AAAI 2026 Papers — Page 30
AAAI Conference on Artificial Intelligence · 4149 papers
Promises Made, Promises Kept: Safe Pareto Improvements via Ex Post Verifiable Commitments
Nathaniel Sauerberg (University of Texas at Austin), Caspar Oesterheld (Carnegie Mellon University)
🎯 What it does: This paper explores three forms of secure Pareto improvement (SPI) achieved through post-hoc verifiable commitments: elimination commitment of disabled actions, token game commitment, and default strategy mapping commitment, and provides computational complexity results for determining the existence and construction methods of SPI under different game settings.
Promoting Efficient Reasoning with Verifiable Stepwise Reward
Chuhuai Yue (Meituan), Guojun Yin (Meituan)
Computational EfficiencyReinforcement LearningTextBenchmark
🎯 What it does: Designed a verifiable step-wise reward mechanism (VSRM) to enhance the inference efficiency of large reasoning models and reduce overthinking.
PromptEmo: Learning Emotion with Bilateral Textual Prompts in Multi-Domain Open-set Scenarios
Xinyi Zeng (Sichuan University), Yan Wang (Chengdu University of Information Technology)
ClassificationRecognitionDomain AdaptationTransformerLarge Language ModelPrompt EngineeringVision Language ModelContrastive LearningImageMultimodality
🎯 What it does: Proposed a bidirectional text prompting framework called PromptEmo based on CLIP to address the multi-domain open-set facial expression recognition (MO-FER) task.
Prompting Adversarial Transferability via Path Flatness Attack
Zeze Tao (Hebei University), Huibing Wang (Dalian Maritime University)
Adversarial AttackImage
🎯 What it does: This paper proposes a path-based flatness attack method called PFA, which significantly enhances the transfer attack effectiveness against black-box models by constraining the flatness of the loss surface along the entire path from the current point to the local minimum.
PromptMoE: Generalizable Zero-Shot Anomaly Detection via Visually-Guided Prompt Mixtures
Yuheng Shao (Tongji University), Qinyuan Liu (Tongji University)
Anomaly DetectionPrompt EngineeringMixture of ExpertsVision Language ModelImageBiomedical Data
🎯 What it does: This paper proposes a vision-guided compositional prompt generation framework named PromptMoE for zero-shot anomaly detection.
Promptus: Can Prompt Streaming Replace Video Streaming
Jiangkai Wu (Wangxuan Institute of Computer Technology, Peking University), Xinggong Zhang (Wangxuan Institute of Computer Technology, Peking University)
GenerationCompressionPrompt EngineeringDiffusion modelVideo
🎯 What it does: Convert video frames into prompts for Stable Diffusion, generating video at the receiver end to achieve ultra-low bitrate video streaming.
Proof Systems for Tensor-based Model Counting
Olaf Beyersdorff (Friedrich Schiller University Jena), Christoph Staudt (Friedrich Schiller University Jena)
Computational Efficiency
🎯 What it does: This paper conducts a theoretical complexity analysis of model counting algorithms based on tensor networks, proposes two new tensor proof systems T dense and T sparse, and proves their correspondence with tensor compression solving.
Proof Systems That Tightly Characterise Model Counting Algorithms
Olaf Beyersdorff (Friedrich Schiller University Jena), Kaspar Kasche (Friedrich Schiller University Jena)
🎯 What it does: This paper proposes two new #SAT proof systems, KC PS Syn and KC PS Sem, and proves their equivalence with existing systems such as MICE Ref and proof systems based on Decision-DNNF. It further clarifies the theoretical foundations of these systems for current state-of-the-art model counters.
ProPL: Universal Semi-Supervised Ultrasound Image Segmentation via Prompt-Guided Pseudo-Labeling
Yaxiong Chen (Wuhan University of Technology), Lichao Mou (Technical University of Munich)
SegmentationConvolutional Neural NetworkTransformerPrompt EngineeringVision Language ModelImageBiomedical DataUltrasound
🎯 What it does: This paper proposes the ProPL framework, achieving multi-organ, multi-task unsupervised semi-supervised ultrasound image segmentation.
ProRec-Video: Guiding Hierarchical Interest Transitions for Proactive Short Video Recommendation with Dynamic Feedback Adaptation
Weizhi Chen, Houjie Qiu (National University Of Defense Technology)
Recommendation SystemLarge Language ModelAgentic AIVideoMultimodalityRetrieval-Augmented Generation
🎯 What it does: Propose the ProRec-Video framework for active recommendation in short videos, supporting personalized interest transfer and dynamic feedback adaptation.
Prototype Entropy Alignment: Reinforcing Structured Uncertainty in LLM Reasoning
Zhengyuan Pan (Xiamen University), Qingqiang Wu (University of Chinese Academy of Sciences)
Explainability and InterpretabilityRepresentation LearningReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningText
🎯 What it does: Proposed Prototype Entropy Alignment (PEA), a reinforcement learning framework that clusters high-entropy token distributions from expert reasoning trajectories into dynamically updatable prototypes, using these prototypes as process-level rewards to guide large language models' reasoning processes.
Prototype-Based Semantic Consistency Alignment for Domain Adaptive Retrieval
Tianle Hu, Guoxu Zhou (Harbin Institute Of Technology)
RetrievalDomain AdaptationAuto EncoderContrastive LearningImage
🎯 What it does: Proposed a two-stage prototype-based semantic consistency alignment framework called PSCA to address issues of semantic alignment, pseudo-label reliability, and quantization errors in domain adaptive retrieval.
Prototype-Driven Active Domain Adaptation with Density Consideration
Zeyu Zhang (Jilin University), Shuai Lü
Domain AdaptationGenerative Adversarial NetworkImageBenchmark
🎯 What it does: This paper proposes an active domain adaptation framework called PDADA, which enhances target domain performance by leveraging prototype-driven density-aware sample selection and adversarial training.
Prototype-Guided Supervision for Graph Learning with Noisy and Sparse Labels
Qiyu Li (Guangxi Normal University), Jinyan Wang (Guangxi Normal University)
ClassificationRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: Proposed a prototype-guided supervised framework that trains graph neural networks using prototypes instead of direct labels, enhanced by dual-branch mixup and radius regularization to improve representation learning.
ProtSAE: Disentangling and Interpreting Protein Language Models via Semantically-Guided Sparse Autoencoders
Xiangyu Liu (Nanjing University), Wei Hu (Nanjing University)
Explainability and InterpretabilityTransformerLarge Language ModelAuto EncoderBiomedical Data
🎯 What it does: Semantic-guided training of sparse autoencoders on protein language models to remove semantic contamination, enhance interpretability, and support generation control.
Provably Data-Driven Projection Method for Quadratic Programming
Anh Tuan Nguyen (Carnegie Mellon University), Viet Anh Nguyen (Chinese University of Hong Kong)
OptimizationComputational Efficiency
🎯 What it does: Learn a projection matrix to reduce the dimension of convex quadratic programming (QP), thereby improving solving speed
Provably Efficient Multi-Objective Bandit Algorithms Under Preference-Centric Customization
Linfeng Cao (Ohio State University), Ness B. Shroff (Ohio State University)
OptimizationReinforcement Learning
🎯 What it does: This paper proposes a user-preference-aware multi-objective multi-armed bandit (PAMO-MAB) framework and presents provably effective algorithms for unknown and hidden preferences;
Proxy Zero-Shot Hashing with Multimodal Fusion via Stable Diffusion
Hui Zhang (Ocean University of China), Yuan Cao (Ocean University of China)
RetrievalDiffusion modelContrastive LearningMultimodalityBenchmark
🎯 What it does: Propose a zero-shot hashing framework called PZSH based on Stable Diffusion for multi-modal pseudo image generation and alignment;
ProxyTTT: Proxy-driven Test-Time Training for Multi-modal Re-identification
Aihua Zheng (Anhui University), Yan Yan (Anhui University)
RecognitionTransformerVision Language ModelAuto EncoderContrastive LearningImageMultimodality
🎯 What it does: Propose a unified framework named ProxyTTT, which combines multi-modal proxy learning (MCP+MSP) with selective adaptation based on proxy entropy (PESA), enhancing the discriminability of multi-modal identity features during training and further reducing domain discrepancies at test time by adaptively selecting high-confidence samples.
Prune&Comp: Free Lunch for Layer-Pruned LLMs via Iterative Pruning with Magnitude Compensation
Xinrui Chen (Tsinghua University), Chun Yuan (Tsinghua University)
Computational EfficiencyTransformerLarge Language ModelTextBenchmark
🎯 What it does: This paper proposes a training-agnostic iterative layer pruning method called PRUNE&COMP, which utilizes amplitude compensation techniques to recover the hidden state amplitude gaps caused by removed layers during pruning, thereby maintaining model performance;
Prune4Web: DOM Tree Pruning Programming for Web Agent
Jiayuan Zhang (Beihang University), Jing Zhang (Beihang University)
AI Code AssistantReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningAgentic AIVision-Language-Action ModelImageTextMultimodalityBenchmark
🎯 What it does: Propose the Prune4Web framework, which generates executable Python programs via LLM to perform programmatic pruning of web DOM trees, significantly reducing candidate elements and improving subtask localization accuracy.
PSA-MF: Personality-Sentiment Aligned Multi-Level Fusion for Multimodal Sentiment Analysis
Heng Xie, Changsheng Li (Beijing Institute Of Technology)
ClassificationRecurrent Neural NetworkTransformerContrastive LearningVideoTextMultimodalityAudio
🎯 What it does: Propose a multi-level fusion framework PSA-MF based on personality-emotion alignment for multimodal sentiment analysis.
PSEO: Optimizing Post-hoc Stacking Ensemble Through Hyperparameter Tuning
Beicheng Xu (Peking University), Bin Cui (Peking University)
OptimizationHyperparameter SearchTabular
🎯 What it does: Propose the PSEO framework, which enhances AutoML's predictive performance through hyperparameter optimization of post-stacking ensembles.
Pseudo Multi-view K-means Clustering
Jinqian Chen (Xi'an Jiaotong University), Qinghai Zheng (Fuzhou University)
OptimizationRepresentation LearningTabular
🎯 What it does: Proposed a pseudo-multi-view K-means clustering (PMKC) that simulates multi-view learning by generating multiple soft k-means decompositions on single-view data to enhance clustering performance.
Pseudo-Spiking Neurons: A Noise-Based Training Framework for Heterogeneous-Latency Spiking Neural Networks
Yuxuan Zhang (Beihang University), Wen Yao (Defense Innovation Institute, Chinese Academy of Military Science)
Computational EfficiencyConvolutional Neural NetworkSpiking Neural NetworkImage
🎯 What it does: Proposes Pseudo Spiking Neuron (PseudoSN), a noise-based training proxy that approximates the temporal dynamics of spiking neural networks (SNNs) within a single forward pass, treating neuronal delays as learnable parameters; simultaneously optimizes accuracy, delay, and energy consumption through a hardware-aware objective; performs post-training fine-tuning to obtain deployable SNNs with heterogeneous delays;
PSPO: Prompt-Level Prioritization and Experience-Weighted Smoothing for Efficient Policy Optimization
Xinxin Zhu (Shenzhen University), F. Richard Yu (Guangdong Laboratory of Artificial Intelligence and Digital Economy)
OptimizationComputational EfficiencyReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningPrompt EngineeringTextBenchmark
🎯 What it does: Proposed a lightweight PSPO algorithm that improves data efficiency and training stability in LLM alignment through experience-weighted reward smoothing and prompt-level prioritized sampling.
PsyPARSE: Retrieval-Augmented Slow Thinking for Personalized Empathetic Counseling
Longxiang Wang (Chongqing University), Ronghao Chen (Peking University)
Prompt EngineeringTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Proposes a training-free psychological counseling framework, PsyPARSE, that simulates the deep thinking and personalized responses of human psychotherapists.
Public Goods Games in Directed Networks with Constraints on Sharing
Argyrios Deligkas (Royal Holloway University of London), Anders Yeo (Independent Researcher)
Graph
🎯 What it does: This paper studies public goods games in directed networks with a capacity constraint k, and analyzes the existence, computational complexity, and efficiency of pure and mixed Nash equilibria;
PUFM: Efficient Point Cloud Upsampling via Flow Matching
Zhi-Song Liu (Lappeenranta-Lahti University Of Technology), Lei Li (University Of Leicester)
GenerationComputational EfficiencyGraph Neural NetworkFlow-based ModelPoint Cloud
🎯 What it does: Propose a point cloud up-sampling framework PUFM based on flow matching, which directly learns the optimal transport path from sparse point clouds to dense point clouds, achieving efficient and fine-grained point cloud up-sampling.
PulseMind: A Multi-Modal Medical Model for Real-World Clinical Diagnosis
Jiao Xu (Dalian University of Technology), Ping Wang (Peking University)
TransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningVision Language ModelMultimodalityBiomedical DataBenchmark
🎯 What it does: This paper proposes PulseMind, a multimodal medical large model designed for real-world clinical diagnosis, encompassing the newly constructed MediScope dialogue dataset, the PulseMind Benchmark evaluation framework, and the CRPO reinforcement learning training framework.
PUNO: A Neural Operator Framework for Point Cloud Upsampling
Zijian Xiao (Southeast University), Li Yao (Southeast University)
Super ResolutionConvolutional Neural NetworkGraph Neural NetworkPoint Cloud
🎯 What it does: Proposes the PUNO framework, which utilizes neural operators to perform high-quality upsampling of sparse point clouds in function space.
PurMM: Attention-Guided Test-Time Backdoor Purification in Multimodal Large Language Models
Wenzheng Jiang (National University of Defense Technology), Ji Wang (Wuhan University)
Safty and PrivacyAdversarial AttackTransformerLarge Language ModelVision Language ModelMultimodality
🎯 What it does: To address the vulnerability of multi-modal large language models to backdoor attacks during downstream fine-tuning, this paper proposes PurMM, an attention-guided test-time backdoor purification framework.
Pushing Rendering Boundaries: Hard Gaussian Splatting
Qingshan Xu, Hanwang Zhang (Nanyang Technological University)
GenerationNeural Radiance FieldGaussian SplattingImage
🎯 What it does: Propose the Hard Gaussian Splatting (HGS) framework, which leverages multi-view significant location gradients and rendering errors to mine and grow hard Gaussians, significantly improving the quality of Neural View Synthesis (NVS) for 3D scenes;
Put the Space of LoRA Initialization to the Extreme to Preserve Pre-trained Knowledge
Pengwei Tang (Renmin University of China), Debing Zhang (Xiaohongshu Inc)
Computational EfficiencyKnowledge DistillationRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Designed a new LoRA initialization method called LoRA-Null, which utilizes the null space of pre-trained knowledge activation to initialize the LoRA adapter, thereby better preserving the knowledge of the pre-trained model during fine-tuning.
Putting Fair Division on the Map
Paula Böhm (Institut für Informatik TU Clausthal), Stanisław Szufa (University of Chicago)
OptimizationExplainability and InterpretabilityTabular
🎯 What it does: This paper constructs two methods to map fair division instances of indivisible goods to a two-dimensional plane, aiming to reveal structural differences between instances and analyze the distributions and properties of real and synthetic instances.
Q Cache: Visual Attention Is Valuable in Less than Half of Decode Layers for Multimodal Large Language Model
Jiedong Zhuang (Zhejiang University), Haoji Hu (Alibaba Cloud Computing)
Computational EfficiencyTransformerLarge Language ModelVision Language ModelMultimodality
🎯 What it does: Studied a cross-layer shared attention mechanism (Lazy Attention) and QCache caching strategy to reduce KV cache usage and computational load during the decoding phase of multi-modal large language models (MLLMs), while maintaining almost identical performance.
qa-FLoRA: Data-free query-adaptive Fusion of LoRAs for LLMs
Shreya Shukla (Mercedes Benz Research and Development India), Hiteshi Jain (Mercedes Benz Research and Development India)
Computational EfficiencyRepresentation LearningTransformerLarge Language ModelText
🎯 What it does: Proposes a query-adaptive LoRA fusion method called qa-FLoRA that requires no training or data, dynamically computing fusion weights using hierarchical KL divergence;
QAPNet: A Quantum-Attentive Patchwise Network for Robust Medical Image Classification Under Noisy Inputs
Maqsudur Rahman (Boise State University), Jun Zhuang (Boise State University)
ClassificationContrastive LearningImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: Designed and implemented QAPNet, which encodes image patches through quantum variational circuits, reweights instances using additive attention, and incorporates supervised contrastive learning with prototype regularization to achieve robust medical image classification under Gaussian noise input.
QiMeng-CRUX: Narrowing the Gap Between Natural Language and Verilog via Core Refined Understanding eXpression
Lei Huang (Chinese Academy of Sciences), Qi Guo (Chinese Academy of Sciences)
GenerationAI Code AssistantTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningPrompt EngineeringText
🎯 What it does: Propose a structured intermediate representation space called CRUX to bridge the gap between natural language descriptions and Verilog code, and build a two-phase training framework (Joint Expression Modeling and Dual-Space Optimization) to achieve more accurate and synthesizable Verilog code generation.
QiMeng-Kernel: Macro-Thinking Micro-Coding Paradigm for LLM-Based High-Performance GPU Kernel Generation
Xinguo Zhu (Chinese Academy of Sciences), Ling Li (Chinese Academy of Sciences)
OptimizationTransformerLarge Language ModelReinforcement LearningPrompt EngineeringTextBenchmark
🎯 What it does: Propose a hierarchical macro thinking micro-encoding (MTMC) framework that leverages LLMs to first generate high-level optimization strategies and then progressively implement low-level code, enabling automated generation of high-performance GPU kernels.
QRShield: Exploiting Vulnerabilities of Latent Diffusion Models for Preventing AI Art Plagiarism
Xunyue Mo (Sun Yat-sen University), Zibin Zheng (Sun Yat-sen University)
GenerationSafty and PrivacyAdversarial AttackDiffusion modelAuto EncoderImage
🎯 What it does: This paper proposes QRShield, a lightweight adversarial artwork protection method targeting Latent Diffusion Models.
Qualitative Analysis of ω-Regular Objectives on Robust MDPs
Ali Asadi (Institute of Science and Technology Austria), Ali Shafiee (Institute of Science and Technology Austria)
OptimizationComputational EfficiencyBenchmark
🎯 What it does: For robust Markov decision processes (RMDP), a qualitative analysis algorithm is proposed to solve reachability and parity (ω-regular) objectives, capable of determining whether the objective can be satisfied with probability 1 under all possible transition functions.
Quality-aware and Soft Consistency Driven Representation Fusion for Incomplete Multi-view Multi-label Classification
Yadong Liu (Harbin Institute of Technology), Jie Wen (Harbin Institute of Technology)
ClassificationRepresentation LearningContrastive LearningImage
🎯 What it does: To address the problem of missing views and labels in multi-view multi-label classification, the QUALITY framework is proposed, combining soft consistency constraints and quality-aware instance-level dynamic fusion to achieve robust learning on incomplete data.
Quality-Aware Language-Conditioned Local Auto-Regressive Anomaly Synthesis and Detection
Long Qian (Chinese Academy of Sciences), Jinqiao Wang (Chinese Academy of Sciences)
GenerationData SynthesisAnomaly DetectionTransformerVision Language ModelAuto EncoderImage
🎯 What it does: Proposes ARAS—a language-conditioned local autoregressive anomaly synthesis method, embedded into the QARAD quality-aware weighted detection framework, achieving high-resolution, text-controllable anomaly generation and detection.
QuanTaxo: A Quantum Approach to Self-Supervised Taxonomy Expansion
Sahil Mishra (Indian Institute of Technology Delhi), Tanmoy Chakraborty (Indian Institute of Technology Delhi)
Representation LearningText
🎯 What it does: Proposes QuanTaxo, a self-supervised hierarchical classification expansion framework based on quantum embeddings, designed to insert new concepts into existing taxonomies.
Quantifying and Improving Adaptivity in Conformal Prediction Through Input Transformations
Sooyong Jang (University of Pennsylvania), Insup Lee (University of Pennsylvania)
ClassificationRecognitionImage
🎯 What it does: This paper proposes a transformation ranking and balanced binning method based on input perturbation for constructing adaptive prediction sets in conformal prediction, and introduces two new evaluation metrics, T-CV and T-SS, to more accurately measure the adaptability of prediction sets to sample difficulty. Based on the binning results, it also introduces grouped conditional conformal prediction, leading to improved adaptive prediction set algorithms such as O-LAC and O-SAPS.
Quantifying the Potential to Escape Filter Bubbles: A Behavior-Aware Measure via Contrastive Simulation
Difu Feng (Chinese Academy of Sciences), Qingming Huang (University of Chinese Academy of Sciences)
Recommendation SystemTransformerLarge Language ModelContrastive LearningTabular
🎯 What it does: Conduct a quantitative evaluation of filter bubbles, proposing a behavior-based contrast metric called Bubble Escape Potential (BEP).
Quantum Algorithms for Spectral Sums
Alessandro Luongo (National University of Singapore), Changpeng Shao (Chinese Academy of Sciences)
Computational EfficiencyGraphPhysics Related
🎯 What it does: Proposed a series of quantum algorithms for estimating spectral sums (such as log determinant, von Neumann entropy, inverse trace, Schatten p-norm) of positive semi-definite matrices and graph theory-related metrics (such as triangle count, effective resistance, spanning tree count)
Quantum Lipschitz Bandits
Bongsoo Yi (University of North Carolina at Chapel Hill), Yao Li (University of North Carolina at Chapel Hill)
OptimizationPhysics Related
🎯 What it does: This paper proposes two quantum Lipschitz bandit algorithms, Q-LAE and Q-Zooming, and verifies their superiority both theoretically and experimentally.
Quantum Non-Linear Bandit Optimization
Zakaria Shams Siam (University at Albany State University of New York), Chong Liu (University at Albany State University of New York)
OptimizationHyperparameter SearchTabular
🎯 What it does: Proposed a quantum nonlinear bandit optimization algorithm, Q-NLB-UCB, which achieves logarithmic cumulative regret independent of input dimensions in high-dimensional spaces;
Quantum Transformer for Molecular Learning: Multi-Configuration Ground-State Energy Prediction
Yuichi Kamata (Fujitsu Limited), Hirotaka Oshima (Fujitsu Limited)
TransformerTabularPhysics Related
🎯 What it does: Proposed the Molecular Quantum Transformer (MQT) model, which uses a quantum attention mechanism to predict the ground state energy of various molecules across different configurations in a single step.
QuantVSR: Low-Bit Post-Training Quantization for Real-World Video Super-Resolution
Bowen Chai (Shanghai Jiao Tong University), Yulun Zhang (Shanghai Jiao Tong University)
Super ResolutionComputational EfficiencyVideo
🎯 What it does: This paper proposes QuantVSR, a low-bit post-training quantization framework for real-world video super-resolution.
QuEPT: Quantized Elastic Precision Transformers with One-Shot Calibration for Multi-Bit Switching
Ke Xu (Anhui University), Xingyi Zhang (Anhui University)
Computational EfficiencyTransformerLarge Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Propose QuEPT, a Transformer-based elastic multi-bit post-training quantization framework that supports real-time switching between multi-bit widths and quantization of weights and activations with a single calibration.
Query-Efficient Domain Knowledge Stealing Against Large Language Models
Zhengao Li (Florida State University), Yushun Dong (Florida State University)
Knowledge DistillationTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringBiomedical DataFinance RelatedChain-of-Thought
🎯 What it does: Developed a framework for domain knowledge extraction in a black-box LLM environment, leveraging self-guided exploration, structured questioning, and feedback loops to achieve low query volume.
Query-Routed Activation Editing with Truth-hierarchical Preference Optimization
Kewei Liao (Beihang University), Xianglong Liu (Beihang University)
TransformerLarge Language ModelReinforcement LearningText
🎯 What it does: Proposes the Query Routing and Activation Editing (QRAE) framework to dynamically route queries to the most relevant attention heads and customize activation edits to reduce hallucinations in large language models (LLMs).
QueryAligner: Customizing User Query to Match LLMs Preferences for Better Intent Recognition
Yunlong Ma (Tianjin University), Yuexian Hou (Harbin Institute of Technology)
RecognitionTransformerSupervised Fine-TuningReinforcement LearningPrompt EngineeringMixture of ExpertsContrastive LearningText
🎯 What it does: This paper proposes QueryAligner, which dynamically rewrites user queries to enhance intent recognition by addressing the language preferences of large language models.
QueryCraft: Transformer-Guided Query Initialization for Enhanced Human-Object Interaction Detection
Yuxiao Wang (South China University of Technology), Qi Liu (South China University of Technology)
Object DetectionKnowledge DistillationTransformerVision-Language-Action ModelImageMultimodality
🎯 What it does: Proposed the QueryCraft framework, significantly enhancing DETR-based human-object interaction detection through semantic prior-driven query initialization.
QuMAB: Query-based Multi-annotator Behavior Pattern Learning
Liyun Zhang (University of Tokyo), Yuta Nakashima (Tongji University)
Explainability and InterpretabilityRepresentation LearningTransformerVision Language ModelImageVideo
🎯 What it does: Proposed QuMAB, a query-based multi-annotator behavior pattern learning framework, focusing on capturing individual discriminative behaviors of each annotator rather than simply seeking sample-level consistency.
QuoTA: Query-oriented Token Assignment via CoT Query Decouple for Long Video Comprehension
Yongdong Luo (Xiamen University), Jiebo Luo (University of Rochester)
TransformerLarge Language ModelVision Language ModelVideoTextBenchmarkChain-of-Thought
🎯 What it does: Proposes a training-free query-oriented visual token allocation framework named QuoTA, enhancing the efficiency and effectiveness of long video understanding.
R-AVST: Empowering Video-LLMs with Fine-Grained Spatio-Temporal Reasoning in Complex Audio-Visual Scenarios
Zhu Lu, Feng Zheng (ZTE Corporation)
Object DetectionSegmentationLarge Language ModelReinforcement LearningVision Language ModelVideoMultimodality
🎯 What it does: This paper constructs the R-AVST dataset and proposes the AVST-Zero model based on this dataset, aiming to enhance the spatiotemporal reasoning capabilities of video large language models in complex audio-visual scenarios.
R-Tuning: Wavelet-Decomposed Replay and Semantic Alignment for Continual Adaptation of Pretrained Time-Series Models
Tianyi Yin (Tongji University), Weiming Shen (Zhongguancun Academy)
Domain AdaptationKnowledge DistillationTransformerTime Series
🎯 What it does: Proposes the Replay Tuning (R-Tuning) framework, which enables continuous adaptation of pre-trained time series models without original data by utilizing frequency-aware replay samples and semantic alignment.
R²D-LPCC: Relevance-Ranking Guided Region-Adaptive Dynamic LiDAR Point Cloud Compression
Fangzhe Nan (Zhejiang Gongshang University), Changshuo Wang (University College London)
CompressionAutonomous DrivingConvolutional Neural NetworkPoint CloudBenchmark
🎯 What it does: Propose a region-adaptive dynamic LiDAR point cloud compression framework R²D-LPCC based on semantic relevance ranking, which can significantly reduce the bit rate while preserving details in safety-critical regions.
RAA: Achieving Interactive Remove/Add Anything via Fully Synthetic Data
Delong Liu (Beijing University of Posts and Telecommunications), Zhicheng Zhao (Beijing University of Posts and Telecommunications)
Image HarmonizationGenerationData SynthesisTransformerLarge Language ModelDiffusion modelFlow-based ModelImageMultimodality
🎯 What it does: Propose a fully automated, self-improving synthetic data generation pipeline that generates over 514,510 pairs of high-quality image editing samples, and build the RAA framework to achieve precise object addition/removal editing based on this;
RABot: Reinforcement-Guided Graph Augmentation for Imbalanced and Noisy Social Bot Detection
Longlong Zhang (Northwestern Polytechnical University), Yang Liu (Xi'an Jiaotong University)
ClassificationData SynthesisAnomaly DetectionGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: Propose a RABot framework based on reinforcement learning and neighborhood-aware oversampling for social bot detection
RAC-DMVC: Reliability-Aware Contrastive Deep Multi-View Clustering Under Multi-Source Noise
Shihao Dong (Nanjing University of Information Science and Technology), Xinzhong Zhu (Zhejiang Normal University)
Representation LearningAuto EncoderContrastive LearningMultimodalityBenchmark
🎯 What it does: Proposes a Reliability-Aware Contrastive Deep Multi-View Clustering Framework (RAC-DMVC), capable of achieving unsupervised clustering in multi-source noisy environments where observation noise and missing noise coexist.
RacketVision: A Multiple Racket Sports Benchmark for Unified Ball and Racket Analysis
Linfeng Dong (University Of Science And Technology Of China), Xiao Sun (Zhejiang University)
Object DetectionObject TrackingPose EstimationConvolutional Neural NetworkRecurrent Neural NetworkTransformerVideoBenchmark
🎯 What it does: Proposes the RacketVision dataset and benchmark, covering three sports (table tennis, tennis, badminton), providing fine-grained annotations for ball position and racket pose, supporting three tasks: ball tracking, racket pose estimation, and trajectory prediction;
RaCoT: Plug-and-Play Contrastive Example Generation Mechanism for Enhanced LLM Reasoning Reliability
Kaitong Cai (Sun Yat-sen University), Keze Wang (Sun Yat-sen University)
RetrievalComputational EfficiencyTransformerLarge Language ModelPrompt EngineeringContrastive LearningTextRetrieval-Augmented Generation
🎯 What it does: RaCoT improves the reasoning reliability of retrieval-augmented generation by generating semantically similar but answer-different contrastive questions before retrieval and using difference prompts to guide LLMs to focus more precisely on key information.
Radar-APLANC: Unsupervised Radar-based Heartbeat Sensing via Augmented Pseudo-Label and Noise Contrast
Ying Wang (Nanjing University of Information Science & Technology), Xiaobai Li (Zhejiang University)
Data-Centric LearningContrastive LearningTime SeriesBiomedical Data
🎯 What it does: Proposed and implemented an unsupervised radar heartbeat detection framework called Radar-APLANC, which extracts heartbeat signals using enhanced pseudo labels and noise contrastive learning.
RadarLLM: Empowering Large Language Models to Understand Human Motion from Millimeter-wave Point Cloud Sequence
Zengyuan Lai (Shanghai Jiao Tong University), Ling Pei (Shanghai Jiao Tong University)
GenerationData SynthesisSafty and PrivacyRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelAuto EncoderTextPoint Cloud
🎯 What it does: Propose the RadarLLM framework, which converts millimeter-wave radar point cloud sequences into natural language descriptions, achieving privacy-preserving human motion semantic understanding.
RadarMP: Motion Perception for 4D mmWave Radar in Autonomous Driving
Ruiqi Cheng, Wei Liang (Beijing Racobit Electronic Information Technology Co Ltd)
Object DetectionAutonomous DrivingTransformerOptical FlowMultimodalityPoint Cloud
🎯 What it does: Propose the RadarMP framework, which utilizes 4D mmWave radar tensors from consecutive frames to simultaneously perform radar target detection and 3D scene flow estimation, achieving motion perception;
Rademacher Complexity for Distributionally Robust Learning
Zhengyu Zhou (Wuhan University), Weiwei Liu (Wuhan University)
Domain AdaptationOptimizationTabular
🎯 What it does: Study the upper bounds of generalization error in distributionally robust learning (DRL), adopting Rademacher complexity instead of traditional covering number arguments, and derive a convergence rate with error order O(n^{-1/(2k^*)}).
RAG-Enhanced Collaborative LLM Agents for Drug Discovery
Namkyeong Lee (Korea Advanced Institute of Science and Technology), Gabriele Scalia (Genentech)
Drug DiscoveryTransformerAgentic AITextGraphBiomedical DataRetrieval-Augmented Generation
🎯 What it does: Proposed a multi-agent, retrieval-augmented generation (RAG)-driven CLADD framework for zero-shot drug discovery question-answering tasks, capable of integrating external knowledge graphs and annotations without domain-specific fine-tuning.
RAG-R1:Incentivizing the Search and Reasoning Capabilities of LLMs Through Multi-Query Parallelism
Zhiwen Tan, Jinjie Gu (Ant Group)
RetrievalComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose the RAG-R1 framework, which enables LLMs to adaptively leverage internal and external knowledge during inference and support think-then-search interaction through two-phase training (Format Learning SFT + Retrieval-Enhanced RL);
RAGAR: Retrieval Augmented Personalized Image Generation Guided by Recommendation
Run Ling (Northeastern University), Xingwei Wang (Zhejiang University)
GenerationRecommendation SystemTransformerLarge Language ModelVision Language ModelDiffusion modelMultimodalityRetrieval-Augmented Generation
🎯 What it does: Proposed a retrieval-enhanced personalized image generation framework called RAGAR, which combines ideas from recommendation systems to perform semantic retrieval and weighting of user history, followed by generating images that meet user preferences using LLM and diffusion models.
RAGFort: Dual-Path Defense Against Proprietary Knowledge Base Extraction in Retrieval-Augmented Generation
Qinfeng Li (Zhejiang University), Xuhong Zhang (Zhejiang University)
Safty and PrivacyTransformerContrastive LearningTextRetrieval-Augmented Generation
🎯 What it does: To address knowledge base reconstruction attacks in RAG systems, this paper proposes RAGFort, a dual-path defense framework that combines mutual class isolation and intra-class suppression. It systematically evaluates the contributions of two attack paths and achieves comprehensive protection.
RAIN: Redundancy-Aware Latent Injection for Quality-Preserving Image Watermarking
Yehan Sun (Beijing Jiaotong University), Yao Zhao (Beijing Jiaotong University)
GenerationDiffusion modelAuto EncoderImage
🎯 What it does: Designed the RAIN framework in the latent space of diffusion models to inject high-capacity RGB watermarks while maintaining high visual quality of generated images.
RaLD: Generating High-Resolution 3D Radar Point Clouds with Latent Diffusion
Ruijie Zhang (Huazhong University of Science and Technology), Wei Wang (Wuhan University)
GenerationData SynthesisTransformerDiffusion modelAuto EncoderPoint Cloud
🎯 What it does: Propose the RaLD framework, which utilizes latent diffusion models to generate high-resolution LiDAR-like 3D point clouds from millimeter-wave radar spectra;
RaLiFlow: Scene Flow Estimation with 4D Radar and LiDAR Point Clouds
Jingyun Fu (Zhejiang University), Na Zhao (Singapore University of Technology and Design)
Autonomous DrivingConvolutional Neural NetworkRecurrent Neural NetworkOptical FlowMultimodalityPoint Cloud
🎯 What it does: Proposes the RaLiFlow framework, achieving the first joint scene flow estimation of 4D millimeter-wave radar and LiDAR point clouds;
Random Amalgamation of Adapters for Flatter Loss Landscapes: Towards Class-Incremental Learning with Better Stability
Yao Deng (Huazhong University of Science and Technology), Jiaqi Gui (Huazhong University of Science and Technology)
ClassificationOptimizationRepresentation LearningTransformerImage
🎯 What it does: This paper proposes a category-incremental learning method for pre-trained models, significantly enhancing the model's stability and performance during the incremental learning process through the combination of Randomly Mixed Adapters (RAA) and an improved Sharpness-Aware Minimization (SAM+).
Random is Faster than Systematic in Multi-Objective Local Search
Zimin Liang (University of Birmingham), Miqing Li (University of Birmingham)
OptimizationBenchmark
🎯 What it does: Investigate the efficiency differences between random exploration of neighborhoods and systematic exploration in multi-objective local search.
RankList – a Listwise Preference Learning Framework for Predicting Subjective Preferences
Abinay Reddy Naini (Carnegie Mellon University), Carlos Busso (Carnegie Mellon University)
Recommendation SystemComputational EfficiencyData-Centric LearningConvolutional Neural NetworkTransformerSupervised Fine-TuningContrastive LearningImageAudio
🎯 What it does: Propose RankList, a method that extends RankNet's contrastive loss into a complete list-level preference learning framework, enhancing global ranking consistency through log-sum-exp approximation and skip-wise comparisons.
RAPTOR: Real-Time High-Resolution UAV Video Prediction with Efficient Video Attention
Zhan Chen (Aerospace Information Research Institute Chinese Academy of Sciences), Yidan Zhang (Aerospace Information Research Institute Chinese Academy of Sciences)
GenerationAutonomous DrivingComputational EfficiencyVideo
🎯 What it does: Proposes the RAPTOR framework for single-channel real-time generation of high-resolution future video frames, addressing the speed-resolution-quality trilemma in UAV navigation.
RareAgents: Autonomous Multi-disciplinary Team for Rare Disease Diagnosis and Treatment
Xuanzhong Chen (Tsinghua University), Ting Chen (Tsinghua University)
TransformerLarge Language ModelAgentic AIBiomedical DataRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Developed a multi-disciplinary team (MDT) intelligent agent system called RareAgents based on a large language model for the diagnosis and treatment decision support of rare diseases.
Rational Revision of Group Intentions
Nima Motamed, Dragan Doder (Utrecht University)
🎯 What it does: Proposes a formal framework based on ATL-strategy context for rational revision of group intentions in multi-agent systems, along with corresponding representative theorems.
RatioSketch: Towards More Accurate Frequency Estimation in Data Streams via a Lightweight Neural Network
Mengbo Wang (Tsinghua University), Mingwei Xu (Peking University)
Computational EfficiencySupervised Fine-TuningTime Series
🎯 What it does: Proposed a lightweight neural network correction framework called RatioSketch to enhance the frequency estimation accuracy of traditional hand-crafted sketches (e.g., CMS, TowerSketch) and existing neural sketches (MetaSketch, LegoSketch).
RAW-Flow: Advancing RGB-to-RAW Image Reconstruction with Deterministic Latent Flow Matching
Zhen Liu (University of Electronic Science and Technology of China), Shuaicheng Liu (Independent Researcher)
Image TranslationRestorationFlow-based ModelAuto EncoderImage
🎯 What it does: Propose a RGB→RAW image reconstruction framework (RAW-Flow) based on potential flow matching, treating inverse ISP as a deterministic latent space transmission problem.
RayD3D: Distilling Depth Knowledge Along the Ray for Robust Multi-View 3D Object Detection
Rui Ding (Xi'an Jiaotong University), Gang Hua (Amazon)
Object DetectionKnowledge DistillationContrastive LearningMultimodalityPoint Cloud
🎯 What it does: Propose the RayD3D method, which enhances the robustness of multi-view 3D detection in noisy environments through cross-modal knowledge distillation along camera rays.
RcAE: Recursive Reconstruction Framework for Unsupervised Industrial Anomaly Detection
Rongcheng Wu (University of Technology Sydney), Ye Lin (University of Technology Sydney)
Anomaly DetectionConvolutional Neural NetworkRecurrent Neural NetworkDiffusion modelAuto EncoderImage
🎯 What it does: This paper proposes a recursive reconstruction framework, RcAE, for unsupervised industrial defect detection;
RCAFlow: A Workflow-Informed Hierarchical Planning Multi-Agent System for Root Cause Analysis
Yufei Gao (Zhejiang University), Bowei Yang (Zhejiang University)
Explainability and InterpretabilityLarge Language ModelBenchmark
🎯 What it does: Proposed a multi-agent system named RCAFlow, leveraging structured workflow knowledge and hierarchical planning to achieve explainable and efficient diagnosis for complex root cause analysis in cloud microservices.
RCMoE: A Communication-Efficient Random Compression Framework for Resource-Constrained Mixture-of-Experts Training
Donglei Wu (Guangzhou University), Zhihong Tian (Guangzhou University)
OptimizationComputational EfficiencyMixture of ExpertsImageText
🎯 What it does: Propose a communication compression framework named RCMoE, specifically designed for addressing communication bottlenecks in training resource-constrained Mixture-of-Experts (MoE) models, employing random compression techniques to enhance compression rates while reducing error propagation;
RCP-LO: A Relative Coordinate Prediction Framework for Generalizable Deep LiDAR Odometry
Chen Liu (Xiamen University), Cheng Wang (Xiamen University)
Pose EstimationAutonomous DrivingConvolutional Neural NetworkDiffusion modelSimultaneous Localization and MappingPoint Cloud
🎯 What it does: This paper proposes a framework called RCP-LO, which rewrites LiDAR pose estimation as relative coordinate prediction. It generates robust coordinates using diffusion models and solves the pose in one step through differentiable geometric weighted singular value decomposition, achieving end-to-end training and efficient inference.
RCP-Merging: Merging Long Chain-of-Thought Models with Domain-Specific Models by Considering Reasoning Capability as Prior
Junyao Yang (South China University of Technology), Ziqian Zeng (South China University of Technology)
Computational EfficiencyKnowledge DistillationTransformerLarge Language ModelTextTabularBiomedical DataFinance RelatedChain-of-Thought
🎯 What it does: This paper proposes a new model merging framework called RCP-Merging, which treats the parameters of long-chain reasoning models as priors, preserving reasoning capabilities during merging while incorporating domain knowledge sensitivity and reasoning retention metrics to achieve a balance between domain knowledge and reasoning capabilities.
Re-architecting Personalized Federated Learning for Demanding Edge Environments
Quyang Pan (Chinese Academy of Sciences), Jingyuan Wang (Beihang University)
Federated LearningComputational EfficiencyKnowledge DistillationImageAudio
🎯 What it does: Proposes DistilCacheFL, a personalized edge learning architecture that integrates dataset distillation with knowledge caching.
Re-SpS: A Reinforcement Learning Approach to Speculative Sampling
Chenan Wang (William & Mary), Haipeng Chen (William & Mary)
Computational EfficiencyRepresentation LearningLarge Language ModelReinforcement LearningText
🎯 What it does: To address the inference latency issue in large language models, we propose the Re-SpS framework, which dynamically adjusts the hyperparameters of speculative sampling using reinforcement learning.
REACT-LLM: A Benchmark for Evaluating LLM Integration with Causal Features in Clinical Prognostic Tasks
Linna Wang (Sichuan University), Li Lu (Sichuan University)
ClassificationExplainability and InterpretabilityData-Centric LearningTransformerLarge Language ModelPrompt EngineeringBiomedical DataElectronic Health RecordsBenchmarkChain-of-Thought
🎯 What it does: Built a benchmark framework named REACT-LLM to evaluate the synergistic effects of large language models (LLMs) and causal features in clinical prognosis tasks.
ReACT: Reward-informed Autoregressive Decision CAD Transformer
Yijie Ding (Nanyang Technological University), Jianmin Zheng (Nanyang Technological University)
GenerationConvolutional Neural NetworkTransformerReinforcement LearningPoint Cloud
🎯 What it does: Reverse modeling of point clouds to automatically generate CAD modeling sequences.
REACTION: Parameter-Efficient Learning for Recommendation
Song-Li Wu (Tsinghua University), Zhenhua Dong (Huawei Noah's Ark Lab)
Recommendation SystemContrastive LearningTabular
🎯 What it does: This paper proposes a parameter-efficient recommendation learning framework called REACTION, aiming to reduce feature redundancy and structural redundancy in deep recommendation models.
READ: Real-time and Efficient Asynchronous Diffusion for Audio-driven Talking Head Generation
Haotian Wang (University of Science and Technology of China), Qingfeng Liu (iFLYTEK)
GenerationTransformerDiffusion modelFlow-based ModelAuto EncoderVideoMultimodalityAudio
🎯 What it does: This paper proposes the READ framework, achieving real-time audio-driven speaker head video generation, capable of producing smooth, well-aligned animated videos at a 1:1 time ratio.
Real Garment Benchmark (RGBench): A Comprehensive Benchmark for Robotic Garment Manipulation Featuring a High-Fidelity Scalable Simulator
Wenkang Hu (Shanghai Jiao Tong University), Ruigang Yang (Shanghai Jiao Tong University)
Robotic IntelligencePoint CloudMeshBenchmark
🎯 What it does: Proposed the RGBench benchmark, containing over 6,000 clothing models, real-world robot grasping/throwing/folding experimental data, and a unified evaluation framework.
Real Noise Decoupling for Hyperspectral Image Denoising
Yingkai Zhang (Beijing Institute of Technology), Ying Fu (Beijing Institute of Technology)
RestorationConvolutional Neural NetworkTransformerImage
🎯 What it does: Proposes a multi-stage noise decomposition framework for denoising real hyperspectral images.