AAAI 2025 Papers — Page 13
AAAI Conference on Artificial Intelligence · 3028 papers
FROC: Building Fair ROC from a Trained Classifier
Avyukta Manjunatha Vummintala (International Institute of Information Technology), Sujit Gujar (International Institute of Information Technology)
ClassificationOptimizationTabular
🎯 What it does: A post-processing method called FROC is proposed to convert the scoring function of a trained binary classifier into a probabilistic classifier that satisfies the 'ε-Equalized ROC' fairness constraint without retraining the model, while ensuring minimal AUC loss.
From 2D CAD Drawings to 3D Parametric Models: A Vision-Language Approach
Xilin Wang (Beihang University), Zihan Zhou (Manycore Tech Inc.)
GenerationRetrievalTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodality
🎯 What it does: This paper proposes CAD2PROGRAM, a method for recovering 3D parametric models from 2D CAD drawings based on a vision-language model.
From Coarse to Fine: A Matching and Alignment Framework for Unsupervised Cross-View Geo-Localization
Xueyi Wang (Beijing Institute of Technology), Fang Deng (Beijing Institute of Technology)
RetrievalDomain AdaptationConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: An unsupervised cross-view geographic localization framework is proposed, which achieves cross-view feature alignment through clustering alignment and intermediate state transformation.
From Logistic Regression to the Perceptron Algorithm: Exploring Gradient Descent with Large Step Sizes
Alexander Tyurin (AIRI Moscow Russia Skoltech Moscow Russia)
ClassificationOptimizationImage
🎯 What it does: This study investigates the behavior of using large step size gradient descent to solve logistic regression on separable data and reveals its equivalence to the batch perceptron algorithm.
From Pairwise to Ranking: Climbing the Ladder to Ideal Collaborative Filtering with Pseudo-Ranking
Yuhan Zhao (Harbin Engineering University), Hongtao Song (Harbin Engineering University)
Recommendation SystemTabular
🎯 What it does: Proposes the Pseudo Ranking Paradigm (PRP), which addresses the issue of traditional collaborative filtering lacking complete ranking information by generating pseudo rankings and using ranking loss.
From PEFT to DEFT: Parameter Efficient Finetuning for Reducing Activation Density in Transformers
Bharat Runwal (Independent Researcher), Pin-Yu Chen (IBM Research)
Computational EfficiencyTransformerSupervised Fine-TuningText
🎯 What it does: This paper proposes two methods, DEFT and ADA-DEFT, which induce activation sparsity in the MLP of Transformers through a differentiable sparse loss within the framework of parameter-efficient fine-tuning (PEFT), thereby improving inference efficiency while maintaining downstream task performance.
From Representation Space to Prognostic Insights: Whole Slide Image Generation with Hierarchical Diffusion Model for Survival Prediction
Zhihao Tang (Beijing University of Posts and Telecommunications), Chaozhuo Li (Beijing University of Aeronautics and Astronautics)
GenerationData SynthesisDiffusion modelImageBiomedical Data
🎯 What it does: Based on a hierarchical diffusion model, generate missing WSI representations and integrate them into the survival prediction process.
From Words to Worth: Newborn Article Impact Prediction with LLM
Penghai Zhao (Nankai University), Xiang Li (Nankai University)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Using titles and abstracts, we predict the future impact of newly published articles through low-rank adaptation and quantization fine-tuning of large language models; simultaneously, we improve and adopt a new metric TNCSI SP;
From Your Block to Our Block: How to Find Shared Structure Between Stochastic Block Models over Multiple Graphs
Iiro Kumpulainen (University of Helsinki), Nikolaj Tatti (University of Helsinki)
OptimizationGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: This paper proposes the Shared Stochastic Block Model (SSBM) for identifying shared block structures in multiple unaligned and differently sized graphs, and presents a two-phase solution framework: first, using MCMC or single-layer/multi-layer methods to obtain block partitions for each graph, and then selecting shared blocks through Integer Linear Programming (ILP) or greedy heuristics, ultimately achieving the maximum likelihood shared SBM.
Frozen Language Models Are Gradient Coherence Rectifiers in Vision Transformers
Lichen Bai (Shenzhen International Graduate School, Tsinghua University), Hong-Gee Kim (Seoul National University)
ClassificationRecognitionOptimizationTransformerLarge Language ModelSupervised Fine-TuningImageVideo
🎯 What it does: The study inserts frozen LLM blocks into the Vision Transformer (ViT), analyzes the effect of gradient consistency improvement during the training process, and proposes an auxiliary training strategy that uses LLM only during the training phase to avoid additional costs during inference.
FSL-Rectifier: Rectify Outliers in Few-Shot Learning via Test-Time Augmentation
Yunwei Bai (National University of Singapore), Tsuhan Chen (National University of Singapore)
ClassificationAnomaly DetectionMeta LearningGenerative Adversarial NetworkImage
🎯 What it does: A testing-time augmentation-based FSL Rectifier framework is proposed, which corrects the outlier effects of test samples by generating synthetic samples.
FSTA-SNN:Frequency-Based Spatial-Temporal Attention Module for Spiking Neural Networks
Kairong Yu (Zhejiang University), Qi Xu (Dalian University of Technology)
Spiking Neural NetworkImage
🎯 What it does: This paper explores the learning preferences of intermediate output spikes in SNNs through frequency domain analysis and proposes a Frequency-based Space-Time Attention (FSTA) module that effectively suppresses redundant spikes and enhances feature learning.
Fully Test-time Adaptation for Tabular Data
Zhi Zhou (Nanjing University), Yu-Feng Li (Nanjing University)
Domain AdaptationTabularBenchmark
🎯 What it does: This paper studies the Full-Time Testing Adaptation (FTTA) problem on tabular data and proposes the FTAT method.
Fully-Scalable Massively Parallel Algorithm for k-center with Outliers
Di Wu (Central South University), Jianxin Wang (Central South University)
Optimization
🎯 What it does: This paper presents the first algorithm for solving the k-center and outlier (k, z)-center problems under a Fully-Scalable MPC model, capable of performing clustering with sublinear (Θ(n^δ)) storage space on each machine.
Functional Connectomes of Neural Networks
Tananun Songdechakraiwut (Duke University), Yutong Wu (Duke University)
Explainability and InterpretabilityComputational EfficiencyGraph Neural NetworkImage
🎯 What it does: This study investigates a method that combines functional connectivity groups of neural networks with those of humans, characterizing the functional topology of large neural networks through persistent graph homology and Wasserstein statistics, and validating its interpretability using unsupervised clustering.
FunEditor: Achieving Complex Image Edits via Function Aggregation with Diffusion Models
Mohammadreza Samadi (Huawei Technologies), Di Niu (University of Alberta)
Image TranslationImage HarmonizationGenerationComputational EfficiencyDiffusion modelImage
🎯 What it does: We propose FunEditor, which achieves complex image editing tasks by aggregating atomic editing functions, supporting multi-task execution simultaneously and completing it in few-step inference.
Fusing Pruned and Backdoored Models: Optimal Transport-based Data-free Backdoor Mitigation
Weilin Lin (Hong Kong University of Science and Technology), Hui Xiong (Hong Kong University of Science and Technology)
OptimizationAdversarial AttackImage
🎯 What it does: This paper proposes a completely data-independent post-training anti-poisoning method called OTBR. It first prunes the model using randomly generated untrained neuron weight changes (NWC), and then aligns and merges the pruned model with the original model through optimal transport (OT) to restore performance and suppress backdoors.
Future Sight and Tough Fights: Revolutionizing Sequential Recommendation with FENRec
Yu-Hsuan Huang (National Yang Ming Chiao Tung University), Wen-Huang Cheng (National Taiwan University)
Recommendation SystemTransformerContrastive LearningSequential
🎯 What it does: A new framework called FENRec is proposed in sequential recommendation, which utilizes future interaction information to generate time-dependent soft labels and incorporates persistent hard negative samples in contrastive learning to address the sparse data problem.
Fuzzy Collaborative Reasoning
Huanhuan Yuan (Soochow University), Victor S. Sheng (Texas Tech University)
Recommendation SystemTabularSequential
🎯 What it does: Reformulate the sequential recommendation task as first-order logic query answering, and implement collaborative reasoning using fuzzy logic operations, proposing the FuzzCR model.
G-VEval: A Versatile Metric for Evaluating Image and Video Captions Using GPT-4o
Tony Cheng Tong (Hong Kong University of Science and Technology), Dit-Yan Yeung (Hong Kong University of Science and Technology)
GenerationRetrievalTransformerLarge Language ModelPrompt EngineeringImageVideoTextChain-of-Thought
🎯 What it does: A new visual subtitle evaluation metric G-VEval has been designed, supporting the evaluation of subtitles for images and videos.
G2LDetect: A Global-to-Local Approach for Hallucination Detection
Xiaoxia Cheng (Zhejiang University), Weiming Lu (Zhejiang University)
Anomaly DetectionTransformerLarge Language ModelTextBenchmark
🎯 What it does: A global-to-local hallucination detection framework G2LDetect is proposed, which first constructs a hierarchical tree structure for global representation of text, and then extracts local details for detection and aggregates results along paths from the tree.
GaitCycFormer: Leveraging Gait Cycles and Transformers for Gait Emotion Recognition.
Qingyang Zeng (Nanjing University), Lin Shang (Nanjing University)
RecognitionGraph Neural NetworkTransformerTime Series
🎯 What it does: The GaitCycFormer framework is proposed, utilizing gait cycles and a dual-layer Transformer for the recognition of 3D skeletal gait emotions.
Game4Loc: A UAV Geo-Localization Benchmark from Game Data
Yuxiang Ji (Xiamen University), Liaoni Wu (Xiamen University)
RetrievalDomain AdaptationTransformerContrastive LearningImageBenchmark
🎯 What it does: A UAV geographic positioning benchmark dataset GTA-UAV was constructed based on the simulation of 'GTA V', and a weighted contrastive learning method, weighted-InfoNCE, was introduced to achieve partial matching cross-view retrieval and positioning.
GapMatch: Bridging Instance and Model Perturbations for Enhanced Semi-Supervised Medical Image Segmentation
Wei Huang (Sichuan University), Yan Wang (Institute of High Performance Computing)
SegmentationImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: Proposes the GapMatch framework, which combines instance perturbation and model perturbation for semi-supervised medical image segmentation.
GarFast: Realistic and Fast Garment Transfer with a Simplified Parser-Free Approach
Chenghu Du (Wuhan University of Technology), Shengwu Xiong (Wuhan College)
Image TranslationGenerationConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: Designed and implemented a lightweight single network GarFast, completing high-resolution realistic clothing try-on tasks without a parser.
GARLIC: GPT-Augmented Reinforcement Learning with Intelligent Control for Vehicle Dispatching
Xiao Han (City University of Hong Kong), Jieping Ye (Harbin Institute of Technology)
Autonomous DrivingOptimizationRecurrent Neural NetworkGraph Neural NetworkTransformerLarge Language ModelReinforcement LearningContrastive LearningGraphTabularTime Series
🎯 What it does: A multi-view graph + GPT reinforcement learning framework named GARLIC has been designed and implemented for online ride-hailing vehicle scheduling, taking into account traffic conditions and driver behavior.
GAS: Generative Activation-Aided Asynchronous Split Federated Learning
Jiarong Yang (South China University of Technology), Yuan Liu (South China University of Technology)
Federated LearningImage
🎯 What it does: An asynchronous split federated learning framework GAS is proposed, which utilizes activation buffers, model buffers, and generative activation techniques to mitigate model update bias caused by slow devices and improve convergence speed.
Gaussian Graphical Modelling Without Independence Assumptions for Uncentered Data
Bailey Andrew (University of Leeds), Luisa Cutillo (University of Leeds)
OptimizationGraph Neural NetworkVideoGraphBiomedical Data
🎯 What it does: This paper proposes the incorporation of a mean estimation wrapper into the multi-axis Gaussian graphical model with a Kronecker-sum structure to correct the bias caused by the traditional zero-mean assumption.
GaussianPainter: Painting Point Cloud into 3D Gaussians with Normal Guidance
Jingqiu Zhou (Beihang University), Hongsheng Li (Beihang University)
GenerationData SynthesisGaussian SplattingImagePoint Cloud
🎯 What it does: Generate high-quality 3D Gaussian fields from point clouds through a single forward inference guided by a given reference image, achieving an efficient conversion from point clouds to Gaussians.
GaussianSR: High Fidelity 2D Gaussian Splatting for Arbitrary-Scale Image Super-Resolution
Jintong Hu (Tsinghua University), Lei Zhang (Hong Kong Polytechnic University)
RestorationSuper ResolutionGaussian SplattingImage
🎯 What it does: This paper proposes GaussianSR, which performs arbitrary scale super-resolution reconstruction of low-resolution images through 2D Gaussian splatting.
Gaze Label Alignment: Alleviating Domain Shift for Gaze Estimation
Guanzhong Zeng (Hikvision Research Institute), Jiang Zhu (Hikvision Research Institute)
Pose EstimationDomain AdaptationImage
🎯 What it does: Proposes the Gaze Label Alignment algorithm (GLA), which eliminates cross-domain errors by aligning label distributions from different domains, and positions it as a key technology for domain generalization.
GBRIP: Granular Ball Representation for Imbalanced Partial Label Learning
Jintao Huang (Hong Kong Baptist University), Wenbin Qian (University of Macau)
ClassificationImage
🎯 What it does: The GB-RIP framework is proposed, using coarse-grained granular balls to represent label disambiguation in imbalanced partial label learning with multi-center loss.
GCAD: Anomaly Detection in Multivariate Time Series from the Perspective of Granger Causality
Zehao Liu (Hangzhou Dianzi University), Pengfei Jiao (Hangzhou Dianzi University)
Anomaly DetectionTime Series
🎯 What it does: A multivariate time series anomaly detection framework based on Granger causality, GCAD, is proposed, which utilizes a deep predictor gradient to dynamically discover causal graphs and detect anomalies through deviation detection.
GCD-Sampling: A General Cross-scale Decoupled Sampling for Point Cloud
Tao Dai (Shenzhen University), Zexuan Zhu (Tsinghua University)
ClassificationSegmentationPoint Cloud
🎯 What it does: A general cross-scale decoupled point cloud sampling method called GCD-sampling is proposed, which can adaptively sample point clouds without changing the target network structure, and achieve fine control over the sampling point positions through cross-scale feature fusion and convex combination learning.
GCD: Advancing Vision-Language Models for Incremental Object Detection via Global Alignment and Correspondence Distillation
Xu Wang (University of Science and Technology of China), Zihan Lin (University of Science and Technology of China)
Object DetectionKnowledge DistillationVision Language ModelImage
🎯 What it does: The GCD method is proposed for incremental object detection tasks, aiming to prevent catastrophic forgetting in visual-language detectors through global semantic alignment and semantic correspondence distillation.
GDiffRetro: Retrosynthesis Prediction with Dual Graph Enhanced Molecular Representation and Diffusion Generation
Shengyin Sun (City University of Hong Kong), Chen Ma (City University of Hong Kong)
GenerationDrug DiscoveryGraph Neural NetworkDiffusion modelGraph
🎯 What it does: A backward synthesis prediction framework GDiffRetro is proposed, which is based on dual graph enhanced molecular representation and 3D conditional diffusion. It first identifies the reaction center to generate synthetic fragments, and then uses a diffusion model to generate complete reactants from the synthetic fragments.
GeCC: Generalized Contrastive Clustering with Domain Shifts Modeling
Yujie Chen (Shenzhen University), Debby D. Wang (Hong Kong Metropolitan University)
Representation LearningConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: A general contrastive clustering framework GeCC is proposed, which utilizes a clustering-guided domain transformation modeling module and attention weights to enhance unsupervised clustering performance.
GenAL: Generative Agent for Adaptive Learning
Rui Lv (University of Science and Technology of China), Linbo Zhu (University of Science and Technology of China)
Recommendation SystemTransformerLarge Language ModelReinforcement LearningAgentic AIPrompt EngineeringText
🎯 What it does: A generative intelligent agent GenAL based on large language models has been designed for adaptive learning path recommendation, which includes a global thinking agent and a local teaching agent, capable of dynamically recommending exercises based on learners' historical records and text semantics.
GenAuction: A Generative Auction for Online Advertising
Yuchao Ma (Renmin University of China), Lin Liu (Baidu Inc.)
Recommendation SystemOptimizationTransformerReinforcement LearningTabular
🎯 What it does: A KPI-based advertising auction mechanism called GenAuction is proposed, which uses a Generator-Evaluator structure to directly optimize multi-objective allocation at the page view (PV) level and implements incentive compatibility through VCG billing.
General Uncertainty Estimation with Delta Variances
Simon Schmitt (DeepMind), Hado van Hasselt (DeepMind)
Graph Neural NetworkGraphTime Series
🎯 What it does: This paper proposes and evaluates the Delta Variance family, which is used to estimate the uncertainty in model outputs (including implicit quantification targets) caused by parameter uncertainty when only limited training data is available, and applies it to various quantification targets such as weather forecasting.
Generalizable Sensor-Based Activity Recognition via Categorical Concept Invariant Learning
Di Xiong (Nanjing Normal University), Chaolei Han (Southeast University)
RecognitionDomain AdaptationConvolutional Neural NetworkTime Series
🎯 What it does: A new regularization framework CCIL is proposed, which uses a concept matrix to learn cross-domain consistency, thereby improving the activity recognition performance of wearable sensors.
Generalization Analysis for Deep Contrastive Representation Learning
Nong Minh Hieu (Singapore Management University), Cheng Yeaw Ku (Nanyang Technological University)
OptimizationRepresentation LearningContrastive LearningImage
🎯 What it does: This paper presents a generalization error upper bound for Deep Contrastive Representation Learning (DCRL), revealing the relationship between model complexity, network parameters, spectral norms, and the number of negative samples.
Generalization of Graph Neural Networks Is Robust to Model Mismatch
Zhiyang Wang (University of Pennsylvania), Alejandro Ribeiro (University of Pennsylvania)
ClassificationGraph Neural NetworkGraph
🎯 What it does: Analyzed the generalization performance of Graph Neural Networks (GNNs) on node and graph classification tasks under the condition of model mismatch (i.e., inconsistency between the manifolds of training and testing graphs), and proved that under the constraints of low-pass and integral Lipschitz filters, GNNs can robustly generalize to unseen nodes and graphs.
Generalized Class Discovery in Instance Segmentation
Cuong Manh Hoang (Seoul National University of Science and Technology), Byeongkeun Kang (Chung-Ang University)
Object DetectionSegmentationContrastive LearningImage
🎯 What it does: This paper proposes a Generalized Category Discovery (GCD) framework for instance segmentation, capable of simultaneously identifying known and newly emerging category instances in a mixed scenario of labeled and unlabeled data.
Generalized Convergence Analysis of Tsetlin Automaton Based Algorithms: A Probabilistic Approach to Concept Learning
Mohamed-Bachir Belaid (NILU Climate and Environmental Research Institute), Anis Yazidi (Oslo Metropolitan University)
ClassificationOptimizationTabular
🎯 What it does: Introduces the Probabilistic Concept Learning (PCL) framework and proves its almost certain convergence to any literal conjunction under noise-free samples, while validating its effectiveness through experiments.
Generalized Debiased Semi-Supervised Hashing for Large-Scale Image Retrieval
Xingbo Liu (Shandong Jianzhu University), Yilong Yin (Shandong University)
RetrievalImage
🎯 What it does: A general-purpose debiased semi-supervised hashing framework GDSH is proposed to address pseudo-label noise and optimization bias, enhancing large-scale image retrieval performance.
Generalized Dimension Reduction Using Semi-Relaxed Gromov-Wasserstein Distance
Ranthony A. Clark (Duke University), Thomas Weighill (University of North Carolina at Greensboro)
OptimizationImagePoint Cloud
🎯 What it does: In the high-dimensional point cloud dimensionality reduction task, the authors propose a manifold-valued multidimensional scaling (MDS) framework based on the semi-relaxed Gromov-Wasserstein (srGW) distance, which can solve for optimal embeddings in non-Euclidean target spaces.
Generalized Implicit Neural Representations for Dynamic Molecular Surface Modeling
Fang Wu (Stanford University), Stan Z. Li (Westlake University)
Drug DiscoveryProtein Structure PredictionMixture of ExpertsPoint Cloud
🎯 What it does: This paper studies dynamic protein surface modeling and proposes a dynamic surface representation framework MoE-DSR based on Mixture-of-Experts.
Generalized Zero-Shot Learning for Point Cloud Segmentation with Evidence-Based Dynamic Calibration
Hyeonseok Kim (Seoul National University of Science and Technology), Yeejin Lee (Seoul National University of Science and Technology)
SegmentationGenerative Adversarial NetworkPoint Cloud
🎯 What it does: A dynamic calibration method based on evidence uncertainty estimation, E3DPC-GZSL, is proposed for generalized zero-shot point cloud semantic segmentation.
Generalizing Constraint Models in Constraint Acquisition
Dimos Tsouros (KU Leuven), Tias Guns (University College Cork)
ClassificationOptimizationExplainability and InterpretabilityTabularBenchmark
🎯 What it does: This paper proposes a constraint acquisition method called GENCON, which utilizes constraint-level classification to learn parameterized constraint models, thereby achieving model generalization across instances.
Generating Counterfactual Explanations Under Temporal Constraints
Andrei Buliga (Fondazione Bruno Kessler), Massimiliano Ronzani (Fondazione Bruno Kessler)
OptimizationExplainability and InterpretabilityTime SeriesSequential
🎯 What it does: This paper proposes a counterfactual explanation method for generating process trajectories under time constraints, which ensures that the generated trajectories always satisfy the business process background knowledge expressed by Linear Temporal Logic (LTLp).
Generating Synthetic Data for Unsupervised Federated Learning of Cross-Modal Retrieval
Tianlong Zhang (Beijing University of Posts and Telecommunications), Yuankai Qi (Macquarie University)
Data SynthesisRetrievalFederated LearningGenerative Adversarial NetworkContrastive LearningImageTextMultimodality
🎯 What it does: A FedWGAN method is proposed, which generates high-quality pseudo image-text pairs through locally trained WGANs within a federated learning framework, addressing the issue of non-IID data in unsupervised cross-modal retrieval, and uses prototype filtering of generated samples on the server side for global model training.
Generating Traffic Scenarios via In-Context Learning to Learn Better Motion Planner
Aizierjiang Aiersilan (University of Macau)
Data SynthesisAutonomous DrivingLarge Language ModelVideoText
🎯 What it does: Developed the AutoSceneGen framework, which utilizes in-context learning of large language models to automatically convert user-input text descriptions into scripts executable in simulators like CARLA, thereby generating diverse and safety-critical traffic scenarios for training more robust motion planners.
Generative Medical Segmentation
Jiayu Huo (King's College London), Rachel Sparks (Shanghai United Imaging Intelligence Co., Ltd.)
SegmentationGenerationDomain AdaptationConvolutional Neural NetworkTransformerDiffusion modelAuto EncoderImageBiomedical DataUltrasound
🎯 What it does: This paper proposes Generative Medical Segmentation (GMS), which utilizes a pre-trained vision foundation model (Stable Diffusion VAE) to extract latent representations of images and masks, trains a lightweight latent mapping network to achieve mapping from image latent to mask latent, and then decodes to obtain segmentation results.
Generative Planning with 3D-Vision Language Pre-training for End-to-End Autonomous Driving
Tengpeng Li (Tongji University), Pai Peng (COWAROBOT)
Autonomous DrivingTransformerLarge Language ModelVision Language ModelContrastive LearningMultimodality
🎯 What it does: An end-to-end autonomous driving framework GPVL is proposed, combining 3D vision with a language pre-trained generative planning method.
Generative Video Diffusion for Unseen Novel Semantic Video Moment Retrieval
Dezhao Luo (Queen Mary University of London), Yang Liu (Peking University)
GenerationRetrievalDomain AdaptationDiffusion modelVideoText
🎯 What it does: This paper proposes a Fine-grained Video Editing framework (FVE) that utilizes a generative video diffusion model to generate video segments that align with new semantics based on target text sentences from source domain videos, thereby enhancing cross-domain generalization ability for video moment retrieval in the absence of target video data.
GenesisTex2: Stable, Consistent and High-Quality Text-to-Texture Generation
Jiawei Lu (Zhejiang University), Tianjia Shao (University College London)
GenerationData SynthesisDiffusion modelMesh
🎯 What it does: This paper proposes a text-to-texture generation framework based on a pre-trained stable diffusion model, capable of generating consistent and high-quality textures for 3D meshes without additional training.
GenHMR: Generative Human Mesh Recovery
Muhammad Usama Saleem (University of North Carolina), Chen Chen (University of Central Florida)
GenerationPose EstimationTransformerImageMesh
🎯 What it does: A method for single image human mesh recovery based on generative models, GenHMR, is proposed, utilizing pose tokenization and image-conditioned masked transformers to achieve uncertainty modeling and iterative sampling.
Genomics Data Lossless Compression with (S, K)-Mer Encoding and Deep Neural Networks
Hui Sun (Nankai University), Xiaoguang Liu (Nankai University)
CompressionRecurrent Neural NetworkTransformerBiomedical Data
🎯 What it does: A lossless compression framework for genomic data called DeepGeCo with three modes (MINI, PLUS, ULTRA) is proposed, balancing compression ratio, throughput, and memory usage;
GenPlan: Generative Sequence Models as Adaptive Planners
Akash Karthikeyan (University of Waterloo), Yash Vardhan Pant (University of Waterloo)
OptimizationRobotic IntelligenceReinforcement Learning from Human FeedbackTransformerFlow-based ModelSequential
🎯 What it does: A generative planning method called GenPlan based on a discrete flow model is proposed, utilizing an energy-guided denoising process to achieve multi-task adaptive planning.
GENTEEL-NEGOTIATOR: LLM-Enhanced Mixture-of-Expert-Based Reinforcement Learning Approach for Polite Negotiation Dialogue
Priyanshu Priya (Indian Institute of Technology Patna), Asif Ekbal (Indian Institute of Technology Patna)
TransformerLarge Language ModelReinforcement LearningMixture of ExpertsText
🎯 What it does: This study proposes a GENTEEL-NEGOTIATOR framework based on LLM-enhanced hybrid expert reinforcement learning for generating polite and goal-oriented negotiation dialogues.
GeoAggregator: An Efficient Transformer Model for Geo-Spatial Tabular Data
Rui Deng (University of Glasgow), Mingshu Wang (Florida State University)
TransformerTabular
🎯 What it does: A lightweight Transformer architecture called GeoAggregator is proposed for regression of geospatial tabular data.
GeoBEV: Learning Geometric BEV Representation for Multi-view 3D Object Detection
Jinqing Zhang (Beihang University), Yunhong Wang (Beihang University)
Object DetectionAutonomous DrivingTransformerImagePoint Cloud
🎯 What it does: This paper proposes GeoBEV, a 3D object detection framework based on multi-view images, achieving finer geometric recovery through high-resolution BEV representation.
Geodesic Flow Kernels for Semi-Supervised Learning on Mixed-Variable Tabular Dataset
Yoontae Hwang (University of Oxford), Yongjae Lee (Ulsan National Institute of Science and Technology)
ClassificationRepresentation LearningTransformerTabular
🎯 What it does: Aiming at semi-supervised learning for mixed-type tabular data, the GFTab framework is proposed to learn robust representations and perform classification from a limited number of labeled samples and a large number of unlabeled samples.
Geolocation Representation from Large Language Models Are Generic Enhancers for Spatio-Temporal Learning
Junlin He (Hong Kong Polytechnic University), Wei Ma (Hong Kong Polytechnic University)
Representation LearningData-Centric LearningGraph Neural NetworkTransformerLarge Language ModelPrompt EngineeringTabularTime Series
🎯 What it does: A training-independent geographic location embedding method based on large language models and OpenStreetMap, called LLMGeovec, is proposed and applied as a general enhancer for geographic prediction (GP), long-term sequence forecasting (LTSF), and graph-based spatiotemporal forecasting (GSTF).
GeoMamba: Towards Multi-granular POI Recommendation with Geographical State Space Model
Yifang Qin (Peking University), Ming Zhang (Peking University)
Recommendation SystemGraph Neural NetworkGraph
🎯 What it does: This paper proposes the GeoMamba model for next point of interest recommendation based on geographic information and multi-scale state space.
Geometry-Aware 3D Salient Object Detection Network
Chen Wang (Northwestern Polytechnical University), Yuchao Dai (Northwestern Polytechnical University)
Object DetectionSegmentationConvolutional Neural NetworkPoint Cloud
🎯 What it does: By introducing a superpoint segmentation and geometric enhancement module, a Geometry-aware 3D salient object detection network is constructed for salient object segmentation in point clouds.
GeoPro-Net: Learning Interpretable Spatiotemporal Prediction Models Through Statistically-Guided Geo-Prototyping
Bang An (Georgia Institute of Technology), Jun Luo (Fudan University)
Explainability and InterpretabilityConvolutional Neural NetworkSupervised Fine-TuningTabularTime Series
🎯 What it does: In urban event prediction, GeoPro-Net is proposed, which extracts Geo-concepts through statistical significance testing and combines it with prototype learning to achieve interpretable spatiotemporal event prediction.
Germane Conflicts: Desirable Properties for Localising Inconsistency
Glauber de Bona (Polytechnic School of the University of São Paulo), Anthony Hunter (University College London)
🎯 What it does: A set of desirable properties for evaluating inconsistency localization in knowledge bases is proposed, and a new concept of 'alternative conflict' is defined based on these properties.
GFlow: Recovering 4D World from Monocular Video
Shizun Wang (National University of Singapore), Xinchao Wang (National University of Singapore)
Object TrackingSegmentationDepth EstimationOptimizationGaussian SplattingOptical FlowVideo
🎯 What it does: This paper proposes the GFlow framework, which simultaneously recovers 4D (3D + time) scenes and camera trajectories from monocular uncalibrated videos, and supports tracking, segmentation, editing, and novel view rendering.
GGS: Generalizable Gaussian Splatting for Lane Switching in Autonomous Driving
Huasong Han (Wuhan University), Chunxia Xiao (Wuhan University)
GenerationData SynthesisAutonomous DrivingDiffusion modelGaussian SplattingImage
🎯 What it does: A 3D Gaussian Spray (GGS) framework based on MVS fusion is proposed, utilizing virtual lane generation, diffusion loss, and depth refinement to achieve high-quality lane switching and new perspective synthesis under single-lane data.
Ghidorah: Towards Robust Multi-Scale Information Diffusion Prediction via Test-Time Training
Wenting Zhu (Beijing University of Posts and Telecommunications), Xi Zhang (Beijing University of Posts and Telecommunications)
Domain AdaptationRecommendation SystemRecurrent Neural NetworkGraph Neural NetworkSupervised Fine-TuningContrastive LearningGraphTime Series
🎯 What it does: This study investigates the out-of-distribution (OOD) problem in information diffusion prediction and proposes the Ghidorah model based on test-time training. During the testing phase, a self-supervised task dynamically fine-tunes the feature extractor to achieve multi-scale predictions at both macro and micro levels.
GHOST: Gaussian Hypothesis Open-Set Technique
Ryan Rabinowitz (University of Colorado), Terrance E. Boult (University of Colorado)
ClassificationRecognitionAnomaly DetectionConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a post-processing algorithm called GHOST, which models each class feature using a multivariate Gaussian (diagonal covariance) and performs Z-score normalization on logits to distinguish between known and unknown samples, thereby enhancing open-set recognition and OOD detection.
GigaGS: 3D Gaussian Based Planar Representation for Large-Scene Surface Reconstruction
Junyi Chen (Shanghai Jiao Tong University), Tong He (Shanghai Artificial Intelligence Laboratory)
Gaussian SplattingPoint CloudMesh
🎯 What it does: GigaGS is proposed for large-scale scene surface reconstruction using 3D Gaussian Splatting.
GIM: A Million-scale Benchmark for Generative Image Manipulation Detection and Localization
Yirui Chen (Shanghai Jiao Tong University), Jie Hu (Huawei Noah's Ark Lab)
Image TranslationAnomaly DetectionTransformerDiffusion modelImageBenchmark
🎯 What it does: This work constructs a large-scale GIM dataset and evaluation benchmark for the detection and localization of generative image tampering;
GLAD: Improving Latent Graph Generative Modeling with Simple Quantization
Van Khoa Nguyen (Geneva School for Business Administration), Alexandros Kalousis (University of Geneva)
GenerationData SynthesisDrug DiscoveryGraph Neural NetworkTransformerDiffusion modelGraphStochastic Differential Equation
🎯 What it does: Designed and implemented a model GLAD for graph generation in a discrete latent space.
GLAM: Global-Local Variation Awareness in Mamba-based World Model
Qian He (Chinese Academy of Sciences), Jiandong Tian (Chinese Academy of Sciences)
Convolutional Neural NetworkReinforcement LearningWorld ModelVideoBenchmark
🎯 What it does: The GLAM (Global-Local variation Awareness Mamba-based world model) framework is proposed, utilizing two parallel Mamba modules, GMamba and LMamba, to perceive state changes at both global and local levels, thereby achieving high-quality world model inference and sample-efficient model-based reinforcement learning.
GLCF: A Global-Local Multimodal Coherence Analysis Framework for Talking Face Generation Detection
Xiaocan Chen (Sun Yat-sen University), Jiantao Zhou (University of Macau)
GenerationAnomaly DetectionConvolutional Neural NetworkVideoMultimodalityAudio
🎯 What it does: A large-scale multi-scene speech-driven speech generation (MSTF) dataset was constructed, and a Global-Local Multi-Modal Consistency Analysis Framework (GLCF) was proposed to detect lip-synced videos generated by speech. The framework includes three key modules: RSFDM, DCTAM, and V-AFM.
GLEN: Generalized Focal Loss Ensemble of Low-Rank Networks for Calibrated Visual Question Answering
Mahsa Mozaffari (Rochester Institute of Technology), Qi Yu (Rochester Institute of Technology)
ClassificationRecognitionImageMultimodality
🎯 What it does: To address the issue of overconfidence in visual question answering models, a low-rank decomposition of the final layer is proposed, and a diversified low-rank network ensemble is constructed through generalized focal loss to enhance model calibration.
GLIC: General Format Learned Image Compression
MingSheng Zhou (XiHua University), MingMing Kong (XiHua University)
CompressionAuto EncoderImage
🎯 What it does: A general format learning-based image compression model GLIC is designed, which can unify any channel image into a single channel for efficient compression, and proposes an adaptive attention residual block and a uniform grouping cross-channel context module to achieve high-quality progressive preview.
Global Attribute-Association Pattern Aggregation for Graph Fraud Detection
Mingjiang Duan (Zhejiang University), Zunlei Feng (Zhejiang University)
Anomaly DetectionGraph Neural NetworkGraphFinance Related
🎯 What it does: A graph fraud detection framework based on Attribute-Association Pattern Aggregation (GAAP) is proposed, which extracts attribute patterns using dynamic binning embedding, obtains association patterns through GraphSAGE, and aggregates global graph patterns using global cross-attention, ultimately improving detection performance.
Global Graph Propagation with Hierarchical Information Transfer for Incomplete Contrastive Multi-view Clustering
Guoqing Chao (Harbin Institute of Technology), Yongyong Chen (Ningbo University)
ClassificationOptimizationGraph Neural NetworkAuto EncoderContrastive LearningGraph
🎯 What it does: An end-to-end GHICMC framework is proposed, utilizing global graph propagation and hierarchical information transfer to achieve incomplete multi-view clustering.
Global-Semantic Alignment Distillation for Partial Multi-view Classification
Xiaoli Wang (Nanjing University of Science and Technology), Jun Liu (Lancaster University)
ClassificationKnowledge DistillationTransformerImage
🎯 What it does: This paper proposes a global semantic alignment distillation framework (GLAD) without interpolation, aimed at addressing the partial multi-view classification (PMvC) problem.
GlyphDraw2: Automatic Generation of Complex Glyph Posters with Diffusion Models and Large Language Models
Jian Ma (OPPO AI Center), Zhenyu Yang (OPPO AI Center)
GenerationData SynthesisTransformerLarge Language ModelDiffusion modelImageTextMultimodalityBenchmark
🎯 What it does: The GlyphDraw2 framework is proposed, which enables the automatic generation of posters using large language models and achieves precise text rendering.
GlyphSR: A Simple Glyph-Aware Framework for Scene Text Image Super-Resolution
Baole Wei (Peking University), Zhi Tang (Peking University)
RecognitionSuper ResolutionRecurrent Neural NetworkTransformerImage
🎯 What it does: A Glyph-aware scene text image super-resolution framework, GlyphSR, is proposed, significantly improving the readability of low-resolution text images and the accuracy of subsequent text recognition.
GMAP: Generalized Manipulation of Articulated Objects in Robotic Using Pre-trained Model
Hongliang Zeng (South China University of Technology), Jiahua Wang (South China University of Technology)
SegmentationRobotic IntelligenceTransformerSupervised Fine-TuningPoint Cloud
🎯 What it does: The GMAP framework is proposed to achieve a full process from instruction to perception to manipulation, used for the segmentation of joint objects, estimation of joint parameters, and prediction of interactive usability for robotic arms.
GNN-Transformer Cooperative Architecture for Trustworthy Graph Contrastive Learning
Jianqing Liang (Shanxi University), Jiye Liang (Shanxi University)
ClassificationRepresentation LearningGraph Neural NetworkTransformerContrastive LearningGraph
🎯 What it does: This paper proposes a new graph contrastive learning framework GTCA, which combines GCN and NodeFormer as two encoders, and generates reliable positive and negative samples through non-random augmentation and topology-based k-NN intersection, constructing a multi-positive sample contrastive loss.
GNN-Transformer Task Planning Enhanced with Semantic-Driven Data Augmentation
Soojin Jeong (Hanyang University), Yoonseon Oh (Hanyang University)
Robotic IntelligenceGraph Neural NetworkTransformerLarge Language ModelTextGraph
🎯 What it does: A high-level task planner based on Graph Neural Networks and Transformer (GTTP) has been designed and implemented, capable of predicting robot sub-goals based on natural language instructions, and its executability has been validated on both simulated and real robots.
GNS: Solving Plane Geometry Problems by Neural-Symbolic Reasoning with Multi-Modal LLMs
Maizhen Ning (Xi'an Jiaotong-Liverpool University), Kaizhu Huang (University of Liverpool)
TransformerLarge Language ModelPrompt EngineeringTextMultimodalityBenchmark
🎯 What it does: The GNS (Geometry Neural-Symbolic) framework is proposed, utilizing multimodal LLMs for knowledge prediction, symbolic analysis, reasoning, and symbolic computation of plane geometry problems, ultimately obtaining answers through a symbolic solver.
Goal-Driven Reasoning in DatalogMTL with Magic Sets
Shaoyu Wang (Shanghai Jiao Tong University), Pan Hu
Time Series
🎯 What it does: A method is proposed to extend magic set rewriting techniques to DatalogMTL, achieving goal-driven reasoning.
GoBERT: Gene Ontology Graph Informed BERT for Universal Gene Function Prediction
Yuwei Miao (University of Texas at Arlington), Junzhou Huang (University of Texas at Arlington)
TransformerLarge Language ModelGraphBiomedical Data
🎯 What it does: We propose GoBERT, a model that combines the Gene Ontology graph structure with BERT to predict unknown gene functions based on known gene functions.
GoHD: Gaze-oriented and Highly Disentangled Portrait Animation with Rhythmic Poses and Realistic Expressions
Ziqi Zhou (Institute of Automation, Chinese Academy of Sciences), Dong-Ming Yan (State Key Laboratory of Virtual Reality Technology and Systems, Beihang University)
GenerationData SynthesisPose EstimationRecurrent Neural NetworkTransformerDiffusion modelVideoAudio
🎯 What it does: The GoHD framework is proposed, which implements audio-based facial animation capable of generating high-quality, expressive, and controllable speaker videos from any reference identity.
GPU-Accelerated Parallel Bilevel Optimization for Roubst 6G ISAC
Xingdi Chen (Tongji University), Kai Yang (Tongji University)
Optimization
🎯 What it does: To address the channel estimation error problem in 6G ISAC systems, a GPU-accelerated bi-level optimization framework is proposed for robust beamforming design, along with an algorithm BOBLRBF that can generate the Pareto front and its deep learning implementation BOBLRBF-DNN.
Gradient Alignment Improves Test-Time Adaptation for Medical Image Segmentation
Ziyang Chen (Northwestern Polytechnical University), Yong Xia (Northwestern Polytechnical University)
SegmentationDomain AdaptationConvolutional Neural NetworkImageBiomedical Data
🎯 What it does: This paper proposes a test-time adaptation (TTA) method named GraTa, which utilizes gradient alignment and dynamic learning rates to enhance the performance of medical image segmentation models under domain shifts.
Gradient Weight-normalized Low-rank Projection for Efficient LLM Training
Jia-Hong Huang (University of Amsterdam), Evangelos Kanoulas (University of Amsterdam)
Computational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: A method called GradNormLoRP is proposed, which combines weight normalization and low-rank gradient projection to enhance the parameter and memory efficiency of LLM training.
Gradient-Based Nonlinear Rehearsal Learning with Multivariate Alterations
Tian Qin (Nanjing University), Zhi-Hua Zhou (Nanjing University)
OptimizationFlow-based ModelTabular
🎯 What it does: This paper proposes Grad-Rh, a gradient-based practice learning method for finding decision solutions under multivariate, nonlinear, and non-Gaussian structured practice models (SRM).
Gradient-Based Sample Selection for Black-Box Universal Domain Adaptation
Qiuyan He (Peking University), Minghua Deng (Peking University)
Domain AdaptationKnowledge DistillationImage
🎯 What it does: A Gradient-Based Sample Selection (GSS) method is proposed for black-box universal domain adaptation (B-UniDA), which derives sample selection metrics using gradient descent and Bayes' theorem, constructs an open set classifier, and effectively distinguishes and learns unknown and known classes through modules such as self-supervised clustering, knowledge distillation, and balancing mechanisms.
Gradient-Guided Credit Assignment and Joint Optimization for Dependency-Aware Spatial Crowdsourcing
Yafei Li (Zhengzhou University), Mingliang Xu (Zhengzhou University)
Recommendation SystemOptimizationGraph Neural NetworkReinforcement LearningGraphTabular
🎯 What it does: A two-stage recommendation and matching optimization framework (RMO) is proposed to address the spatial crowdsourcing problem with subtask dependencies;
GradQ-ViT: Robust and Efficient Gradient Quantization for Vision Transformers
Dahun Choi (Seoul National University of Science and Technology), Hyun Kim (Seoul National University of Science and Technology)
OptimizationComputational EfficiencyTransformerImage
🎯 What it does: A gradient quantization framework GradQ-ViT for Vision Transformers is proposed, which enables stable end-to-end training at INT8 low precision.
GRAIN: Multi-Granular and Implicit Information Aggregation Graph Neural Network for Heterophilous Graphs
Songwei Zhao (Jilin University), Hechang Chen (Jilin University)
Representation LearningGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: A new graph neural network model GRAIN is proposed for heterophilous graphs, which can adaptively aggregate multi-granularity information and consider the implicit relationships of distant nodes, thereby generating smoother and more accurate node embeddings.