural NetworkContrastive LearningMultimodalityBenchmark
🎯 What it does: Investigates the robustness of multimodal entity alignment (MMEA) under scenarios with missing visual modalities and noisy environments, proposing the rMMEA method that combines self-distilled ranking distillation with mutual information estimation.
RMO: Towards Better LLM Alignment via Reshaping Reward Margin Distributions
Yanchi Ru (Xi'an Jiaotong University), Xiangliang Zhang (University of Notre Dame)
Reinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposed the Reward Margin Optimization (RMO) framework, which enhances preference alignment in large language models (LLMs) by reshaping the reward margin distribution during data preparation, batch construction, and training processes.
RMSAGen: Integrating Multiple Sequence Alignment for Function RNA Design
Jiyue Jiang (Chinese University of Hong Kong), Yu Li (Chinese University of Hong Kong)
GenerationTransformerLarge Language ModelBiomedical Data
🎯 What it does: Developed a generative model called RMSAGen based on RNA multi-sequence alignment (MSA) for designing functional RNA sequences.
RoadSceneVQA: Benchmarking Visual Question Answering in Roadside Perception Systems for Intelligent Transportation System
Runwei Guan (Hong Kong University of Science and Technology), Yutao Yue (Hong Kong University of Science and Technology)
Autonomous DrivingTransformerLarge Language ModelVision Language ModelImageTextMultimodalityBenchmarkChain-of-Thought
🎯 What it does: This paper proposes the RoadSceneVQA dataset and the RoadMind model, aiming to enhance visual question answering and reasoning capabilities for roadside perception.
Robust Causal Discovery Under Imperfect Structural Constraints
Zidong Wang (City University of Hong Kong), Xiaoguang Gao (Northwestern Polytechnical University)
OptimizationExplainability and InterpretabilityGraphBiomedical Data
🎯 What it does: Propose a robust causal discovery framework named RoaDs, based on prior alignment and multi-task learning, for causal graph learning under imperfect structural constraints.
Robust Decentralized Multi-armed Bandits: From Corruption-Resilience to Byzantine-Resilience
Zicheng Hu (East China Normal University), Cheng Chen (East China Normal University)
OptimizationFederated LearningGraph
🎯 What it does: Proposed the DeMABAR algorithm to address robustness in decentralized multi-agent multi-armed bandits under adversarial corruption and Byzantine attacks.
Robust Detection of Synthetic Tabular Data Under Schema Variability
G. Charbel N. Kindji (Orange Labs Lannion), Tanguy Urvoy (Universit' e de Rennes)
Domain AdaptationAnomaly DetectionTransformerTabular
🎯 What it does: Propose a table data detection model called Datum-wise Transformer, specifically designed to identify real rows and synthetic rows;
Robust Domain Adaptive Hashing via Structural Noise Modeling and Correction
Junsheng Wang (Yangzhou University), Xiaobing Sun (Jiangxi Institute of Technology)
RetrievalDomain AdaptationContrastive LearningImage
🎯 What it does: Propose a robust domain adaptive hashing method, RDAH, aiming to simultaneously suppress the degradation of hash code quality caused by source domain label noise and cross-domain differences.
Robust Fusion Controller: Degradation-Aware Image Fusion with Fine-Grained Language Instructions
Hao Zhang (Wuhan University), Jiayi Ma (Wuhan University)
RestorationConvolutional Neural NetworkTransformerPrompt EngineeringVision Language ModelContrastive LearningImageText
🎯 What it does: Designed a robust fusion controller RFC to achieve degradation-aware image fusion through fine-grained language instructions.
Robust High-Order Tensor Compressive Sensing Based on M-Estimators
Xiaowei Wang (Engineering Research Center of Intelligent Technology for Agriculture Ministry of Education), Yulong Wang (Engineering Research Center of Intelligent Technology for Agriculture Ministry of Education)
RestorationCompressionImageVideo
🎯 What it does: Propose a robust high-order tensor compressed sensing method RTCS, which utilizes M-estimators to adaptively model low-rank structures and noise, supporting high-order tensor reconstruction;
Robust Integrative Analysis of Multi-omics Datasets via Nuclear-norm Maximization
Meng-Zhu Wang (Hebei University of Technology), Hongxing Zhang (National Center for Protein Sciences)
OptimizationRepresentation LearningGraph Neural NetworkContrastive LearningBiomedical Data
🎯 What it does: Developed a spatial multi-omics integration method RIA based on batch nuclear norm maximization (BNM), leveraging adaptive graph learning and dynamic prototype contrastive learning to construct more discriminative and diverse latent representations.
Robust Lazy Conflict Detection via Multi-Conflict Extraction and Genetic Diversity Control
Viet-Man Le (Graz University of Technology), Alexander Felfernig (Graz University of Technology)
OptimizationTabular
🎯 What it does: This study proposes an improved lazy conflict detection method, enhancing conflict set coverage and robustness through multi-conflict extraction and genetic diversity control.
Robust Learning from Noisily Labeled Long-Tailed Data via Fairness Regularizer
Jiaheng Wei (Hong Kong University of Science and Technology (Guangzhou)), Yang Liu (University of California, Santa Cruz)
ClassificationData-Centric LearningImage
🎯 What it does: This paper studies the robust learning problem in scenarios where label noise and long-tailed distribution coexist, proposing a Fair Regularization (FR) mechanism that encourages the model to maintain balanced performance across subgroups, thereby improving overall and tail performance.
Robust Long-Term Test-Time Adaptation for 3D Human Pose Estimation Through Motion Discretization
Yilin Wen (University of Tokyo), Yusuke Sugano (University of Tokyo)
Pose EstimationDomain AdaptationConvolutional Neural NetworkAuto EncoderVideo
🎯 What it does: Online adaptive 3D human pose estimation on unlabeled real-time video streams, continuously self-supervised model updates to bridge the training-test domain gap.
Robust Multiagent Combinatorial Path Finding
Yehonatan Kidushim (Ben Gurion University of Negev), Meir Kalech (Ben Gurion University of Negev)
OptimizationRobotic IntelligenceGraphBenchmark
🎯 What it does: This paper proposes a robust multi-agent combined path planning framework named Robust CBSS, and implements two algorithms, RCbssBase and RCbssEff, which can accomplish dynamic task allocation and path planning in real robot environments with directional constraints and probabilistic delays.
Robust Noise Modeling for Spike Camera via Time-Interval Quantification and Spike-DSLR Multimodal Dataset in Low-Light Imaging
Yue Cao (Harbin Engineering University), Liguo Zhang (Harbin Engineering University)
RestorationConvolutional Neural NetworkMultimodality
🎯 What it does: This paper addresses noise modeling and image reconstruction for pulse cameras under extremely low-light conditions, proposing a time-interval-based noise quantification method and constructing the Spike-DSLR multimodal dataset.
Robust Out-of-Order Retrieval for Grid-Based Storage at Maximum Capacity
Tzvika Geft (Rutgers University), Kostas Bekris (Rutgers University)
OptimizationSequential
🎯 What it does: Propose a robust framework for two-dimensional grid high-density storage systems that achieves zero or minimal relocations when storage and retrieval sequences experience k-bound perturbations.
Robust Pedestrian Detection with Uncertain Modality
Qian Bie (Wuhan University of Science and Technology), Xin Xu (Wuhan University of Science and Technology)
Object DetectionConvolutional Neural NetworkImageMultimodality
🎯 What it does: Constructed the TRNT dataset containing three modalities (RGB, NIR, TIR), and proposed AUNet to achieve robust pedestrian detection under any modality combination;
Robust Pseudo-Labeling via Decoupled Class-Aware Filtering and Dynamic Category Correction
Jianghang Lin (Xiamen University), Liujuan Cao (Xiamen University)
SegmentationTransformerVision Language ModelImage
🎯 What it does: To address the pseudo-label noise issue in semi-supervised instance segmentation, the PL-DC framework is proposed, which includes decoupled filtering, dynamic class correction, and pixel-level uncertainty weighting.
Robust SDE Parameter Estimation Under Missing Time Information Setting
Van Long Tran (Deakin University), Phuoc Nguyen (Deakin University)
OptimizationDrug DiscoveryScore-based ModelTime SeriesBiomedical DataStochastic Differential Equation
🎯 What it does: Studied a method that first restores the temporal order in observation sequences with missing or scrambled timestamps before performing stochastic differential equation (SDE) parameter estimation.
Robust Semi-paired Multimodal Learning for Cross-modal Retrieval
Yang Qin, Peng Hu (Sichuan University)
RetrievalContrastive LearningMultimodality
🎯 What it does: This paper studies image-text retrieval in the semi-aligned multimodal learning (SPL) scenario and proposes a robust cross semi-aligned learning (RCSL) framework;
Robust Single-Stage Fully Sparse 3D Object Detection via Detachable Latent Diffusion
Wentao Qu, Liang Xiao (Nanjing University Of Science And Technology)
Object DetectionAutonomous DrivingDiffusion modelAuto EncoderPoint Cloud
🎯 What it does: Proposes RSDNet, a single-stage fully sparse 3D object detection network based on a detachable latent framework, which utilizes diffusion models to learn multi-level denoising in the feature space, achieving robustness against various perturbations.
Robust Watermarking on Gradient Boosting Decision Trees
Jun Woo Chung (Rochester Institute of Technology), Weijie Zhao (Rochester Institute of Technology)
Safty and PrivacySupervised Fine-TuningTabular
🎯 What it does: Propose a robust watermarking scheme for Gradient Boosted Decision Trees (GBDT) that can be maintained after deployment
Robust-R1: Degradation-Aware Reasoning for Robust Visual Understanding
Jiaqi Tang (Hong Kong University of Science and Technology), Qifeng Chen (Nanjing University of Science and Technology)
Explainability and InterpretabilitySupervised Fine-TuningReinforcement LearningVision Language ModelMultimodalityChain-of-Thought
🎯 What it does: Propose the Robust-R1 framework, achieving robust understanding of multi-modal large language models under real-world visual degradation through structured degradation perception chains.
RobusTor3D: Robust Multimodal 3D Object Detector for Autonomous Driving by Vision-Language Knowledge Blending
Ying Yang (Beijing Jiaotong University), Zhengyin Liang (Beijing Jiaotong University)
Autonomous DrivingConvolutional Neural NetworkTransformerContrastive LearningTextMultimodalityPoint Cloud
🎯 What it does: Proposed a robust multi-modal 3D detector called RobusTor3D, which enhances model robustness against long-tail distributions, adverse weather, sensor errors, modality missing, and cross-domain scenarios through knowledge from vision-language models at both structural and supervisory layers.
Role Hypergraph Contrastive Learning for Multivariate Time-Series Analysis
Rundong Xue (Xi'an Jiaotong University), Yue Gao (Tsinghua University)
ClassificationGraph Neural NetworkTransformerContrastive LearningTime Series
🎯 What it does: Propose a contrastive learning framework based on role hypergraphs, leveraging role hypergraphs to capture higher-order associations in multivariate time series, and achieving learning of spatial short-term consistency and long-term evolution through structural and feature contrast.
RoS-Guard: Robust and Scalable Online Change Detection with Delay-Optimal Guarantees
Zelin Zhu (Tongji University), Kai Yang (Tongji University)
Anomaly DetectionOptimizationComputational Efficiency
🎯 What it does: The paper addresses the online change detection problem under system parameter uncertainty by proposing the RoS-Guard algorithm and implementing a GPU-accelerated neural unfolding solver;
RoSA: Enhancing Parameter-Efficient Fine-Tuning via RoPE-aware Selective Adaptation in Large Language Models
Dayan Pan (Beihang University), Xiangyu Zhao (City University of Hong Kong)
OptimizationComputational EfficiencyRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Designed a new parameter-efficient fine-tuning framework named RoSA, which integrates RoPE-aware attention enhancement and dynamic layer selection;
RoSE: A Role Correlation Structure-Enhanced Model for Multi-Event Argument Extraction
Geting Huang (Sichuan University), Xiuyuan Xu (Sichuan University)
RecognitionGraph Neural NetworkTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Propose a multi-event argument extraction model called RoSE, which explicitly models the heterogeneity and overlap of event structures through role-related structures to improve the accuracy of multi-event argument extraction.
RouterNet: Hierarchical Point Routing Network for Robust Vertebral Landmark Localization on AP X-ray Images
Yingjie Guo (Huazhong University of Science and Technology), Zhiwei Wang (Huazhong University of Science and Technology)
Pose EstimationConvolutional Neural NetworkImageBiomedical Data
🎯 What it does: Proposed a hierarchical point routing network called RouterNet for precise localization of vertebral landmark points in AP X-ray images, and utilized the localization results for automatic assessment of scoliosis.
ROVER: Robust Generative Continual Identity Unlearning Against Relearning Attacks
Tairan Huang (Central South University), Xiu Su (Central South University)
GenerationAuto EncoderGenerative Adversarial NetworkImage
🎯 What it does: Designed and implemented the ROVER framework to achieve robust multi-identity continual forgetting in generative models, addressing two major challenges: identity conflicts and relearnable attacks.
RPE-PAD: Relative Pose Estimation for Pose-agnostic Anomaly Detection
Zhipeng Zhang (East China Normal University), Zhi Li (East China Normal University)
Pose EstimationAnomaly DetectionConvolutional Neural NetworkGaussian SplattingImage
🎯 What it does: This paper proposes a query- and pose-agnostic anomaly detection framework called RPE-PAD based on relative pose estimation, designed to detect and locate anomalies in test sets without pose information.
RPGen: Robust and Differentially Private Synthetic Image Generation
Zihao Wang (Nanyang Technological University), Zhengtao Yu (Nanyang Technological University)
GenerationData SynthesisSafty and PrivacyDiffusion modelImage
🎯 What it does: Proposed the RPGen framework, which enhances the robustness of diffusion models against DP noise during the public pre-training phase by leveraging adversarial model perturbation, and achieves high-quality synthetic image generation under different privacy budgets through privacy-friendly domain-aligned data selection.
RPM-MCTS: Knowledge-Retrieval as Process Reward Model with Monte Carlo Tree Search for Code Generation
Yuanyuan Lin (Huazhong University of Science and Technology), Kaixin Sui (ByteDance Seed)
GenerationTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: This paper proposes a Monte Carlo Tree Search method (RPM-MCTS) that combines knowledge base retrieval with sandbox feedback for generating accurate code, and implements a process reward model without requiring an additional training process.
RPTS: Tree-Structured Reasoning Process Scoring for Faithful Multimodal Evaluation
Haofeng Wang (Harbin Institute of Technology), Yu Zhang (Harbin Institute of Technology)
Explainability and InterpretabilityLarge Language ModelMultimodalityBenchmarkChain-of-Thought
🎯 What it does: This paper proposes a tree-structured reasoning process scoring metric called RPTS, and builds a multimodal reasoning benchmark named RPTS-Eval based on it.
RS2-SAM2: Customized SAM2 for Referring Remote Sensing Image Segmentation
Fu Rong (Wuhan University), Lefei Zhang (Wuhan University)
SegmentationTransformerVision Language ModelImageMultimodalityBenchmark
🎯 What it does: Propose the RS2-SAM2 framework, adapting the Segment Anything Model 2 (SAM2) to remote sensing image referring expression segmentation tasks through components such as a joint encoder, bidirectional hierarchical fusion module, and mask prompt generator.
RSA-CR: Resisting Shilling Attacks in Citation Recommendation via Dumbbell Inductive Learning
Xiyue Gao (Xidian University), Jiangtao Cui (Xidian University)
Recommendation SystemAdversarial AttackGraph Neural NetworkTextGraphBenchmark
🎯 What it does: This study first formalizes shilling attacks targeting citation recommendation in academic papers and proposes the RSA-CR algorithm, which achieves global citation recommendation with robustness against attacks through dual-layer academic graphs and confidence-guided aggregation (Dumbbell Inductive Learning).
RSOD: Reliability-Guided Sonar Image Object Detection with Extremely Limited Labels
Chengzhou Li (Dalian University of Technology), Xin Fan (Dalian University of Technology)
Object DetectionConvolutional Neural NetworkImage
🎯 What it does: Proposes the RSOD semi-supervised sonar image object detection framework, which uses a teacher-student model to evaluate the reliability of pseudo-labels and generate hybrid pseudo-labels, significantly enhancing detection performance even with minimal annotations.
RSPlace: Rotation Sensing Macro Placement via Bidirectional Tree Expansion
Tianyi Liu (Nanjing University of Aeronautics and Astronautics), Yu Wang (Nanjing University of Aeronautics and Astronautics)
OptimizationReinforcement LearningTabularBenchmark
🎯 What it does: This paper proposes a macro placement method called RSPlace based on reinforcement learning, which first incorporates the rotation angle of macros into the search space;