AAAI 2025 Papers with AI Summaries
AAAI Conference on Artificial Intelligence · 3028 papers
→ AAAI 2025 papers with code (1442)
Each paper below shows an AI-generated one-line summary. Get the full 6-part summary (innovation, method, data, results, limitations) and search all 3028 AAAI 2025 papers by keyword, author or institution —
free trial on arXivSub.
(Almost Full) EFX for Three (and More) Types of Agents
Pratik Ghosal (Indian Institute of Technology), Nithin Varma (University of Cologne)
🎯 What it does: The study shows that with n agents having k different valuation functions, there exists an EFX allocation with at least k-2 unallocated items; it is also proven that if there are only two distinct agents, a complete EFX allocation exists.
3CAD: A Large-Scale Real-World 3C Product Dataset for Unsupervised Anomaly Detection
Enquan Yang (Shanghai University), Dan Zeng (Shanghai University)
Anomaly DetectionKnowledge DistillationImage
🎯 What it does: The first large-scale industrial defect detection dataset for 3C products, 3CAD, is proposed, and an unsupervised anomaly detection framework, CFRG, is developed based on this dataset.
3D Annotation-Free Learning by Distilling 2D Open-Vocabulary Segmentation Models for Autonomous Driving
Boyi Sun (Institute of Automation, Chinese Academy of Sciences), Fei-Yue Wang (Institute of Automation, Chinese Academy of Sciences)
SegmentationAutonomous DrivingKnowledge DistillationContrastive LearningPoint Cloud
🎯 What it does: A two-stage 3D annotation-free learning framework AFOV is constructed, which generates pseudo-labels for point cloud learning through a 2D open vocabulary segmentation model and performs knowledge distillation.
3D Denoisers Are Good 2D Teachers: Molecular Pretraining via Denoising and Cross-Modal Distillation
Sungjun Cho (University of Wisconsin Madison), Moontae Lee (LG AI Research)
Knowledge DistillationDrug DiscoveryGraph Neural NetworkTransformerGraph
🎯 What it does: This paper proposes a two-stage self-supervised pre-training framework D&D: first, it uses a 3D conformer to learn force field knowledge through a noise removal task, and then transfers the representations of the 3D encoder to the 2D molecular graph encoder via cross-modal knowledge distillation, enabling downstream property prediction using only 2D graphs.
3D Measurement of Complex Textured Objects Based on Bidirectional Fringe Projection
Yuchong Chen (Southeast University), Feipeng Da (Southeast University)
Point Cloud
🎯 What it does: A phase error correction and point cloud fusion method based on bidirectional stripe projection (ELPC‑PCF) is proposed to improve the accuracy of structured light measurement for complex textured objects.
3D-aware Select, Expand, and Squeeze Token for Aerial Action Recognition
Luying Peng (Nanjing University of Science and Technology), Guo-Sen Xie
RecognitionTransformerContrastive LearningVideo
🎯 What it does: A 3D-Tok framework is proposed, which uses a 3D-Token Selector (3TS) to select important tokens at both temporal and spatial levels, and then enhances semantics through an Expand-Squeeze Converter (ESC) to achieve action recognition in drone videos.
3D-RPE: Enhancing Long-Context Modeling Through 3D Rotary Position Encoding
Xindian Ma (Tianjin University), Nan Xu (Beijing Wenge Technology)
TransformerLarge Language ModelText
🎯 What it does: Proposes a 3D Rotational Position Encoding (3D-RPE) on the three-dimensional sphere to enhance the long-context modeling capabilities of large language models.
3D²-Actor: Learning Pose-Conditioned 3D-Aware Denoiser for Realistic Gaussian Avatar Modeling
Zichen Tang (Beihang University), Di Huang (Beihang University)
RestorationGenerationData SynthesisPose EstimationSupervised Fine-TuningDiffusion modelGaussian SplattingVideo
🎯 What it does: A 3D-aware denoiser is constructed based on pose-guided 2D denoising and 3D reconstruction interleaving, capable of generating high-fidelity, animatable human 3D avatars from sparse multi-view RGB videos.
3DHumanEdit: Multi-modal Body Part-aware Conditioning Information Integration for 3D Human Manipulation
FeiFan Xu (South China University of Technology), Si Wu (South China University of Technology)
GenerationData SynthesisPose EstimationTransformerVision Language ModelGenerative Adversarial NetworkImageTextMultimodality
🎯 What it does: This paper proposes the 3DHumanEdit method, which enables fine-grained, body-part-aware editing of 3D humans based on text descriptions and segmentation images, while maintaining multi-view consistency.
3DMambaIPF: A State Space Model for Iterative Point Cloud Filtering via Differentiable Rendering
Qingyuan Zhou (Fudan University), Ying He (Nanyang Technological University)
RestorationGraph Neural NetworkPoint Cloud
🎯 What it does: This paper proposes an iterative point cloud filtering framework based on Mamba, called 3DMambaIPF, and introduces a differentiable rendering loss to enhance surface detail recovery.
3DPGS: 3D Probabilistic Graph Search for Archaeological Piece Grouping
Junfeng Cheng (Imperial College London), Tania Stathaki (Imperial College London)
Graph Neural NetworkPoint CloudBenchmark
🎯 What it does: A new benchmark called 'Archaeological Fragment Grouping' is proposed, along with the corresponding dataset ArcPie and new evaluation metrics;
3SAT: A Simple Self-Supervised Adversarial Training Framework
Jiang Fang (Institute of Information Engineering, Chinese Academy of Sciences), Wei Ma (Institute of Information Engineering, Chinese Academy of Sciences)
Representation LearningAdversarial AttackConvolutional Neural NetworkGenerative Adversarial NetworkContrastive LearningImage
🎯 What it does: A 3SAT framework is proposed, utilizing raw unaugmented samples for self-supervised adversarial training, significantly enhancing the model's robustness and generalization performance.
4D Diffusion for Dynamic Protein Structure Prediction with Reference and Motion Guidance
Kaihui Cheng (Fudan University), Yuan Qi (Fudan University)
Protein Structure PredictionTransformerDiffusion modelBiomedical Data
🎯 What it does: A 4D diffusion model is proposed, capable of generating dynamic trajectory sequences of proteins that include both the main chain and side chains in one go.
A (1+ε)-Approximation for Ultrametric Embedding in Subquadratic Time
Gabriel Bathie (LaBRI Universite de Bordeaux), Guillaume Lagarde (LaBRI Universite de Bordeaux)
Tabular
🎯 What it does: An optimal supermetric embedding algorithm that can achieve arbitrary approximation quality in sub-quadratic time is proposed, along with the corresponding time and space complexity.
A Black-Box Evaluation Framework for Semantic Robustness in Bird’s Eye View Detection
Fu Wang (University of Exeter), Wenjie Ruan (University of Exeter)
Object DetectionAutonomous DrivingOptimizationPoint Cloud
🎯 What it does: A black-box semantic robustness evaluation framework is designed to quantify the worst-case robustness of BEV detection models under three types of semantic disturbances: geometric transformations, color shifts, and motion blur.
A Compact Implicit Neural Representation for Efficient Storage of Massive 4D Functional Magnetic Resonance Imaging
Ruoran Li (Tsinghua University), Jinli Suo (Tsinghua University)
CompressionBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A compression framework based on Implicit Neural Representation (INR) is proposed for 4D fMRI data compression.
A Compact Model for Mathematics Problem Representations Distilled from BERT
Hao Ming (Central China Normal University), Xiaopan Lyu (Central China Normal University)
CompressionKnowledge DistillationTransformerLarge Language ModelText
🎯 What it does: Based on the large language model BERT, a lightweight mathematical problem representation model with only three layers and a hidden dimension of 312 was constructed using knowledge distillation and multi-task pre-training.
A Complete Algorithm for Optimization Modulo Nonlinear Real Arithmetic
Fuqi Jia (Institute of Software Chinese Academy of Sciences), Jian Zhang (Institute of Software Chinese Academy of Sciences)
Optimization
🎯 What it does: The first complete OMT(NRA) solver is proposed, with the core algorithm being Optimization Cylindrical Algebraic Covering (OCAC), which is embedded in the CDCL(T) framework to form the CDCL(OCAC) solution.
A Comprehensive Evaluation on Event Reasoning of Large Language Models
Zhengwei Tao (Peking University), Kam-Fai Wong (Chinese University of Hong Kong)
TransformerLarge Language ModelPrompt EngineeringTextGraphBenchmark
🎯 What it does: This paper proposes a benchmark, EV 2, for evaluating the event reasoning capabilities of large language models and systematically assesses event reasoning performance across various relationships and reasoning paradigms.
A Comprehensive Overhaul of Multimodal Assistant with Small Language Models
Minjie Zhu (East China Normal University), Yaxin Peng (Shanghai University)
OptimizationTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelTextMultimodality
🎯 What it does: This paper proposes and implements a series of multimodal assistants based on small language models (≤3B parameters) — Mipha, systematically studying the impact of language models, visual representations, and optimization strategies on the performance of multimodal small language models (MSLMs), and validating their effectiveness through extensive benchmarking.
A Computationally Grounded Framework for Cognitive Attitudes
Tiago de Lima (University Artois), François Schwarzentruber (École Normale Supérieure Lyon)
🎯 What it does: A belief-based modal logic for cognitive attitudes is proposed, along with its semantics, proof, and model checking methods.
A Continuous-time Tractable Model for Present-biased Agents
Yasunori Akagi (NTT Corporation), Takeshi Kurashima (NTT Corporation)
OptimizationStochastic Differential EquationOrdinary Differential Equation
🎯 What it does: A continuous-time analytically tractable progress task model is proposed, studying agent behavior with immediate bias and solving the optimal goal setting and reward scheduling problem.
A Deep Probabilistic Framework for Continuous Time Dynamic Graph Generation
Ryien Hosseini (University of Chicago), Henry Hoffmann (University of Chicago)
GenerationData SynthesisGraph Neural NetworkGraphTime Series
🎯 What it does: This paper proposes a continuous-time dynamic graph generation framework called DG-Gen, which directly models event probabilities and supports autoregressive generation and link prediction.
A Diffusion-Based Framework for Occluded Object Movement
Zheng-Peng Duan (Nankai University), Chongyi Li (Tsinghua University)
RestorationGenerationDiffusion modelImage
🎯 What it does: A dual-branch framework called DiffOOM based on diffusion models is designed to complete the movement and repair of occluded objects while maintaining the integrity of the target object.
A Gaussian Filter-Based 3D Registration Method for Series Section Electron Microscopy
Zhenbang Zhang (Shandong University), Renmin Han (Shandong University)
Image TranslationRestorationOptimizationComputational EfficiencyConvolutional Neural NetworkOptical FlowImage
🎯 What it does: This study proposes a three-dimensional electron microscopy slice image registration method based on Gaussian filtering, which separates nonlinear distortion from natural deformation using a frequency domain approach and estimates the deformation field through a convolutional network, ultimately achieving fast and accurate three-dimensional reconstruction.
A Generalizable Anomaly Detection Method in Dynamic Graphs
Xiao Yang (Nanyang Technological University), Zhiqi Shen (Nanyang Technological University)
Anomaly DetectionGraph Neural NetworkTransformerGraph
🎯 What it does: This paper proposes a method called GeneralDyG for detecting anomalies in nodes and edges of dynamic graphs.
A Knowledge Distillation-Based Approach to Enhance Transparency of Classifier Models
Yuchen Jiang (Guilin University of Electronic Technology), Ahmad Chaddad (Guilin University of Electronic Technology)
Explainability and InterpretabilityComputational EfficiencyKnowledge DistillationConvolutional Neural NetworkImageBiomedical DataMagnetic Resonance ImagingAlzheimer's Disease
🎯 What it does: A simplified CNN model is trained using knowledge distillation, and hierarchical interpretability is achieved through average feature maps, intuitively demonstrating the diagnostic decision-making process.
A Label-free Heterophily-guided Approach for Unsupervised Graph Fraud Detection
Junjun Pan (Griffith University), Shirui Pan (University of Adelaide)
Anomaly DetectionGraph Neural NetworkGraph
🎯 What it does: This paper proposes an unsupervised graph fraud detection framework called HUGE, which utilizes a label-agnostic heterogeneity metric HALO and alignment loss to identify fraudulent nodes.
A Layer Selection Approach to Test Time Adaptation
Sabyasachi Sahoo (Universite Laval), Christian Gagné (Universite Laval)
Domain AdaptationConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: This study addresses the mid-layer selection problem in Test-Time Adaptation (TTA) and proposes a dynamic selection of the most suitable layers for updates based on gradient alignment, thereby improving the model's performance in new domains.
A Lightweight Sparse Interaction Network for Time Series Forecasting
Xu Zhang (Fudan University), Wei Wang (Fudan University)
TransformerTime Series
🎯 What it does: A lightweight sparse interaction network (LSINet) is proposed, achieving time series prediction through a multi-head sparse interaction mechanism and shared interaction learning.
A Logical Analysis of Hanabi
Elise Perrotin (Artificial Intelligence and Society Tokyo)
🎯 What it does: Formal modeling of the Hanabi game is conducted using the lightweight knowledge logic EL-O, and the reasoning limitations are addressed by introducing special reasoning actions, ultimately resulting in a finite semantics that requires only second-order modal depth.
A Lottery Ticket Hypothesis Approach with Sparse Fine-tuning and MAE for Image Forgery Detection and Localization
Jiaying Zhu (University of Science and Technology of China), Zheng-Jun Zha (University of Science and Technology of China)
Anomaly DetectionTransformerMixture of ExpertsAuto EncoderImage
🎯 What it does: Proposes the Forgery Masked Autoencoder (FMAE), which utilizes the Lottery Ticket Hypothesis to select forgery-sensitive parameters for sparse fine-tuning based on MAE, and incorporates a multi-source noise extractor to achieve image forgery detection and localization;
A Many-Objective Problem Where Crossover Is Provably Indispensable
Andre Opris (University of Passau)
Optimization
🎯 What it does: This paper studies the role of crossover operators in evolutionary multi-objective optimization (EMO), particularly the necessity of crossover in multi-objective optimization. Through theoretical analysis, it demonstrates that using crossover can significantly accelerate runtime, especially when the number of objectives exceeds two.
A Matching-Based Algorithm for the Traveling Tournament Problem
Jingyang Zhao (University of Electronic Science and Technology of China), Mingyu Xiao (University of Electronic Science and Technology of China)
Optimization
🎯 What it does: An effective algorithm based on minimum weight matching is designed to solve the travel tournament problem with a maximum of 4 consecutive away/home games (TTP-4).
A Method for Enhancing Generalization of Adam by Multiple Integrations
Long Jin (Lanzhou University), Zhenming Su (Lanzhou University)
OptimizationConvolutional Neural NetworkTransformerImageText
🎯 What it does: This paper proposes an optimizer called MIAdam, which adds multiple integral terms based on Adam. It utilizes the low-pass filtering effect of integrals to suppress high-frequency noise in gradients, guiding the training process towards flat minima, thereby enhancing the model's generalization ability and robustness to label noise while maintaining Adam's fast convergence characteristics.
A Modal Logic for Joint Abilities of Structured Strategies with Bounded Complexity
Ruiqi Jin (Sun Yat-sen University), Liping Xiong (Wuyi University)
🎯 What it does: A new modal logic SJAADL is proposed, which describes the joint capabilities in multi-agent systems using structured (non-deterministic) strategies, and its semantics and model checking algorithms are provided.
A Multi-Focus-Driven Multi-Branch Network for Robust Multimodal Sentiment Analysis
Chuanqi Tao (Nanjing University of Aeronautics and Astronautics), Peng Gao (Nanjing University of Aeronautics and Astronautics)
ClassificationRecognitionRecurrent Neural NetworkTransformerMultimodality
🎯 What it does: This paper proposes MFMB-Net, a multi-focus multi-branch network that combines fusion and reconstruction to achieve robust multimodal emotion analysis, particularly targeting scenarios with missing modalities.
A Multiagent Path Search Algorithm for Large-Scale Coalition Structure Generation
Redha Taguelmimt (Clermont Auvergne University), Tuomas Sandholm (Carnegie Mellon University)
OptimizationAgentic AIGraph
🎯 What it does: Developed the SALDAE algorithm, which adopts the multi-agent pathfinding concept to gradually expand nodes on the coalition structure graph and uses heuristic strategies to find high-value coalition structures.
A New Adversarial Perspective for LiDAR-based 3D Object Detection
Shijun Zheng (Xiamen University), Cheng Wang (Xiamen University)
Object DetectionAutonomous DrivingAdversarial AttackGenerative Adversarial NetworkPoint Cloud
🎯 What it does: A framework for LiDAR perception adversarial attacks based on random objects (water mist and smoke) is proposed, utilizing the generative adversarial network PCSGAN to generate random object point cloud sequences, which are then overlaid onto vehicle point clouds through range image rendering to attack various 3D object detection models.
A New Federated Learning Framework Against Gradient Inversion Attacks
Pengxin Guo (University of Hong Kong), Liangqiong Qu (University of Hong Kong)
Federated LearningSafty and PrivacyImage
🎯 What it does: This paper proposes a federated learning framework called HyperFL that utilizes hypernetworks to generate local model parameters, aiming to break the direct link between shared gradients and local private data, thereby defending against gradient inversion attacks.
A New Formula for Sticker Retrieval: Reply with Stickers in Multi-Modal and Multi-Session Conversation
Bingbing Wang (Harbin Institute of Technology), Ruifeng Xu (Harbin Institute of Technology)
RetrievalTransformerLarge Language ModelContrastive LearningTextMultimodality
🎯 What it does: A multi-modal multi-conversation sticker retrieval framework IGSR is proposed, and the MultiChat dataset is constructed.
A No Free Lunch Theorem for Human-AI Collaboration
Kenny Peng (Cornell University), Jon Kleinberg (Cornell University)
Classification
🎯 What it does: This paper proves that in a human-machine collaborative binary classification setting, if the collaboration strategy does not essentially revert to the decision of a single agent, then this strategy will, in certain situations, be worse than the worst agent, proposing a 'no free lunch' theorem;
A Novel Diffusion Model for Pairwise Geoscience Data Generation with Unbalanced Training Dataset
Junhuan Yang (George Mason University), Lei Yang
GenerationData SynthesisConvolutional Neural NetworkTransformerDiffusion modelMultimodality
🎯 What it does: A diffusion model called UB-Diff is proposed for generating multimodal imbalanced scientific data, capable of generating corresponding paired data using only a single modality.
A Pioneering Neural Network Method for Efficient and Robust Fuel Sloshing Simulation in Aircraft
Yu Chen (Xi'an Jiaotong University), Yan Chang (Xi'an Jiaotong University)
Convolutional Neural NetworkVideo
🎯 What it does: Designed and implemented the first neural network method capable of robustly simulating fuel fluctuations in complex fuel tanks.
A Plug-and-Play Bregman ADMM Module for Inferring Event Branches in Temporal Point Processes
Qingmei Wang (Renmin University of China), Hongteng Xu (Renmin University of China)
OptimizationExplainability and InterpretabilityTransformerReinforcement LearningTime SeriesSequential
🎯 What it does: This paper proposes a plugin module based on Bregman ADMM for inferring event branching structures in temporal point processes (TPP);
A Practical Approach to Causal Inference over Time
Martina Cinquini (University of Pisa), Isabel Valera (Saarland University)
Time Series
🎯 What it does: This paper studies the causal effects over time in discrete-time dynamic systems, defines temporal interventions, and maps vector autoregression (VAR) models to structural causal models (SCM), achieving causal inference based on observed time series data.
A Robust Prototype-Based Network with Interpretable RBF Classifier Foundations
Sascha Saralajew (NEC Laboratories Europe), Ammar Shaker (NEC Laboratories Europe)
ClassificationExplainability and InterpretabilityConvolutional Neural NetworkImage
🎯 What it does: This paper proposes an improved Component-Based Classification Network (CBC), closely related to RBF networks, and achieves interpretability and robustness through negative reasoning;
A Sample-Level Evaluation and Generative Framework for Model Inversion Attacks
Haoyang Li (Hong Kong Polytechnic University), Jianliang Xu (Hong Kong Baptist University)
GenerationData SynthesisSafty and PrivacyAdversarial AttackConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: This study investigates sample-level privacy assessment in model inversion attacks (MI) and proposes a new evaluation metric, DDCS, along with a GAN transfer learning enhancement framework based on natural gradient and entropy loss.
A Scalable and Effective Alternative to Graph Transformers
Kaan Sancak (Georgia Institute of Technology), Ümit V. Çatalyürek (Georgia Institute of Technology)
Computational EfficiencyRepresentation LearningGraph Neural NetworkTransformerGraphBenchmark
🎯 What it does: A new graph model GECO is proposed to replace the self-attention mechanism of traditional Graph Transformers, significantly improving computational efficiency and scalability while maintaining high-quality predictions.
A Similarity Paradigm Through Textual Regularization Without Forgetting
Fangming Cui (Shanghai Jiao Tong University), Jun Yu (Harbin Institute of Technology)
Domain AdaptationRepresentation LearningTransformerPrompt EngineeringContrastive LearningImage
🎯 What it does: On the pre-trained CLIP model, the SPTR method is proposed by fine-tuning text prompts and combining multiple handcrafted prompts to enhance the generalization ability of downstream tasks while maintaining general knowledge.
A Simple and Comprehensive Benchmark for Single-Cell Transcriptomics
Jiaxin Qi (Computer Network Information Center, Chinese Academy of Sciences), Gaogang Xie (Computer Network Information Center, Chinese Academy of Sciences)
Convolutional Neural NetworkBiomedical DataBenchmark
🎯 What it does: This paper systematically identifies and addresses three overlooked issues in single-cell transcriptome data: long-tail distribution, model selection, and evaluation, proposing weighted sampling, model structure adaptation, and a unified benchmark.
A Simple Graph Contrastive Learning Framework for Short Text Classification
Yonghao Liu (Jilin University), Renchu Guan (Jilin University)
ClassificationGraph Neural NetworkContrastive LearningText
🎯 What it does: A contrastive learning framework for short text classification called SimSTC is proposed, which does not require data augmentation and utilizes multi-view graph structures.
A Stochastic Approach to Bi-Level Optimization for Hyperparameter Optimization and Meta Learning
Minyoung Kim (Samsung AI Center), Timothy Hospedales (University of Edinburgh)
OptimizationHyperparameter SearchMeta LearningTransformerImageStochastic Differential Equation
🎯 What it does: Transforming the bi-level optimization problem into a probabilistic distribution form, using SGLD to sample the internal distribution and deriving a scalable hypergradient recursive algorithm.
A Syntactic Approach to Computing Complete and Sound Abstraction in the Situation Calculus
Liangda Fang (Jinan University), Quanlong Guan (Jinan University)
🎯 What it does: A syntactic method for synthesizing abstraction from low-level basic action theory to high-level action theory is proposed, utilizing linear integer situation calculus and Golog programs to achieve complete and provable abstraction.
A Systematic Exploration of Knowledge Graph Alignment with Large Language Models in Retrieval Augmented Generation
Shiyu Tian (Beijing University of Posts and Telecommunications), Xiaojie Wang (Beijing University of Posts and Telecommunications)
GenerationRetrievalTransformerLarge Language ModelTextGraphRetrieval-Augmented Generation
🎯 What it does: The system explores the role of Knowledge Graph Alignment (KGA) with large language models in Retrieval-Augmented Generation (RAG), subdividing KGA into two phases: graph transformation and textification, and comprehensively evaluates the impact of 81 graph features, 13 formats, 13 orders, and 14 templates.
A Theoretical Framework for an Efficient Normalizing Flow-Based Solution to the Electronic Schrödinger Equation
Daniel Freedman (Independent Researcher), Alex Bronstein (Technion - Israel Institute of Technology)
Flow-based ModelPhysics RelatedOrdinary Differential Equation
🎯 What it does: A method for solving the electronic Schrödinger equation based on normalizing flow is proposed, constructing a wave function Ansatz that can be directly sampled and satisfies fermionic symmetry and normalization conditions.
A Theory of Formalisms for Representing Knowledge
Heng Zhang (Zhejiang Lab), Donghui Quan (Nankai University)
🎯 What it does: A unified knowledge representation framework is proposed, defining the Knowledge Representation Form (KRF) and its general form, and proving that all general KRFs are equal under recursive isomorphism;
A Thorough Comparison Between Independent Cascade and Susceptible-Infected-Recovered Models
Panfeng Liu (Shanghai Jiao Tong University), Kuan Yang (Shanghai Jiao Tong University)
Graph
🎯 What it does: This paper systematically compares the Independent Cascade (IC) model with the Susceptible-Infected-Recovered (SIR) model, exploring the impact of edge dependence caused by node recovery in the SIR model on information dissemination and influence maximization.
A Training-free Synthetic Data Selection Method for Semantic Segmentation
Hao Tang (China University of Petroleum), Bingfeng Zhang (China University of Petroleum)
SegmentationData SynthesisContrastive LearningImage
🎯 What it does: A method for selecting synthetic data without training (SDS) is proposed, utilizing CLIP for perturbation similarity assessment (PCS) and mIoU-based annotation filtering (ASF) to filter high-quality samples from a large number of synthetic image-annotation pairs for training semantic segmentation models.
A Trusted Lesion-assessment Network for Interpretable Diagnosis of Coronary Artery Disease in Coronary CT Angiography
Xinghua Ma (Harbin Institute of Technology), Shuo Li (Case Western Reserve University)
Domain AdaptationExplainability and InterpretabilityConvolutional Neural NetworkTransformerImageBiomedical DataComputed Tomography
🎯 What it does: The paper proposes a trustworthy lesion assessment network (TLA-Net) for interpretable CAD diagnosis in coronary CT angiography.
A Unified Degradation-Robust Approach to SSL and UDA for 3D Medical Images
Suruchi Kumari (Indian Institute of Technology Roorkee), Pravendra Singh (Indian Institute of Technology Roorkee)
SegmentationDomain AdaptationKnowledge DistillationConvolutional Neural NetworkDiffusion modelImageBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: A unified framework is proposed that combines semi-supervised learning and unsupervised domain adaptation for 3D medical image segmentation, capable of handling limited annotations, domain shifts, and degraded images simultaneously.
A Unified Framework for Human-Allied Learning of Probabilistic Circuits
Athresh Karanam (University of Texas at Dallas), Sriraam Natarajan (University of Texas at Dallas)
OptimizationData-Centric LearningPoint CloudTabularBiomedical Data
🎯 What it does: A unified framework is proposed to encode domain knowledge as differentiable equality/inequality constraints, and to achieve constraint satisfaction in the parameter learning of probabilistic circuits (PC) through penalty terms.
A Unified Loss for Handling Inter-Class and Intra-Class Imbalance in Medical Image Segmentation
Feilong Xu (Jilin University), Xiaoli Zhang (Jilin University)
SegmentationConvolutional Neural NetworkImageBiomedical DataUltrasound
🎯 What it does: A unified 'Balance loss' is proposed to address the inter-class and intra-class imbalance issues in medical image segmentation, and its effectiveness is validated on five different segmentation tasks.
A Unified Model of Direct and Indirect Reciprocity in Multichannel Games
Juan Shi (Northwestern Polytechnical University), Zhen Wang (Northwestern Polytechnical University)
🎯 What it does: This paper proposes a unified framework that takes into account both direct and indirect reciprocity, dynamically assessing player behavior and reputation in a multi-channel repeated prisoner's dilemma while seeking the evolution of cooperation.
A Unifying Information-theoretic Perspective on Evaluating Generative Models
Alexis Fox (Duke University), Abhijin Adiga (University of Virginia)
GenerationData SynthesisDiffusion modelImage
🎯 What it does: A unified information theory-based k-nearest neighbors precision/recall evaluation framework is proposed, and three-dimensional metrics (PCE, RCE, RE) are designed to simultaneously assess the authenticity, mode loss, and mode collapse of generative models.
A Variable Occurrence-Centric Framework for Inconsistency Handling
Yakoub Salhi (University Artois)
🎯 What it does: A syntactic framework based on variable occurrences is proposed, using Minimum Inconsistent Relation (MIR) and Maximum Consistent Relation (MCR) to capture and restore inconsistencies in the propositional knowledge base, followed by the construction of various non-explosive inference relations and occurrence-based semantics based on MCR.
A Video-grounded Dialogue Dataset and Metric for Event-driven Activities
Wiradee Imrattanatrai (National Institute of Advanced Industrial Science and Technology), Teruko Mitamura (Language Technologies Institute Carnegie Mellon University)
Large Language ModelVision Language ModelVideoText
🎯 What it does: The VDAct video dialogue dataset and VDEval evaluation metrics are proposed to support event-driven multi-scenario video dialogue tasks.
A Wander Through the Multimodal Landscape: Efficient Transfer Learning via Low-rank Sequence Multimodal Adapter
Zirun Guo (Zhejiang University), Tao Jin (Zhejiang University)
Computational EfficiencyPrompt EngineeringMultimodality
🎯 What it does: Proposes a low-rank sequence multimodal adapter (Wander) to achieve parameter-efficient transfer learning for multimodal models.
A Wiener Process Perspective on Local Intrinsic Dimension Estimation Methods
Piotr Tempczyk (Institute of Informatics, University of Warsaw), Adam Kurpisz (ETH Zurich)
Diffusion modelImageStochastic Differential EquationAudio
🎯 What it does: This paper conducts a theoretical analysis of the local intrinsic dimension (LID) estimation method from the perspective of the Wiener process, providing a closed-form error description for the LIDL and FLIPD algorithms, and exploring the relationship between noise perturbation and data density evolution.
A-VL: Adaptive Attention for Large Vision-Language Models
Junyang Zhang (University of Science and Technology of China), Xiang-Yang Li (University of Science and Technology of China)
Computational EfficiencyTransformerVision Language ModelImageTextMultimodality
🎯 What it does: This paper proposes A-VL, an adaptive attention mechanism suitable for large visual-language models (LVLMs), which can significantly reduce KV cache storage and computation during the inference phase.
A²RNet: Adversarial Attack Resilient Network for Robust Infrared and Visible Image Fusion
Jiawei Li (University of Science and Technology Beijing), Huimin Ma (Dalian University of Technology)
Image TranslationObject DetectionSegmentationAdversarial AttackConvolutional Neural NetworkTransformerGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes an adversarial attack-resistant network A2RNet for the fusion of infrared and visible light images, which can maintain high-quality fusion results and sustain downstream task performance in the presence of adversarial perturbations.
AAKR: Adversarial Attack-based Knowledge Retention for Continual Semantic Segmentation
Zhidong Yu (University of Science and Technology of China), Wei Yang (University of Science and Technology of China)
SegmentationKnowledge DistillationAdversarial AttackConvolutional Neural NetworkImage
🎯 What it does: A knowledge retention framework based on adversarial attacks, AAKR, is proposed to enhance the retention of old knowledge in continuous semantic segmentation by applying attacks on new samples.
ABC3: Active Bayesian Causal Inference with Cohn Criteria in Randomized Experiments
Taehun Cha (Korea University), Donghun Lee (Korea University)
OptimizationTabular
🎯 What it does: A Bayesian active learning-based random experiment design algorithm ABC3 is proposed for efficiently estimating the Conditional Average Treatment Effect (CATE) under limited samples.
ABQ-LLM: Arbitrary-Bit Quantized Inference Acceleration for Large Language Models
Chao Zeng (ByteDance Inc), Xing Mei (ByteDance Inc)
OptimizationComputational EfficiencyTransformerLarge Language ModelText
🎯 What it does: Proposes ABQ-LLM, an arbitrary bit quantization inference framework that addresses performance degradation caused by low-bit quantization and the limitations of GPU quantization units;
Accelerated Diffusion via High-Low Frequency Decomposition for Pan-Sharpening
Ge Meng (Xiamen University), Xinghao Ding (Xiamen University)
RestorationSuper ResolutionConvolutional Neural NetworkDiffusion modelImage
🎯 What it does: By performing wavelet decomposition on multispectral images and panchromatic images, low-frequency coefficients are reconstructed using a diffusion model, and high-frequency coefficients are reconstructed using a high-frequency information recovery module, thus completing the reconstruction of high-resolution multispectral images;
Accelerated Methods with Compressed Communications for Distributed Optimization Problems Under Data Similarity
Dmitry Bylinkin (Moscow Institute of Physics and Technology), Aleksandr Beznosikov (Ivannikov Institute for System Programming of the Russian Academy of Sciences)
OptimizationTabular
🎯 What it does: A distributed optimization algorithm that balances unbiased/bias compression and acceleration is proposed under the condition of data similarity;
ACCon: Angle-Compensated Contrastive Regularizer for Deep Regression
Botao Zhao (Ping An Technology), Jianzong Wang (Ping An Technology)
Representation LearningContrastive LearningImageText
🎯 What it does: This paper proposes the ACCon angle compensation contrastive regularizer to improve the feature representation of deep regression models, maintaining the similarity and distance relationships between continuous labels.
Accurate and Regret-Aware Numerical Problem Solver for Tabular Question Answering
Yuxiang Wang (University of Melbourne), Junhao Gan (University of Melbourne)
Large Language ModelSupervised Fine-TuningTabularChain-of-Thought
🎯 What it does: To address the issue of inaccurate numerical calculations in table-based question answering, the TabLaP model is proposed: it uses a large language model (LLM) as a planner to generate Python calculation scripts, while the actual calculations are performed by an interpreter; it also integrates a state-of-the-art (SOTA) question answering branch, an answer selector (AnsSelector), and a credibility evaluator (TwEvaluator), forming a multi-branch, multi-LLM framework.
Accurate Estimation of Feature Importance Faithfulness for Tree Models
Mateusz Gajewski (University of Warsaw), Piotr Sankowski (University of Warsaw)
Tabular
🎯 What it does: This paper proposes a new prediction accuracy evaluation metric PGI² and provides an exact quadratic time complexity algorithm for tree models, while also designing a greedy feature ranking method based on this metric.
Accurate Link Prediction for Edge-Incomplete Graphs via PU Learning
Junghun Kim (Seoul National University), U Kang (Seoul National University)
Recommendation SystemAnomaly DetectionComputational EfficiencyGraph Neural NetworkSupervised Fine-TuningGraph
🎯 What it does: For graphs with missing edges, an iterative link prediction method based on PU learning, called PULL, is proposed. It utilizes the expected graph structure to perform soft inference on unobserved edges and continuously updates the predictor.
Accurate Nucleic Acid-Binding Residue Identification Based Domain-Adaptive Protein Language Model and Explainable Geometric Deep Learning
Wenwu Zeng (Hunan University), Shaoliang Peng (Hunan University)
Domain AdaptationExplainability and InterpretabilityProtein Structure PredictionGraph Neural NetworkSupervised Fine-TuningBiomedical Data
🎯 What it does: Developed the GeSite model, which combines domain-adaptive protein language models and E(3) equivariant graph neural networks to predict the binding residues of proteins with DNA or RNA;
Achieving Balanced Representation in School Choice with Diversity Goals
Zhaohong Sun (Kyushu University), Makoto Yokoo (Kyushu University)
OptimizationComputational EfficiencyFlow-based Model
🎯 What it does: Proposes the attribute of 'balanced representativeness' and designs a selection function that satisfies maximum diversity, non-wastefulness, justified envy-freeness, and balanced representativeness, implemented using flow networks and rank maximum matching;
Achieving Ensemble-Like Performance in a Single Model: A Feature Diversification Framework for Image-Text Matching
Zhao Zhou (Fudan University), Cheng Jin (Videt Lab)
RetrievalContrastive LearningImageText
🎯 What it does: A Feature Diversification Framework (FDF) is proposed under a single model, which simulates the ensemble effect of multiple models through shared feature extractors, feature transformations, alignment training strategies, and a re-weighting module, achieving ensemble-like performance in image-text matching.
Achieving Lightweight Super-Resolution for Real-Time Computer Graphics
Yu Wen (University of Houston), Xin Fu (University of Houston)
Super ResolutionComputational EfficiencyNeural Architecture SearchImage
🎯 What it does: A lightweight real-time super-resolution framework CGSR is proposed, which integrates rendering information (depth, normals, edges) into the network and achieves low parameters and efficient SR through NAS and dynamic pruning.
Achieving Maximin Share and EFX/EF1 Guarantees Simultaneously
Hannaneh Akrami (Max Planck Institute for Informatics), Nidhi Rathi (Max Planck Institute for Informatics)
Optimization
🎯 What it does: This paper proposes a new algorithm that can simultaneously satisfy the 2/3 approximate maximum minimum share (MMS) and EFX/EF1 fairness constraints when allocating indivisible items.
Achieving Speed-Accuracy Balance in Vision-based 3D Occupancy Prediction via Geometric-Semantic Disentanglement
Yulin He (National University of Defense Technology), Yusong Tan (National University of Defense Technology)
SegmentationDepth EstimationAutonomous DrivingComputational EfficiencyConvolutional Neural NetworkVideoBenchmark
🎯 What it does: A 3D occupancy prediction method based on geometric-semantic decoupling, GSD-Occ, is proposed to achieve both real-time performance and high accuracy.
ACPBench: Reasoning About Action, Change, and Planning
Harsha Kokel (IBM Research), Shirin Sohrabi (IBM Research)
Large Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: This paper presents ACPBench—a large-scale benchmark based on PDDL, focused on seven reasoning tasks related to actions, changes, and planning, for evaluating the capabilities of LLMs in planning-related reasoning.
Acting Beyond Learning: Imagination-Assisted Decision-Making in the Visual-based Multi-Agent Cooperative Scenarios
Huanhuan Yang (National University of Defense Technology), Shaowu Yang (Academy of Military Sciences)
OptimizationRobotic IntelligenceReinforcement LearningAuto EncoderContrastive LearningImage
🎯 What it does: This paper proposes a latent world model CLWPO based on the Contrastive Variational Bound (CVB), which combines model-free learning with heuristic strategy optimization (HPO) to enhance sample efficiency and performance in multi-agent visual cooperation tasks.
Action-Agnostic Point-Level Supervision for Temporal Action Detection
Shuhei M. Yoshida (NEC Corporation), Masashi Sugiyama (The University of Tokyo)
RecognitionObject DetectionConvolutional Neural NetworkContrastive LearningVideo
🎯 What it does: A motion-agnostic point-level (AAPL) annotation scheme without human intervention is proposed, along with a corresponding detection model and training method, achieving action instance detection under lightweight annotation.
Active Fourier Auditor for Estimating Distributional Properties of ML Models
Ayoub Ajarra (Inria), Debabrota Basu (Max Planck Institute for Software Systems)
Tabular
🎯 What it does: This paper proposes the Active Fourier Auditor (AFA), a unified auditing framework for black-box machine learning models that estimates the robustness, individual fairness, and group fairness of the model's distributional properties using Fourier coefficients.
Active Large Language Model-Based Knowledge Distillation for Session-Based Recommendation
Yingpeng Du (Nanyang Technological University), Yew-Soon Ong (Nanyang Technological University)
Recommendation SystemKnowledge DistillationTransformerLarge Language ModelPrompt EngineeringSequential
🎯 What it does: A knowledge distillation method for LLM based on active learning (ALKDRec) is proposed, which efficiently distills LLM knowledge to lightweight recommendation models using a small number of instances in the conversational recommendation task.
Active Reinforcement Learning Strategies for Offline Policy Improvement
Ambedkar Dukkipati (Indian Institute of Science), Prabhas Reddy Onteru (Indian Institute of Science)
Robotic IntelligenceReinforcement LearningContrastive LearningTabular
🎯 What it does: This paper proposes an active reinforcement learning strategy that enhances the performance of offline policies by actively collecting information-rich trajectories using existing offline data under a limited budget.
Active Symbolic Discovery of Ordinary Differential Equations via Phase Portrait Sketching
Nan Jiang (Purdue University), Yexiang Xue (Cornell University)
OptimizationTransformerReinforcement LearningTime SeriesPhysics RelatedOrdinary Differential Equation
🎯 What it does: This paper proposes a method for discovering active symbolic ODEs (Ordinary Differential Equations) through phase diagram sketching, utilizing an active learning strategy to select information-rich regions in phase space, and then sampling batches of initial conditions from these regions to query real trajectory data, thereby more efficiently discovering the symbolic dynamical equations of the system.
AD4CD: Causal-Guided Anomaly Detection for Enhancing Cognitive Diagnosis
Haiping Ma (Anhui University), Yong Yang (Anhui University)
Anomaly DetectionAuto EncoderTime Series
🎯 What it does: The AD4CD framework is proposed, which combines causal inference and anomaly detection to capture abnormal states of students and questions through response time distribution, thereby improving the accuracy of cognitive diagnosis.
AdaCo: Overcoming Visual Foundation Model Noise in 3D Semantic Segmentation via Adaptive Label Correction
Pufan Zou (Xiamen University), Cheng Wang (Xiamen University)
SegmentationAutonomous DrivingPoint Cloud
🎯 What it does: A label-free 3D semantic segmentation framework called AdaCo is proposed, which utilizes visual foundation models to generate cross-modal pseudo-labels and iteratively improves segmentation accuracy through adaptive noise correction and adaptive robust loss.
AdaDiff: Adaptive Step Selection for Fast Diffusion Models
Hui Zhang (Fudan University), Yu-Gang Jiang (ByteDance Inc.)
GenerationOptimizationComputational EfficiencyReinforcement LearningDiffusion modelImageVideo
🎯 What it does: Designed and implemented the AdaDiff framework, which uses reinforcement learning to learn adaptive diffusion step strategies for each text prompt, significantly reducing inference time while maintaining image/video quality.
AdaDPCC: Adaptive Rate Control and Rate-Distortion-Complexity Optimization for Dynamic Point Cloud Compression
Chenhao Zhang (Peking University), Wei Gao (Peking University)
CompressionOptimizationAuto EncoderPoint Cloud
🎯 What it does: The AdaDPCC framework is proposed, achieving dynamic point cloud compression with variable bit rates and variable computational complexity, including multi-path encoding, coarse-to-fine motion estimation and compensation, and a lightweight rate control module.
AdaGK-SGD: Adaptive Global Knowledge Guided Distributed Stochastic Gradient Descent
Hangyu Ye (Xidian University), Leyuan Fang (Hunan University)
ClassificationOptimizationConvolutional Neural NetworkImage
🎯 What it does: Proposes AdaGK-SGD, an algorithm that guides global knowledge in distributed training without additional communication through a maximum lifetime global knowledge module.
AdaO2B: Adaptive Online to Batch Conversion for Out-of-Distribution Generalization
Xiao Zhang (Renmin University of China), Zhenhua Dong (Huawei Noah's Ark Lab)
Domain AdaptationRecommendation SystemReinforcement LearningVideo
🎯 What it does: An adaptive online-to-batch conversion method called AdaO2B is proposed to address the generalization problem of discrete distributions in scenarios such as online recommendation.
Adapting to Non-Stationary Environments: Multi-Armed Bandit Enhanced Retrieval-Augmented Generation on Knowledge Graphs
Xiaqiang Tang (Hong Kong University of Science and Technology), Sihong Xie (Tencent Hunyuan)
RetrievalOptimizationTransformerReinforcement LearningGraphRetrieval-Augmented Generation
🎯 What it does: A multi-armed bandit (MAB) driven retrieval-enhanced generation framework is proposed, utilizing a knowledge graph as the underlying knowledge base to achieve dynamic selection and real-time optimization of various retrieval methods.