AAAI 2026 Papers — Page 26
AAAI Conference on Artificial Intelligence · 4149 papers
Multimodal Robust Prompt Distillation for 3D Point Cloud Models
Xiang Gu (Nanjing University of Science and Technology), Shuchao Pang (Nanjing University of Industry Technology)
ClassificationKnowledge DistillationPrompt EngineeringVision Language ModelContrastive LearningMultimodalityPoint Cloud
🎯 What it does: Propose a multi-modal robust prompt distillation framework MRPD, which enhances the adversarial robustness of point cloud models by distilling robust knowledge from image, text, and 3D teacher models into lightweight prompts;
Multimodal Table Understanding with Difficulty-aware Reinforcement Learning
Chaohu Liu (University of Science and Technology of China), Linli Xu (University of Science and Technology of China)
Reinforcement LearningPrompt EngineeringVision Language ModelTabularChain-of-Thought
🎯 What it does: Propose a multimodal table understanding model MM-Table-R1 based on difficulty-aware reinforcement learning, employing task-level and sample-level curriculum learning to enhance the model's perception and reasoning capabilities for complex table structures.
MultiMotion: Multi Subject Video Motion Transfer via Video Diffusion Transformer
Penghui Liu (Beijing University of Technology), Jack Ma (Hong Kong University of Science and Technology)
GenerationTransformerDiffusion modelRectified FlowAuto EncoderVideoBenchmark
🎯 What it does: In the multi-agent video motion transfer task, a unified framework called MultiMotion is proposed, which can achieve precise and controllable multi-object motion transfer within the Diffusion Transformer (DiT).
Multiple Human Motion Understanding
Lei Li (University of Washington), Jenq-Neng Hwang (University of Washington)
Pose EstimationComputational EfficiencyTransformerLarge Language ModelPrompt EngineeringVision-Language-Action ModelVideoTextMultimodalityBenchmark
🎯 What it does: Proposed LLaMMo, an instruction-tuned multimodal framework for multi-person human motion understanding.
Multiple-Interval Coverage for Resource Management of Passive Surveillance Systems
Jan Pikman (Czech Technical University in Prague), Zdeněk Hanzálek (Czech Technical University in Prague)
Optimization
🎯 What it does: This paper addresses the resource management problem in passive surveillance systems. It first proposes and proves that the Left-Right Algorithm (LRA) can fully construct all necessary receiver coverage intervals. Subsequently, it defines the Multi-Interval Coverage (MIC) optimization problem and systematically analyzes the theoretical complexity of its various variants, proving that if the coverage interval is a single interval and satisfies the 'proper set' condition, it can be solved in polynomial time, but becomes NP-hard when the coverage interval contains two intervals.
Multiple-play Stochastic Bandits with Prioritized Arm Capacity Sharing
Hong Xie (University of Science and Technology of China & State Key Laboratory of Cognitive Intelligence), Defu Lian (University of Science and Technology of China & State Key Laboratory of Cognitive Intelligence)
OptimizationReinforcement Learning
🎯 What it does: Propose a multi-player stochastic multi-armed bandit model named MSB-PRS, applicable in scenarios requiring priority resource sharing such as large language models (LLM) and edge computing;
Multiplex Heterogeneous Graph Neural Networks with Euclidean-Riemannian Mutual Space Synergy
Xiang Li (Ocean University of China), Yanwei Yu (Ocean University of China)
ClassificationRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: Proposed MRiemGNN, a multi-modal heterogeneous graph neural network that co-learns in Euclidean and Riemannian spaces, using relation-aware kernelized message passing to capture composite relationships and enhancing representations through bidirectional mutual distillation.
Multiplicative Orthogonal Sequential Editing for Language Models
Hao-Xiang Xu (University of Science and Technology of China), Jia-Chen Gu (University of California, Los Angeles)
Computational EfficiencyRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposed a new knowledge editing framework called MOSE, which uses the multiplication method of orthogonal matrices to perform sequential editing on large language models, thereby maintaining the numerical stability of model parameters.
MultiTab: A Scalable Foundation for Multitask Learning on Tabular Data
Dimitrios Sinodinos (McGill University), Narges Armanfard (McGill University)
TransformerTabularBenchmark
🎯 What it does: Propose MultiTab-Net, a multi-task Transformer specifically designed for large-scale tabular data, and construct MultiTab-Bench as a synthetic data generation tool that allows adjustable task relevance, difficulty, and quantity.
Multitasks-based Deep Evidential Fusion Network for Blind Image Quality Assessment
Yiwei Lou (Peking University), Yu Huang (Peking University)
ClassificationConvolutional Neural NetworkTransformerVision Language ModelImage
🎯 What it does: A multi-task deep evidence fusion network (DEFNet) was constructed, which achieves no-reference image quality assessment by simultaneously optimizing three tasks: BIQA, scene classification, and distortion type classification, and adopting cross-subregion and local-global two-level trustworthy information fusion.
Multivariate Diffusion Transformer with Decoupled Attention for High-Fidelity Mask-Text Collaborative Facial Generation
Yushe Cao (Tsinghua University), Junliang Xing (National University Of Defense Technology)
Image TranslationGenerationTransformerDiffusion modelMultimodality
🎯 What it does: This paper proposes a multimodal diffusion transformer named MDiTFace, achieving high-fidelity collaborative face generation with semantic masks and text through unified tokenization and decoupled attention.
MuSASplat: Efficient Sparse-View 3D Gaussian Splats via Lightweight Multi-Scale Adaptation
Muyu Xu (Nanyang Technological University), Shijian Lu (Nanyang Technological University)
GenerationComputational EfficiencyTransformerSupervised Fine-TuningGaussian SplattingImage
🎯 What it does: Propose MuSASplat, a lightweight, pose-free sparse-view 3D Gaussian splatting framework that significantly reduces training costs;
MUSE: Multi-Scale Dense Self-Distillation for Nucleus Detection and Classification
Zijiang Yang (University of Science and Technology Beijing), Hui Jiang (Alibaba Group)
ClassificationObject DetectionTransformerBiomedical Data
🎯 What it does: This paper proposes a self-supervised learning framework called MUSE for nucleus detection and classification, employing a multi-scale self-distillation strategy and equipped with an expandable ViT encoder-decoder network.
MUSE: Multimodal Uncertainty-Based Self-Driven Evolution for Robust Physiological-Signal–Based Driver Fatigue Detection
Jiaheng Wang (Shanghai Advanced Research Institute), Honglin Hu (Shanghai Advanced Research Institute)
Autonomous DrivingMultimodalityBiomedical Data
🎯 What it does: Propose an unsupervised multi-modal uncertainty-driven self-driven evolution framework called MUSE, which dynamically reallocates voting weights in real-time based on the uncertainty of each modality to achieve adaptive fusion for driver fatigue detection.
MusicRec: Multi-modal Semantic-Enhanced Identifier with Collaborative Signals for Generative Recommendation
Yuqiu Zhao (Communication University of China), Yanchao Liu (North China University of Technology)
Recommendation SystemTransformerLarge Language ModelAuto EncoderContrastive LearningMultimodality
🎯 What it does: Propose MusicRec, design multimodal semantic-enhanced identifiers and fuse collaborative signals, constructing a generative recommendation framework based on large language models.
MUTrack: A Memory-Aware Unified Representation Framework for Visual Tracking
Weijing Wu (Guangxi Normal University), Yuanliang Xue (Xi'an Research Institute of High Technology)
Object TrackingTransformerVideo
🎯 What it does: This paper proposes a unified memory framework called MUTrack for visual object tracking, which combines long-term and short-term memory to generate a unified target representation, and uses the Perception Interaction Module for deep mutual modulation.
MVGD-Net: A Novel Motion-aware Video Glass Surface Detection Method
Yiwei Lu (Jiangnan University), Tao Yan (Jiangnan University)
SegmentationTransformerOptical FlowVideo
🎯 What it does: Proposed a new network called MVGD-Net for glass surface detection using video motion inconsistency.
MvP-ECR: Multi-Perspective Emotion-Cause Reasoning for Empathetic Dialogue
Yuanyuan He (University of Electronic Science and Technology of China), Fuji Ren (University of Electronic Science and Technology of China)
Large Language ModelPrompt EngineeringText
🎯 What it does: Proposed a multi-perspective emotion-causal reasoning framework, MvP-ECR, for explicitly constructing emotion-causal structures in empathetic dialogues.
MyGram: Modality-aware Graph Transformer with Global Distribution for Multi-modal Entity Alignment
Zhifei Li (Hubei University), Bing Yang (Hubei University)
Representation LearningGraph Neural NetworkTransformerContrastive LearningMultimodality
🎯 What it does: This study proposes the MyGram framework for multimodal entity alignment.
NADIR: Differential Attention Flow for Non-Autoregressive Transliteration in Indic Languages
Lakshya Tomar (RocketFrog AI), Puneet Agarwal (RocketFrog AI)
GenerationTransformerMixture of ExpertsText
🎯 What it does: Proposed the NADIR model for non-autoregressive transcription of Indic scripts in multiple languages, addressing the high latency issue of traditional autoregressive (AR) models.
Navigating the Alpha Jungle: An LLM-Powered MCTS Framework for Formulaic Alpha Factor Mining
Yu Shi (Tsinghua University), Jian Li (Tsinghua University)
TransformerLarge Language ModelTime SeriesFinance Related
🎯 What it does: Built a formula-based alpha mining framework combining large language models and Monte Carlo Tree Search (MCTS), which can iteratively generate and optimize alpha formulas under the drive of financial backtesting feedback.
Navigating Through Paper Flood: Advancing LLM-Based Paper Evaluation Through Domain-Aware Retrieval and Latent Reasoning
Wuqiang Zheng (University Of Science And Technology Of China), Fuli Feng (University Of Science And Technology Of China)
Recommendation SystemTransformerLarge Language ModelTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Proposed and implemented the PaperEval framework, which utilizes large language models (LLMs) for automatic paper evaluation, achieving more accurate assessments through three modules: domain-aware retrieval, latent reasoning, and progressive ranking optimization.
NaVLA$^2$: A Vision-Language-Audio-Action Model for Multimodal Instruction Navigation
Jugang Fan (South China University of Technology), Mingkui Tan (South China University of Technology)
Explainability and InterpretabilityRobotic IntelligenceTransformerLarge Language ModelSupervised Fine-TuningVision-Language-Action ModelImageTextMultimodalityChain-of-ThoughtAudio
🎯 What it does: Designed a multi-modal instruction navigation task, MINav, and constructed a dataset containing 43.9K navigation samples with visual, linguistic, image, and audio prompts; proposed the NaVLA 2 model, integrating spatial semantic audio encoding and CoThinkAct reasoning decoding to achieve interpretable multi-step action planning.
Near-optimal Linear Predictive Clustering in Non-separable Spaces via MIP and QPBO Reductions
Jiazhou Liang (University of Toronto), Scott Sanner (University of Toronto)
OptimizationComputational EfficiencyTabular
🎯 What it does: Propose an approximate global optimal solution framework for linear prediction clustering (LPC) in non-separable feature spaces, containing two improved MIP and QPBO methods.
Negative Entity Suppression for Zero-Shot Captioning with Synthetic Images
Zimao Lu (China University of Mining and Technology), Ke Wang (China University of Mining and Technology)
GenerationData SynthesisRetrievalDomain AdaptationTransformerLarge Language ModelDiffusion modelContrastive LearningImageTextMultimodalityRetrieval-Augmented Generation
🎯 What it does: This paper proposes a Negative Entity Suppression (NES) framework based on synthetic images, leveraging synthetic image retrieval, negative entity filtering, and attention suppression to significantly enhance cross-domain generalization ability in zero-shot image captioning and reduce hallucinations.
Neighbor-aware Instance Refining with Noisy Labels for Cross-Modal Retrieval
Yizhi Liu (Sichuan University), Yuan Sun (Sichuan University)
RetrievalData-Centric LearningMultimodalityBenchmark
🎯 What it does: Propose a robust cross-modal retrieval framework named NIRNL, integrating two modules: cross-modal margin preservation (CMP) and neighborhood-aware instance refinement (NIR). By leveraging neighborhood consistency, the training samples are finely divided into clean, easy-hard, and noisy subsets, with dedicated losses designed for each subset to enhance cross-modal retrieval performance under noisy label environments.
Neighbor-aware Label Refinement: Enhancing Unreliable Instance-Dependent Partial Labels
Xijia Tang (National University of Defense Technology), Chenping Hou (National University of Defense Technology)
ClassificationImage
🎯 What it does: Proposed the UIDPLL (Unreliable Instance Related Partial Labels) problem and designed the NLAP method to incrementally expand and prune candidate labels.
Nested Depth Search
Junkang Li (NukkAI), Veronique Ventos
OptimizationBenchmark
🎯 What it does: This paper proposes an algorithm called Nested Depth Search (NDS), a generalized extension of NMCS, which can explore the search space more extensively in higher-level simulations by setting depth d and step size s; it also provides time complexity analysis of NDS and an exact probability distribution calculation method for the Left Move Problem (LMP).
Nested Graph Pseudo-Label Refinement for Noisy Label Domain Adaptation Learning
Yingxu Wang, Nan Yin (JD Industrial, Inc)
Domain AdaptationGraph Neural NetworkGraph
🎯 What it does: This paper studies the problem of graph layer classification in Graph Domain Adaptation under severe label noise in the source domain, and proposes a dual-branch framework named NeGPR, which achieves robust cross-domain transfer through dual pre-training, nested pseudo-label refinement, and noise-tolerant regularization.
NeSTR: A Neuro-Symbolic Abductive Framework for Temporal Reasoning in Large Language Models
Feng Liang (China Academy of Launch Vehicle Technology), Xiang Zhao (National University of Defense Technology)
Explainability and InterpretabilityLarge Language ModelTextBenchmark
🎯 What it does: Propose NeSTR, a neuro-symbolic temporal reasoning framework that combines structured symbolic representations with reflective reasoning to enhance the temporal reasoning capabilities of large language models.
Neural Architecture and Hyperparameter Selection Through Meta-Learning on Time Series
Erfan Moeini (RWTH Aachen University), Holger H. Hoos (RWTH Aachen University)
ClassificationOptimizationHyperparameter SearchNeural Architecture SearchTime Series
🎯 What it does: Propose a meta-learning based framework that predicts the performance of neural network architectures and their hyperparameters in time series classification and forecasting tasks using historical search results, thereby recommending optimal configurations.
Neural Architecture for Fast and Reliable Coagulation Assessment in Clinical Settings: Leveraging Thromboelastography
Yulu Wang (Zhejiang University), Zhifeng Tang (Huzhou Institute of Zhejiang University)
Convolutional Neural NetworkRecurrent Neural NetworkTransformerBiomedical Data
🎯 What it does: Proposed the PSR framework, integrating multi-domain feature extraction, hierarchical learning, and dynamic adaptation to achieve real-time prediction of hemostatic dynamics under extremely small sample conditions;
Neural Collapse Priors Driven Trust Semi-Supervised Multi-View Classification
Taotao Guo (Southwest University of Science and Technology), Xingfeng Li (Southwest University of Science and Technology)
ClassificationContrastive LearningImageTextMultimodality
🎯 What it does: In semi-supervised multi-view classification, we propose to utilize class prototypes generated by neural network convergence (Neural Collapse) as priors to guide feature learning for unlabeled samples, and combine attention mechanisms, cross-view contrastive learning, and evidence-based reliable fusion to achieve fine-grained calibration of class distributions and uncertainty modeling.
Neural Collapse-Informed Initialization with Perturbation Injection in Classification-based Metric Learning
Jinhee Park (Chung Ang University), Junseok Kwon (Chung Ang University)
ClassificationRetrievalRepresentation LearningConvolutional Neural NetworkImage
🎯 What it does: In the fine-grained image retrieval task, the authors propose initializing the classification head using the neural collapse (NC) direction of a pre-trained model and injecting small Gaussian noise during fine-tuning to maintain the NC structure and enhance retrieval performance.
Neural Graph Navigation for Intelligent Subgraph Matching
Yuchen Ying (Zhejiang University), Mingli Song (Zhejiang University)
Graph Neural NetworkTransformerGraph
🎯 What it does: Proposed Neural Graph Navigation (NeuGN), introducing neural networks into the subgraph matching enumeration phase to achieve structure-aware search navigation.
Neural Outline Cache for Real-time Anti-aliasing Font Rendering
Jiashuaizi Mo (Zhejiang Normal University), Zhonglong Zheng (Zhejiang Normal University)
OptimizationComputational EfficiencyGaussian Splatting
🎯 What it does: Propose Neural Outline Cache (NOC), a lightweight neural font texture that enables real-time anti-aliasing rendering and interactive procedural editing within modern graphics pipelines.
Neural Tangent Kernels Under Stochastic Data Augmentation
Joshua DeOliveira (Worcester Polytechnic Institute), Elke Rundensteiner (Worcester Polytechnic Institute)
OptimizationRepresentation LearningConvolutional Neural NetworkImageStochastic Differential Equation
🎯 What it does: Studied the training dynamics of infinitely wide neural networks under random data augmentation, and proposed a stochastic differential equation model based on the neural tangent kernel (NTK)
Neural Video Compression with Reference Hierarchy
Chuanbo Tang (University of Science and Technology of China), Feng Wu (University of Science and Technology of China)
CompressionMixture of ExpertsOptical FlowVideo
🎯 What it does: This paper proposes a unified reference hierarchy (DRHVC), achieving more efficient temporal prediction by introducing multi-frame feature-level reference management within a neural video compression framework.
Neural-Augmented Kelvinlet for Real-Time Soft Tissue Deformation Modeling
Ashkan Shahbazi (Vanderbilt University), Soheil Kolouri (Vanderbilt University)
Computational EfficiencyGraph Neural NetworkBiomedical DataPhysics Related
🎯 What it does: Developed a real-time soft tissue deformation prediction framework that combines Kelvinlet analytical solutions with neural networks, enhancing model accuracy through residual learning and regularization.
NeuralGS: Bridging Neural Fields and 3D Gaussian Splatting for Compact 3D Representations
Zhenyu Tang (Peking University), Li Yuan (Hong Kong University of Science and Technology)
CompressionGaussian Splatting
🎯 What it does: Propose NeuralGS, a post-training compression method for 3D Gaussian Splatting, which re-encodes high-dimensional Gaussian attributes using multiple small MLPs after clustering, significantly reducing model size.
NeuralOM: Neural Ocean Model for Subseasonal-to-Seasonal Simulation
Yuan Gao (Tsinghua University), Xiaomeng Huang (Tsinghua University)
Graph Neural NetworkTime SeriesPhysics Related
🎯 What it does: Proposed a neural operator framework named NEURALOM for high-accuracy, long-term ocean (Subseasonal-to-Seasonal) simulation and prediction.
Neuro-Symbolic Federated Learning over Heterogeneous Data-Views: A Structured Approach to Distributive EHR Modelling
Soheila Molaei (University of Oxford), David A. Clifton (GlaxoSmithKline)
Federated LearningGraph Neural NetworkMultimodalityTime SeriesElectronic Health Records
🎯 What it does: Under the federated learning framework, distributed electronic health records (EHR) are modeled as typed knowledge graphs to enable collaborative training across institutions and data views.
NeuroBridge: Bio-Inspired Self-Supervised EEG-to-Image Decoding via Cognitive Priors and Bidirectional Semantic Alignment
Wenjiang Zhang (Beijing University of Posts and Telecommunications), Suyu Zhong (Beijing University of Posts and Telecommunications)
RetrievalRepresentation LearningConvolutional Neural NetworkTransformerContrastive LearningImageBiomedical Data
🎯 What it does: Propose the NeuroBridge framework, which leverages self-supervised learning combined with cognitive prior enhancement and shared semantic projection to achieve cross-modal alignment between EEG and images for zero-shot visual decoding.
NeuS-QA: Grounding Long-Form Video Understanding in Temporal Logic and Neuro-Symbolic Reasoning
Sahil Shah (University of Texas at Austin), Sandeep Chinchali (Independent Researcher)
Computational EfficiencyTransformerLarge Language ModelVision Language ModelVideoBenchmark
🎯 What it does: Propose NeuS-QA, a training-agnostic, plug-and-play neuro-symbolic framework for long video question answering (LVQA);
NeuSpring: Neural Spring Fields for Reconstruction and Simulation of Deformable Objects from Videos
Qingshan Xu (Nanyang Technological University), Hanwang Zhang (Nanyang Technological University)
GenerationOptimizationRepresentation LearningGaussian SplattingVideoPhysics Related
🎯 What it does: Reconstruct and simulate the geometry, appearance, and physical properties of deformable objects using video data to build a physical digital twin.
New Synthetic Goldmine: Hand Joint Angle-Driven EMG Data Generation Framework for Micro-Gesture Recognition
Nana Wang (BeiHang University), Hao Su (BeiHang University)
RecognitionGenerationData SynthesisConvolutional Neural NetworkRecurrent Neural NetworkGenerative Adversarial NetworkTime SeriesSequentialBiomedical Data
🎯 What it does: Built a conditional generation framework based on SeqEMG-GAN, which uses hand joint angle sequences to drive the generation of high-fidelity, temporally consistent EMG signals;
Next Generation Active Learning: Mixture of LLMs in the Loop
Yuanyuan Qi (Monash University), Lan Du (Monash University)
ClassificationTransformerLarge Language ModelText
🎯 What it does: Designed a fully manual-annotation-free active learning framework called MoLLIA, which generates high-quality labels by aggregating outputs from multiple lightweight LLMs, and enhances model robustness through negative learning and annotation discrepancy mechanisms;
Next Patch Prediction for AutoRegressive Visual Generation
Yatian Pang (Peking University), Li Yuan (Peking University)
GenerationTransformerAuto EncoderImage
🎯 What it does: Proposed a novel 'Next Patch Prediction' (NPP) method that enhances the quality and efficiency of autoregressive visual generation by aggregating image tokens into high-information-density patches and predicting the next patch.
NGTM: Substructure-based Neural Graph Topic Model for Interpretable Graph Generation
Yuanxin Zhuang (Hong Kong University of Science and Technology), Ying Sun (Hong Kong University of Science and Technology)
GenerationExplainability and InterpretabilityGraph Neural NetworkAuto EncoderGraph
🎯 What it does: Proposed an interpretable graph generation framework NGTM based on topic modeling, generating graph structures through substructure topicization, balancing the interpretability and controllability of local substructures and global themes.
NICE: Neural Implicit Craniofacial Model for Orthognathic Surgery Prediction
Jiawen Yang (ShanghaiTech University), Hongjiang Wei (ShanghaiTech University)
Point CloudBiomedical DataComputed Tomography
🎯 What it does: Proposed a craniofacial model called NICE based on neural implicit functions for predicting facial appearance after orthognathic surgery.
Nighttime Flare Removal via Wavelet-Guided and Gated-Enhanced Spatial-Frequency Fusion Network
Yun Liu (Southwest University), Weisi Lin (Southwest University)
RestorationConvolutional Neural NetworkImage
🎯 What it does: Propose a spatial-frequency fusion network (WGSF-Net) based on multi-level wavelet enhancement and gated attention, specifically designed for nighttime lens flare removal.
NL2CA: Auto-formalizing Cognitive Decision-Making from Natural Language Using an Unsupervised CriticNL2LTL Framework
Zihao Deng (Institute of Automation, Chinese Academy of Sciences), Peijun Ye (Institute of Automation, Chinese Academy of Sciences)
Autonomous DrivingExplainability and InterpretabilityReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextSequentialBenchmark
🎯 What it does: This paper proposes a fully automated framework called NL2CA, which can directly convert natural language descriptions of human experience into production rules for cognitive decision-making, and trains cognitive agents that can imitate human behavior through cognitive reinforcement learning.
No-Regret Strategy Solving in Imperfect-Information Games via Pre-Trained Embedding
Yanchang Fu (University of Chinese Academy of Sciences), Kaiqi Huang (Chinese Academy of Sciences)
OptimizationRepresentation LearningReinforcement LearningSequential
🎯 What it does: Propose the Embedding CFR algorithm, which utilizes a pre-trained low-dimensional embedding space to solve strategies in incomplete information games (e.g., poker).
NODiff: Neural Operator Diffusion for Multispectral Image Fusion
Junming Hou (Southeast University), Liang-Jian Deng (University Of Electronic Science And Technology Of China)
RestorationDiffusion modelImage
🎯 What it does: Proposed a diffusion model called NODiff based on neural operators (FNO) to efficiently perform multispectral image fusion (pansharpening), and achieved parameter-efficient fine-tuning through two-stage pre-training combined with a lightweight adapter.
Noise-Aware Graph-Based Cognitive Diagnostic Framework Through Low-Rank Alignment
Guixian Zhang, Debo Cheng (China University of Mining and Technology)
ClassificationGraph Neural NetworkTabular
🎯 What it does: Research and propose a graph neural network framework based on low-rank reconstruction and self-supervised alignment to enhance the robustness of cognitive diagnostic models in noisy environments.
Noisy Correspondence Learning with Modality Gap Direction Correction
Wuyuqing Wang (Xidian University), Erkun Yang (Xidian University)
RetrievalConvolutional Neural NetworkRecurrent Neural NetworkVision Language ModelContrastive LearningMultimodality
🎯 What it does: Propose a robust learning framework named MGCS for cross-modal retrieval, which corrects cross-modal similarity by leveraging sample-level alignment drift (SAD), and adaptively segments noisy samples through dynamic regularization.
Non-Monotonic S4F Standpoint Logic
Piotr Gorczyca (TUD Dresden University of Technology), Hannes Strass (TUD Dresden University of Technology)
🎯 What it does: Proposed a new framework combining non-monotonic S4F logic with stance logic, termed S4F Stance Logic, and provided its semantics, minimal model determination, and complexity analysis;
Non-Monotonicity in Fair Division of Graphs
Hadi Hosseini (Penn State University), Yu Zhou (Beijing Normal University)
OptimizationGraph
🎯 What it does: This paper studies the allocation of vertices (i.e., items) in a graph, using the 'cut value' as a non-monotonic bundle-dependent value function, and investigates the compatibility and existence between fairness (EF1) and efficiency (such as TS, WTS, SO, PO).
NoReGeo: Non-Reasoning Geometry Benchmark
Irina Abdullaeva (FusionBrain Lab), Andrey Kuznetsov (FusionBrain Lab)
Large Language ModelPrompt EngineeringVision Language ModelContrastive LearningImageTextGraphBenchmark
🎯 What it does: Proposes the NoReGeo benchmark to evaluate whether large language models and vision-language models possess innate geometric intuition without relying on reasoning or algebraic operations;
Not All Distortions Are Created Equal: Distortion-Selective Domain Adaptation for Point Cloud Quality Assessment
Yangwei Li (Beijing Technology and Business University), Haisheng Li (Beijing Technology and Business University)
Domain AdaptationConvolutional Neural NetworkTransformerPoint Cloud
🎯 What it does: This paper proposes a point cloud quality assessment framework DST-PCQA based on distortion selection, which first trains a full distortion classifier to identify target domain-related distortions, and then uses the selected distortion subset to train a dual-branch (2D vision + 3D geometry) quality regression model, significantly improving cross-domain generalization performance.
Not All Inconsistency Is Equal: Decomposing LVLM Uncertainty into Belief Divergence and Belief Conflict
Jie Shi, Feifan Dong (Tongji University)
Explainability and InterpretabilityGraph Neural NetworkVision Language ModelMultimodalityBenchmark
🎯 What it does: Proposed a black-box LVLM uncertainty quantification framework based on Dempster-Shafer evidence theory, decomposing inconsistency into two metrics: belief divergence and belief conflict.
Not All Tokens and Heads Are Equally Important: Dual-Level Attention Intervention for Hallucination Mitigation
Lexiang Tang (Peking University), Yuexian Zou (Peking University)
Explainability and InterpretabilityComputational EfficiencyRepresentation LearningTransformerLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: Propose the VisFlow framework, which directly adjusts the attention distribution of the LVLM decoder during inference through dual-level attention intervention (Token-level TAI and Head-level HAI) to reduce visual hallucinations.
Not Everything Is Permitted: Constrained Cartesian Abstractions for Optimal Classical Planning
Martín Pozo (Universidad Carlos III de Madrid), Carlos Linares López (Universidad Carlos III de Madrid)
OptimizationBenchmark
🎯 What it does: This paper proposes a Cartesian abstraction method constrained by state constraints (such as mutual exclusion pairs and dead pairs), and embeds it into the CEGAR process to generate more informative heuristics;
Not Just for Archiving: Provable Benefits of Reusing the Archive in Evolutionary Multi-objective Optimization
Shengjie Ren (Nanjing University), Chao Qian (University of Birmingham)
OptimizationBenchmark
🎯 What it does: This paper investigates reusing the archive in multi-objective evolutionary algorithms (MOEA) to accelerate the search process, demonstrating that it provides polynomial speedup for SMS-EMOA on OneJumpZeroJump and its variants, and verifying its superiority through experiments.
Not Just What’s There: Enabling CLIP to Comprehend Negated Visual Descriptions Without Fine-Tuning
Junhao Xiao (Central China Normal University), Zejiang He (National University Of Defense Technology)
TransformerVision Language ModelContrastive LearningMultimodality
🎯 What it does: Propose the CLIPGLASSES framework, enabling CLIP to understand visual descriptions with negations without modifying the original parameters.
NOTAM-Evolve: A Knowledge-Guided Self-Evolving Optimization Framework with LLMs for NOTAM Interpretation
Maoqi Liu, Kaiquan Cai (Beihang University)
OptimizationTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextRetrieval-Augmented Generation
🎯 What it does: Proposed and implemented a self-evolving framework called NOTAM-Evolve for deeply parsing highly concise, industry-specific NOTAM text into structured information.
Note2Chat: Improving LLMs for Multi-Turn Clinical History Taking Using Medical Notes
Yang Zhou (Institute of High Performance Computing, Agency for Science, Technology and Research), Yong Liu (Institute of High Performance Computing, Agency for Science, Technology and Research)
TransformerLarge Language ModelSupervised Fine-TuningBiomedical DataElectronic Health Records
🎯 What it does: Proposed the Note2Chat framework based on medical notes, utilizing notes to drive LLMs for multi-round medical history collection and differential diagnosis.
NP-MiSR: Neural Process-based Multi-Interest Learning for Session-Based Recommendation
Jun Bao (Jilin University), Yuanbo Xu (Jilin University)
Recommendation SystemGraph Neural NetworkSequential
🎯 What it does: Designed and implemented a multi-interest learning framework based on neural processes, NP-MiSR, for session recommendation, which can adaptively learn multi-interest representations of sessions and integrate information from similar sessions.
NTSFormer: A Self-Teaching Graph Transformer for Multimodal Isolated Cold-Start Node Classification
Jun Hu (National University of Singapore), Bingsheng He (National University of Singapore)
ClassificationGraph Neural NetworkTransformerMixture of ExpertsMultimodalityGraph
🎯 What it does: This paper proposes a self-teaching framework called NTSFormer based on graph Transformer, aimed at solving the cold start node classification problem in multimodal graphs, capable of simultaneously handling node isolation and modality missing;
NucEL: Single-Nucleotide ELECTRA-Style Genomic Pre-training for Efficient and Interpretable Representations
Ke Ding (Australian National University), Jiayu Wen (Australian National University)
Computational EfficiencyRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningBiomedical Data
🎯 What it does: Proposed and implemented NucEL, a genome pre-training framework based on ELECTRA, which employs single-nucleotide tokenization, is pre-trained specifically on the human genome, and is fine-tuned on multiple genomic functional prediction tasks.
NumCoKE: Ordinal-Aware Numerical Reasoning over Knowledge Graphs with Mixture-of-Experts and Contrastive Learning
Ming Yin (Chinese Academy of Sciences), Neng Gao (Chinese Academy of Sciences)
Representation LearningMixture of ExpertsContrastive LearningGraph
🎯 What it does: Propose a numerical reasoning model for knowledge graphs called NumCoKE, which utilizes a Mixture-of-Experts (MoE) knowledge-aware encoder and ordinal contrastive learning to achieve joint representation of entities, relations, and numerical attributes, as well as fine-grained ordinal discrimination.
NURBGen: High-Fidelity Text-to-CAD Generation Through LLM-Driven NURBS Modeling
Muhammad Usama (German Research Center for Artificial Intelligence), Muhammad Zeshan Afzal (German Research Center for Artificial Intelligence)
GenerationTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringVision Language ModelTextMesh
🎯 What it does: Propose the NURBGen framework, which directly translates natural language prompts into editable NURBS parameters via large language models and generates high-precision 3D CAD models.
O-DisCo-Edit: Object Distortion Control for Unified Realistic Video Editing
Yuqing Chen (Tsinghua University), Yujiu Yang (Huawei Inc)
GenerationDiffusion modelVideo
🎯 What it does: Propose a unified video editing framework O-DisCo-Edit, which performs multi-task video editing using object distortion control based on noise (O-DisCo).
O3SLM: Open Weight, Open Data, and Open Vocabulary Sketch-Language Model
Rishi Gupta (Indian Institute of Science), Anirban Chakraborty (Indian Institute of Science)
RecognitionRetrievalTransformerLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: Proposed a new large visual language model called O3SLM, specialized in understanding and reasoning about hand-drawn sketches, and constructed a large-scale sketch-image-instruction triplet dataset named SketchVCL.
OAD-Promoter: Enhancing Zero-Shot VQA Using Large Language Models with Object Attribute Description
Quanxing Xu, Jinyu Tian
Domain AdaptationTransformerLarge Language ModelPrompt EngineeringVision Language ModelMultimodalityRetrieval-Augmented Generation
🎯 What it does: Propose OAD-Promoter, which enhances the accuracy and domain transferability of LLMs in zero-shot/few-shot knowledge-driven visual question answering (VQA) through multi-granularity visual description generation and memory knowledge assistance.
Object Fusion via Diffusion Time-step for Customized Image Editing with Single Example
Xue Song (Fudan University), Jingjing Chen (Fudan University)
Image HarmonizationGenerationSupervised Fine-TuningDiffusion modelImage
🎯 What it does: Propose a LoRA fusion method called TimeFusion based on time steps and spatial positions, enabling customized editing of single images. It can accomplish tasks such as adding and replacing while preserving the fidelity of the source image and reference objects.
Object-Centric Framework for Video Moment Retrieval
Zongyao Li (NEC Corporation), Mohan Kankanhalli (National University of Singapore)
Object TrackingRetrievalTransformerVideo
🎯 What it does: Propose an object-based framework for modeling object-level feature sequences in video moment retrieval.
Object-Centric Latent Action Learning
Albina Klepach (dunnolab.ai), Vladislav Kurenkov (dunnolab.ai)
Robotic IntelligenceReinforcement LearningVision-Language-Action ModelAuto EncoderVideo
🎯 What it does: Proposes an object-centric latent action learning framework that leverages object slot decomposition from unlabeled videos to learn agents' latent actions;
Object-Centric World Models for Causality-Aware Reinforcement Learning
Yosuke Nishimoto (University of Osaka), Takashi Matsubara (University of Osaka)
TransformerReinforcement LearningWorld ModelImageBenchmark
🎯 What it does: Proposes a reinforcement learning framework called STICA, which combines an object-centric Transformer world model with causal attention mechanisms, achieving end-to-end learning of object representations from pixels and policy training in imagined environments.
ObjectAdv: Object-Level Unrestricted Adversarial Attacks via Diffusion Models
Shijie Zhao (Southwest University of Science and Technology), Hui Zeng (Southwest University of Science and Technology)
Adversarial AttackPrompt EngineeringDiffusion modelImage
🎯 What it does: Proposed ObjectAdv, an object-level unrestricted adversarial attack framework based on diffusion models, which leverages Stable Diffusion's cross-attention to locate attack regions and achieves efficient localized attacks through prompt switching and FFT edge smoothing, generating high-quality UAEs with high attack success rates.
ObjecTok: Learning Holistic and Robust Object Tokens for MLLMs
Sihan Wang (State Key Laboratory of Robotics and Intelligent Systems, Shenyang Institute of Automation, Chinese Academy of Sciences), Zhi Han (State Key Laboratory of Robotics and Intelligent Systems, Shenyang Institute of Automation, Chinese Academy of Sciences)
Object DetectionSegmentationRepresentation LearningTransformerSupervised Fine-TuningVision Language ModelAuto EncoderContrastive LearningImageTextMultimodality
🎯 What it does: Proposes the ObjecTok framework, which encodes each object in an image into a single complete object token and combines it with traditional patch tokens as input to a multimodal large language model, achieving enhanced visual perception capabilities.
Oblivionis: A Lightweight Learning and Unlearning Framework for Federated Large Language Models
Fuyao Zhang (Nanyang Technological University), Qiang Yang (Hong Kong Polytechnic University)
Federated LearningSafty and PrivacyTransformerLarge Language ModelTextBenchmark
🎯 What it does: Developed a lightweight federated learning and machine learning forgetting framework named Oblivionis for training and targeted forgetting in federated LLMs.
OccamVTS: Distilling Vision Models to 1% Parameters for Time Series Forecasting
Sisuo Lyu (Hong Kong University of Science and Technology Guangzhou), Yuxuan Liang (Hong Kong University of Science and Technology Guangzhou)
Knowledge DistillationRepresentation LearningTransformerVision Language ModelAuto EncoderTime Series
🎯 What it does: Transfer 1% of the parameters from a large visual model to a lightweight network through knowledge distillation, achieving efficient time series forecasting.
OceanSplat: Object-aware Gaussian Splatting with Trinocular View Consistency for Underwater Scene Reconstruction
Minseong Kweon (University of Minnesota), Jinsun Park (Pusan National University)
RestorationDepth EstimationNeural Radiance FieldGaussian SplattingImage
🎯 What it does: A 3D Gaussian Splatting method utilizing tri-view (lateral, longitudinal translational virtual cameras) consistency constraints for high-fidelity reconstruction of underwater scenes and image restoration under scattering media.
ODYSSEY: Open-World Quadrupeds Exploration and Manipulation for Long-Horizon Tasks
Kaijun Wang (Zhejiang University), Chunhua Shen (Zhejiang University)
Robotic IntelligenceTransformerLarge Language ModelReinforcement LearningVision Language ModelSimultaneous Localization and MappingMultimodalityBenchmark
🎯 What it does: Proposed the ODYSSEY framework, achieving open-world long-horizon task planning and whole-body control for quadrupedal mobile robots equipped with robotic arms.
Offline Fictitious Self-Play for Competitive Games
Jingxiao Chen (Shanghai Jiao Tong University), Ying Wen (Shanghai Jiao Tong University)
Reinforcement LearningSequential
🎯 What it does: Proposes the Offline Self-Play (OFF-SP) framework and the Offline Fictitious Self-Play (OFF-FSP) algorithm, which use fixed offline data to find approximate Nash equilibria in competitive (zero-sum) games.
Offline Meta-Reinforcement Learning with Flow-Based Task Inference and Adaptive Correction of Feature Overgeneralization
Min Wang (Beijing Institute of Technology), Hasnaa Bennis (University of the Sunshine Coast)
Meta LearningReinforcement LearningFlow-based ModelSequential
🎯 What it does: Proposed the FLORA algorithm, combining flow-based task inference with an adaptive correction mechanism for feature overgeneralization, addressing task distribution inference and extrapolation errors in offline meta reinforcement learning.
Offline Multi-Objective Bandits: From Logged Data to Pareto-Optimal Policies
Ji Cheng (City University of Hong Kong), Bo Xue (City University of Hong Kong)
OptimizationReinforcement Learning
🎯 What it does: Proposed an offline multi-objective contextual bandit algorithm called OffMOB, aiming to learn Pareto optimal strategies and generate the complete Pareto frontier from static log data.
OFL-SAM2: Prompt SAM2 with Online Few-shot Learner for Efficient Medical Image Segmentation
Meng Lan (Hong Kong University of Science and Technology), Xiaomeng Li (Hong Kong University of Science and Technology)
SegmentationMeta LearningTransformerPrompt EngineeringBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: Proposed the OFL-SAM2 framework, leveraging online few-shot learning and adaptive fusion to achieve zero-shot medical image segmentation
Oligodendrocyte-Driven Spiking Neural Model
Mengqiao Han (Northwest Agricultural and Forestry University), Hongming Zhang (Northwest Agricultural and Forestry University)
ClassificationSpiking Neural NetworkTransformerImage
🎯 What it does: Proposed an oligodendrocyte-inspired spiking neuron model and constructed the corresponding Oli-SNN network for efficient classification tasks.
Omni-Effects: Unified and Spatially-Controllable Visual Effects Generation
Fangyuan Mao (Alibaba Group), Xiangxiang Chu (Alibaba Group)
GenerationTransformerPrompt EngineeringMixture of ExpertsDiffusion modelVideo
🎯 What it does: Designed the Omni-Effects unified framework to achieve single-effect, multi-effect, and spatially controllable combination VFX video generation.
OmniBench: A Comprehensive Benchmark Integrating Real-World, Time-sensitive, and Multi-Hop Questions with a Multi-Dimensional Hybrid Evaluation Framework
Wenjie Wang (Ant Group), Bin Chen (Ant Group)
Large Language ModelTextBenchmark
🎯 What it does: Proposed OmniBench, an open-ended dialogue question-answering benchmark based on real user questions, time-sensitive queries, and multi-hop questions.
OmniDPO: A Preference Optimization Framework to Address Omni-Modal Hallucination
Junzhe Chen (Tsinghua University), Lijie Wen (Hong Kong University of Science and Technology)
OptimizationReinforcement Learning from Human FeedbackLarge Language ModelMultimodality
🎯 What it does: Designed and implemented the OMNIDPO framework, leveraging multimodal preference learning (text, video, audio) to reduce multimodal hallucinations and enhance reasoning capabilities.
OmniEvent: Unified Event Representation Learning
Weiqi Yan (Xiamen University), Yu Zang (Jimei University)
ClassificationRepresentation LearningTransformerPoint Cloud
🎯 What it does: Proposes a unified event camera data representation learning framework called OmniEvent, which can be directly applied to multiple tasks such as classification, optical flow estimation, and point cloud registration.
OmniPT: Unleashing the Potential of Large Vision Language Models for Pedestrian Tracking and Understanding
Teng Fu (Fudan University), Bin Li (Fudan University)
Object TrackingSupervised Fine-TuningReinforcement LearningPrompt EngineeringVision Language ModelContrastive LearningVideoText
🎯 What it does: Propose the OmniPT framework, unifying large vision-language models (LVLM) for multi-object tracking, reference tracking, cross-perspective reference tracking, and video/instance semantic understanding, achieving interactive tracking based on instructions.
OmniScale: Scaling Any Modality Model Training with Model-Centric Distributed Recipe Zoo
Qianli Ma (ByteDance Seed), Xin Liu (ByteDance Seed)
Computational EfficiencyLarge Language ModelMixture of ExpertsImageVideoTextMultimodalityAudio
🎯 What it does: Proposed the OmniScale framework for efficient training of large language models (LLMs) across any modality, supporting modular configuration and n-dimensional parallelism.
OmniSparse: Training-Aware Fine-Grained Sparse Attention for Long-Video MLLMs
Feng Chen (University of Adelaide), Qi Wu (University of Adelaide)
Computational EfficiencyTransformerLarge Language ModelVision Language ModelVideoText
🎯 What it does: Propose OmniSparse, a fine-grained sparse attention framework that can be used during training, achieving sparsity across three dimensions: queries, key-value pairs, and attention heads;
OmniVDiff: Omni Controllable Video Diffusion for Generation and Understanding
Dianbing Xi (Zhejiang University), Xuelong Li (China Telecom)
GenerationTransformerDiffusion modelAuto EncoderVideoTextMultimodality
🎯 What it does: Propose OmniVDiff, a unified video diffusion framework capable of performing text-to-video generation, controllable generation based on multimodal inputs (depth, segmentation, edges, etc.), and video multimodal understanding within the same model.
On Condorcet’s Jury Theorem with Abstention
Reshef Meir (Technion-Israel Institute of Technology), Ganesh Ghalme (Indian Institute of Technology Hyderabad)
🎯 What it does: This paper studies the conditions for the Condorcet Jury Theorem (CJT) to hold when abstentions are present, under scenarios where voting costs are heterogeneous and voters have imprecise estimates of pivotality. It proposes a 'Perceivable Pivotality Model' (PPM) and analyzes voting outcomes under different strategic equilibria.
On Coresets for End-to-end Learning from Crowds
Hang Yang (Macau University of Science and Technology), Witold Pedrycz (University of Alberta)
OptimizationData-Centric LearningContrastive LearningImageTabularAudio
🎯 What it does: Investigated the feasibility of using coreset in end-to-end learning with crowdsourced data, and proposed the CrowdCore algorithm with theoretical analysis based on volume regularization under the CCEM (Coupled Cross-Entropy Minimization) framework.