AAAI 2026 Papers — Page 16
AAAI Conference on Artificial Intelligence · 4149 papers
From Single to Societal: Analyzing Persona-Induced Bias in Multi-Agent Interactions
Jiayi Li (Peking University), Yansong Feng (Peking University)
Explainability and InterpretabilityTransformerLarge Language ModelAgentic AIPrompt EngineeringText
🎯 What it does: This paper systematically investigates the biases arising from assigning individual personalities in multi-agent interactions of large language models (LLMs), exploring differences in social traits such as trustworthiness and persistence.
From Solver to Tutor: Evaluating the Pedagogical Intelligence of LLMs with KMP-Bench
Weikang Shi (Chinese University of Hong Kong), Hongsheng Li (Chinese University of Hong Kong)
TransformerLarge Language ModelSupervised Fine-TuningTextSequentialBenchmark
🎯 What it does: Constructed the KMP-Bench evaluation benchmark, comprising two modules: KMP-Dialogue and KMP-Skills, to assess the teaching capabilities of LLMs in K-8 mathematics education.
From Static to Active: Knowledge-Aware Node State Selection in Multi-view Graph Learning
Weiran Liao (Fuzhou University), Shiping Wang (Fuzhou University)
Representation LearningGraph Neural NetworkMultimodalityGraphOrdinary Differential Equation
🎯 What it does: Proposed a Knowledge-Aware Multi-View State Space Model (KAMSSM), enabling nodes to adaptively select activated or static sequences to achieve dynamic information exchange at the node level and directed diffusion across views.
From Stimuli to Minds: Enhancing Psychological Reasoning in LLMs via Bilateral Reinforcement Learning
Yichao Feng (Nanyang Technological University), Anh Tuan Luu (Nanyang Technological University)
Reinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningText
🎯 What it does: Constructed the StimuliQA dataset containing real psychological scenarios and proposed the Psy-Interpreter bilateral reinforcement learning framework (Trajectory Cache + T-GRPO + Bilateral Reward), evaluated on multiple psychological reasoning benchmarks, and achieved continual learning based on self-labeling.
From Subtle to Significant: Prompt-Driven Self-Improving Optimization in Test-Time Graph OOD Detection
Luzhi Wang (Dalian Maritime University), Hongbo Liu (Dalian Maritime University)
Anomaly DetectionOptimizationPrompt EngineeringGraphBiomedical Data
🎯 What it does: Proposed SIGOOD, a graph OOD detection framework that self-improves during testing.
From Text to Simulation: A Multi-Agent LLM Workflow for Automated Chemical Process Design
Xufei Tian (East China University of Science and Technology), Ke Ye (East China University of Science and Technology)
OptimizationTransformerLarge Language ModelAgentic AITextChain-of-Thought
🎯 What it does: Propose a workflow based on a multi-agent large language model (LLM) that automatically converts natural language chemical process descriptions into complete configuration files executable in professional simulation software, achieving closed-loop iterative optimization during this process.
From Tokens to Latent States: Leveraging Pre-trained Language Models for Improving Partially Observable Reinforcement Learning
Meiju Li (Beijing Institute of Technology), Mingzhong Wang (University of the Sunshine Coast)
Convolutional Neural NetworkRecurrent Neural NetworkTransformerLarge Language ModelReinforcement LearningImageText
🎯 What it does: Propose to utilize pre-trained large language models (LLMs) to estimate hidden states in POMDPs and employ them as memory modules in reinforcement learning (RL).
FT-MoE: Sustainable-learning Mixture of Experts for Fault-Tolerant Computing
Wenjing Xiao (Guangxi University), Min Chen (South China University of Technology)
Anomaly DetectionComputational EfficiencyMixture of ExpertsTime Series
🎯 What it does: Constructed the FT-MoE framework, implementing a dual-path hybrid expert network for edge fault detection and classification, and achieving continual learning through offline training + online fine-tuning.
FT-NCFM: An Influence-Aware Data Distillation Framework for Efficient VLA Models
Kewei Chen (Chongqing Institute of Green and Intelligent Technology, Chinese Academy of Sciences), Mingsheng Shang (University of Oulu)
Computational EfficiencyKnowledge DistillationTransformerVision-Language-Action ModelGenerative Adversarial NetworkMultimodality
🎯 What it does: Propose a self-contained evaluation engine FT and an influence-weight-driven generative neural characteristic function matching (NCFM) framework FT-NCFM, which synthesizes information-dense core datasets, significantly reducing the data volume and time required for VLA training.
Full-Atom Peptide Design via Riemannian–Euclidean Bayesian Flow Networks
Hao Qian, Lei Xu (Shanghai Jiao Tong University)
GenerationFlow-based ModelBiomedical Data
🎯 What it does: Propose a full-atom peptide design framework based on Bayesian flow networks (PepBFN), jointly modeling amino acid types, residue orientations, centroid coordinates, and side-chain torsion angles to generate peptide chains in a fully continuous parameter space.
Function-on-Function Bayesian Optimization
Jingru Huang (Tsinghua University), Chen Zhang (Tsinghua University)
OptimizationBiomedical Data
🎯 What it does: Proposes a Bayesian optimization framework (FFBO) for scenarios where both inputs and outputs are functions, constructing a probabilistic model through function-to-function Gaussian processes (FFGP);
FUSE: Fine-Grained and Semantic-Aware Learning for Unified Image Understanding and Generation
Peng Zhang (Zhejiang University), Hao Jiang (Alibaba Group)
GenerationTransformerLarge Language ModelVision Language ModelDiffusion modelFlow-based ModelAuto EncoderMultimodalityBenchmark
🎯 What it does: Propose a unified FUSE framework that can simultaneously perform multimodal understanding and high-quality image generation and editing.
FusedRec: Fused Embedding Communication for Distributed Recommendation Training on GPUs
Xuanteng Huang (Sun Yat-sen University), Xianwei Zhang (Sun Yat-sen University)
Recommendation SystemTabular
🎯 What it does: For embedding communication in distributed deep learning recommendation models, we propose FUSEDREC: fusing multi-class embeddings, deduplication, delayed hashing, and recovery mechanisms to achieve single AlltoAll communication while preserving category information.
FuseMine: Robust Multi-Modal Compound-Protein Interaction Prediction via Differential Attention Feature Mining
Junlin Xu (Wuhan University of Science and Technology), Yajie Meng (Wuhan Textile University)
Drug DiscoveryConvolutional Neural NetworkGraph Neural NetworkTransformerLarge Language ModelMultimodalityBiomedical Data
🎯 What it does: Propose the FuseMine framework, which constructs a multi-modal complex-protein interaction prediction model by jointly encoding molecular structures and sequences through graph neural networks, convolutional networks, and pre-trained language models.
FUSION: Dataset Pruning via Fusing Uncertainty with Structural Information for Optimal Neural Training in Crystal Property Prediction
Xiean Wang (Sun Yat-sen University), Qingsong Zou (Sun Yat-sen University)
OptimizationTabularBenchmarkPhysics Related
🎯 What it does: This paper proposes FUSION, an offline dataset trimming strategy that optimizes the training set for material property prediction by jointly combining uncertainty quantification and crystal structure geometric fingerprints.
FVNet: Harnessing Liquid Neural Dynamics for Lightweight Visual Representation
Zhenzhe Hou (Beijing Institute of Technology), Yutao Liu (Beijing Institute of Technology)
ClassificationObject DetectionSegmentationTransformerImageOrdinary Differential Equation
🎯 What it does: Proposed a lightweight visual backbone network called FVNet, which integrates the continuous-time dynamics of liquid neural networks into visual feature extraction to achieve adaptive spatiotemporal feature encoding.
G-IR: Geometric Image Representation for Learning
Xin Chen (Xi'an Jiaotong University), Zongben Xu (Xi'an Jiaotong University)
RestorationRepresentation LearningAuto EncoderImage
🎯 What it does: This paper proposes an image representation method called G-IR based on geometric optimal transport and quasi-conformal mapping, and combines it with an autoencoder to achieve image restoration and interpolation.
G-UBS: Towards Robust Understanding of Implicit Feedback via Group-Aware User Behavior Simulation
Boyu Chen (Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences), Yali Wang (Tencent)
Recommendation SystemTransformerLarge Language ModelReinforcement LearningVideoTextSequential
🎯 What it does: Proposed a behavior simulation framework called G-UBS based on user groups to achieve individual preference understanding in environments with noisy implicit feedback
GAGS: Granularity-Aware Feature Distillation for Language Gaussian Splatting
Yuning Peng (Wuhan University), Bisheng Yang (Wuhan University)
SegmentationKnowledge DistillationRepresentation LearningVision Language ModelGaussian SplattingImage
🎯 What it does: Proposes the GAGS framework, distilling 2D CLIP features into a 3D Gaussian splat model to achieve out-of-the-box multi-view semantic queries;
GAHMN: A Generative Approach for High-Dimensional Mediation Analysis
Jiaming Zhang, Hanwen Ning (Chinese University of Hong Kong)
Explainability and InterpretabilityGenerative Adversarial NetworkTabularFinance Related
🎯 What it does: Proposes a high-dimensional mediation analysis framework GAHMN based on generative adversarial networks (GANs), which simultaneously models the causal relationships between mediator variables and outcomes.
Gait Recognition via Collaborating Discriminative and Generative Diffusion Models
Haijun Xiong (Huazhong University of Science and Technology), Wenyu Liu (Huazhong University of Science and Technology)
RecognitionConvolutional Neural NetworkDiffusion modelVideo
🎯 What it does: Propose the CoD 2 framework, which jointly trains generative diffusion models with discriminative models to extract more robust gait features.
Gait Transformer: End-to-End Transformer Backbone for Gait Recognition
Saihui Hou (Beijing Normal University), Yongzhen Huang (Institute Of Automation Chinese Academy Of Sciences)
RecognitionTransformerVideo
🎯 What it does: Proposes GaT (Gait Transformer), an end-to-end Transformer backbone network specifically designed for gait recognition based on binary silhouette contours.
Game Ground Bench: Probing the Limits of LVLMs in Complex Semantic Grounding Across Game Universes
Zhangyang Qi (University of Hong Kong), Hengshuang Zhao (University of Hong Kong)
Reinforcement LearningVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Proposed a diverse benchmark called GGBench spanning 10 game categories and designed an extremely low-sample cross-game localization method named Game‑R1;
GARNET: GoT-Based Alert Reduction and Narrative Event Tracing
Yiru Gong (Chinese Academy of Sciences), Zhigang Lu (Chinese Academy of Sciences)
Anomaly DetectionExplainability and InterpretabilityKnowledge DistillationRepresentation LearningGraph Neural NetworkTransformerLarge Language ModelContrastive LearningTextGraphChain-of-Thought
🎯 What it does: Proposes the GARNET framework, which leverages large language models (LLMs) to reason over security incident alert correlation graphs, automatically correlating alerts and generating concise, readable attack path summaries, thereby significantly reducing false positives.
GATCL: An Adaptive Contrastive Learning Framework Based on MHGAT for Spatial Domain Identification in Spatial Transcriptomics
Shilin Zhang, Xiulong Liu (Tianjin University)
ClassificationGraph Neural NetworkContrastive LearningBiomedical Data
🎯 What it does: Propose an adaptive contrastive learning framework called GATCL based on a multi-head graph attention network for spatial domain identification in spatial transcriptomics.
Gated Variational Graph Autoencoders as Experts with Competition and Consensus for Multi-view Clustering
Zhaoliang Chen (Hong Kong Baptist University), Jiming Liu (Hong Kong Baptist University)
Representation LearningGraph Neural NetworkMixture of ExpertsAuto EncoderImageTextTime Series
🎯 What it does: Propose a multi-view clustering framework GVGAE-C based on a gated variational graph autoencoder (VGAE), which utilizes one-to-one corresponding view-specific experts and a structure-aware gating network to achieve sample-level weight allocation, and integrates view-specific and shared information through expert competition and consensus mechanisms.
GateRA: Token-aware Modulation for Parameter-Efficient Fine-tuning
Jie Ou (Yuanzhigu Technology Co., Ltd.), Cees G. M. Snoek (University of Amsterdam)
Computational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose the GateRA framework, introducing token-level adaptive gating and entropy regularization in PEFT, activating low-rank updates only on important or uncertain tokens while preserving pre-trained knowledge;
Gaussian Approximation for Two-Timescale Linear Stochastic Approximation
Bogdan Butyrin (HSE University), Sergey Samsonov (HSE University)
Optimization
🎯 What it does: This paper provides non-asymptotic Gaussian approximation (convex distance) error bounds for the last iteration and Polyak-Ruppert average iteration of two-time-scale stochastic approximation (TTSA) under Markov noise and Markov difference noise.
Gaussian Blending: Rethinking Alpha Blending in 3D Gaussian Splatting
Junseo Koo (Seoul National University), Gunhee Kim (Seoul National University)
Gaussian Splatting
🎯 What it does: Proposes Gaussian Blending as an alternative to traditional scalar alpha blending rendering method, using spatially distributed alpha and transmittance to address attenuation and expansion artifacts in 3D Gaussian Splatting at unseen sampling rates.
Gaussian Uncertainty-Driven Multi-Model Fitting with Graph Neural Network
Ligang Zhang (Quanzhou Institute of Equipment Manufacturing, Haixi Institute, Chinese Academy of Sciences), Qiming Li (Quanzhou Institute of Equipment Manufacturing, Haixi Institute, Chinese Academy of Sciences)
OptimizationGraph Neural NetworkImage
🎯 What it does: This paper proposes a multi-model fitting framework driven by Gaussian uncertainty, which constructs local uncertainty distributions by analyzing the covariance propagation of the Jacobian matrix; and designs a Gaussian Hypothesis Generation Network (GHG-Net) to learn global parameter distributions, combining dynamic graph neural networks and multi-head attention mechanisms to capture spatial relationships between observations; subsequently, the final model is extracted using Mahalanobis distance clustering.
GaussianImage++: Boosted Image Representation and Compression with 2D Gaussian Splatting
Tiantian Li, Yan Wang (Tsinghua University)
CompressionRepresentation LearningGaussian SplattingImage
🎯 What it does: This paper proposes GaussianImage++, a method that enhances image representation and compression performance through 2D Gaussian splatting.
GazeInterpreter: Parsing Eye Gaze to Generate Eye-Body-Coordinated Narrations
Qing Chang (Hong Kong University of Science and Technology), Zhiming Hu (Hong Kong University of Science and Technology)
GenerationTransformerLarge Language ModelVision-Language-Action ModelVideoTextMultimodality
🎯 What it does: Propose a GazeInterpreter framework based on LLM that parses eye movement signals into symbolic events, fuses them with body motion narratives to generate eye-body coordinated narratives, and enhances semantic consistency through self-correcting loops.
GCA: Geometry-aware Conditional Alignment for Partial Domain Adaptation with Coding Rate Reduction
Xiaohui Chen (Sun Yat-Sen University), Chuan-Xian Ren (Sun Yat-Sen University)
Domain AdaptationConvolutional Neural NetworkImage
🎯 What it does: Proposed the Geometry-aware Conditional Alignment (GCA) framework, which integrates conditional alignment with geometric orthogonal discrimination to address partial domain adaptation problems.
GCIB: Causal Intervention Guided Graph Information Bottleneck Framework
Hangyuan Du (Shanxi University), Wenjian Wang (Shanxi University)
Representation LearningDrug DiscoveryGraph Neural NetworkGraphBenchmark
🎯 What it does: Proposed a causal intervention-based graph information bottleneck framework (GCIB), combining subgraph extraction, information bottleneck, and causal intervention to learn graph representations that compress while preserving causal information, thereby enhancing out-of-distribution (OOD) generalization in graph classification tasks.
GCL-OT: Graph Contrastive Learning with Optimal Transport for Heterophilic Text-Attributed Graphs
Yating Ren (Beihang University), Huobin Tan (Beihang University)
ClassificationRepresentation LearningGraph Neural NetworkTransformerLarge Language ModelPrompt EngineeringContrastive LearningTextGraph
🎯 What it does: Integrate large language models with graph neural networks to perform contrastive learning on text-attribute graphs, achieving alignment and fusion of structural and textual perspectives.
GDBA Revisited: Unleashing the Power of Guided Local Search for Distributed Constraint Optimization
Yanchen Deng (Nanyang Technological University), Bo An (Nanyang Technological University)
OptimizationGraphBenchmark
🎯 What it does: Proposed a new distributed guided local search framework called DGLS, addressing the reasons why the original GDBA algorithm performs poorly on general-value DCOPs, solving issues such as over-violation, infinite penalty accumulation, and uncoordinated updates.
GeM-VG: Towards Generalized Multi-image Visual Grounding with Multimodal Large Language Models
Shurong Zheng (Chinese Academy of Sciences), Jinqiao Wang (Chinese Academy of Sciences)
Object DetectionTransformerLarge Language ModelReinforcement LearningMultimodalityBenchmarkChain-of-Thought
🎯 What it does: Proposed GeM-VG, a multimodal large language model capable of performing general multi-image visual grounding;
GEM: A Scale-Aware and Distribution-Sensitive Sparse Fine-Tuning Framework for Effective Downstream Adaptation
Sungmin Kang (University of Southern California), Sunwoo Lee (Inha University)
Computational EfficiencyTransformerSupervised Fine-TuningText
🎯 What it does: Proposes a sparse fine-tuning framework named GEM, which selects parameters with the most impact on the model by utilizing the ratio of gradients to weights, and dynamically allocates the proportion of trainable parameters per layer through entropy, achieving efficient downstream adaptation while maintaining an extremely low parameter update ratio (e.g., 0.1%).
GEM: Gaussian Embedding Modeling for Out-of-Distribution Detection in GUI Agents
Zheng Wu (Shanghai Jiao Tong University), Zhuosheng Zhang (Shanghai Jiao Tong University)
Anomaly DetectionVision Language ModelMultimodality
🎯 What it does: This paper addresses the outlier detection problem for graphical user interface (GUI) agents by proposing a Gaussian Mixture Model (GEM) method based on input embedding distance, which is used to identify instructions that fall outside the training distribution;
GEMA-Score: Granular Explainable Multi-Agent Scoring Framework for Radiology Report Evaluation
Zhenxuan Zhang (Imperial College London), Guang Yang (Imperial College London)
Explainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelAgentic AITextBiomedical DataComputed Tomography
🎯 What it does: Proposes Granular Explainable Multi-Agent Score (GEMA-Score) — a fine-grained medical report evaluation framework based on multi-agent collaboration, capable of objectively measuring pathological entities, locations, severity, and uncertainty, while providing interpretable comprehensive scores through subjective expression assessment (completeness, readability, terminology standardization);
Gene Incremental Learning for Single-Cell Transcriptomics
Jiaxin Qi (Computer Network Information Center Chinese Academy Of Sciences), Gaogang Xie (Computer Network Information Center Chinese Academy Of Sciences)
Knowledge DistillationTransformerBiomedical DataBenchmark
🎯 What it does: Proposed a Gene Incremental Learning (GIL) framework in single-cell transcriptomics, defining a phased gene learning process, baseline methods, data replay and knowledge distillation for transfer learning, as well as two evaluation metrics (gene regression and gene classification).
GenePheno: Interpretable Gene Knockout-Induced Phenotype Abnormality Prediction from Gene Sequences
Jingquan Yan (University of Texas at Arlington), Junzhou Huang (University of Texas Southwestern Medical Center)
Drug DiscoveryTransformerContrastive LearningBiomedical Data
🎯 What it does: Proposes GenePheno, an interpretable multi-label prediction framework that directly utilizes gene sequences to predict phenotypic abnormalities caused by knockouts.
Generalising Traffic Forecasting to Regions Without Traffic Observations
Xinyu Su (University of Melbourne), Jianzhong Qi (University of Melbourne)
Autonomous DrivingConvolutional Neural NetworkGraph Neural NetworkContrastive LearningTime SeriesSequential
🎯 What it does: Propose a traffic prediction model called GenCast, which can perform high-precision predictions in areas lacking sensor observations.
Generalizable Drug–Target Interaction Prediction via ESM-2 Representations and Progressive Contrastive Curriculum Learning
Qianyang Wu (Hainan University), Feifei Cui (Hainan University)
Drug DiscoveryTransformerLarge Language ModelContrastive LearningBiomedical Data
🎯 What it does: Proposed the ESP-DTI framework, combining the ESM-2 protein language model, CLIP-style cross-modal alignment, and progressive adaptive curriculum learning to predict drug-target interactions.
Generalizable Heterogeneity-aware Federated Feature and Basic-matrix Consistency Learning
Xuan Lai (Fuzhou University), Zheyi Chen (Fuzhou University)
Federated LearningKnowledge DistillationRepresentation LearningImage
🎯 What it does: Propose the FBCL framework, which addresses heterogeneity and catastrophic forgetting in federated learning through instance-level feature alignment and feature basis matrix consistency using unlabeled public data.
Generalization Bounds for Semi-supervised Matrix Completion with Distributional Side Information
Antoine Ledent (Singapore Management University), Nong Minh Hieu (Singapore Management University)
Recommendation SystemTabular
🎯 What it does: Propose a semi-supervised matrix completion framework, assuming that the sampling distribution and the true matrix share a low-rank subspace, provide a theoretical upper bound on generalization error, and implement the corresponding DAMC algorithm.
Generalized Geometry Encoding Volume for Real-time Stereo Matching
Jiaxin Liu (Huazhong University of Science and Technology), Xin Yang (Huazhong University of Science and Technology)
Depth EstimationComputational EfficiencyConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: Propose a real-time stereo matching framework named GGEV, which achieves strong generalization for unseen scenes by fusing multi-scale texture features with depth priors from Depth Anything V2, and employs depth-aware dynamic cost aggregation.
Generalized Threshold Optimization with Harmony Multi-Threshold Neurons for Accurate ANN-to-SNN Conversion
Wenhan Zhang (Peking University), Zhaofei Yu (Peking University)
OptimizationComputational EfficiencySpiking Neural NetworkImage
🎯 What it does: Proposed a novel Harmonious Multi-Threshold Spiking Neuron (H-MT) and reduced imbalanced errors through a communication mechanism;
Generalized-Scale Object Counting with Gradual Query Aggregation
Jer Pelhan (University of Ljubljana), Matej Kristan (University of Ljubljana)
Object DetectionTransformerImage
🎯 What it does: Propose an end-to-end few-shot counting and detection framework called GECO2, which utilizes scale-specific query encoders and cross-scale aggregation to generate high-resolution global query maps for accurate counting and localization.
Generalizing Analogical Inference from Boolean to Continuous Domains
Francisco Cunha (University of Lisbon), Zied Bouraoui (University of Artois)
🎯 What it does: Redefine analogical reasoning, demonstrate the failure of existing generalization error bounds in the Boolean domain, and propose a parameterized analogical framework generalizable to the continuous domain (based on the generalized mean), providing a complete characterization of analogical preservation functions and corresponding error upper bounds.
Generalizing Fair Clustering to Multiple Groups: Algorithms and Applications
Diptarka Chakraborty (Nationaly University of Singapore), Tien-Long Nguyen (Pennsylvania State University)
Optimization
🎯 What it does: Studied the problem of nearest fair clustering given an arbitrary number of colors (multi-color) and applied the results to fair-related clustering and fair consensus clustering.
Generalizing Vision-Language Models with Dedicated Prompt Guidance
Xinyao Li (University of Electronic Science and Technology of China), Jingjing Li (University of Electronic Science and Technology of China)
Domain AdaptationTransformerPrompt EngineeringVision Language ModelContrastive LearningImageText
🎯 What it does: Propose a two-step domain expert-guided domain generalization framework called GuiDG, first learning the source domain expert through prompt tuning, and then fine-tuning the visual encoder guided by cross-modal attention;
Generating Attribute-Aware Human Motions from Textual Prompt
Xinghan Wang (Peking University), Yadong Mu (Peking University)
GenerationTransformerPrompt EngineeringVision Language ModelAuto EncoderTextSequential
🎯 What it does: This study proposes an AttrMoGen framework based on structural causal models, capable of generating 3D motion sequences that satisfy specified human attributes such as age and gender under text prompts.
Generating In-Distribution Counterfactual Explanation for Graph Neural Networks
Linmao Chen, Quanlong Guan (South China Normal University)
GenerationExplainability and InterpretabilityGraph Neural NetworkDiffusion modelAuto EncoderGraph
🎯 What it does: Propose a method called ICExplainer, which uses variational inference to obtain the true graph distribution as a prior, and then combines it with a graph diffusion model to generate counterfactual explanations that can alter GNN predictions while maintaining consistency with the original distribution.
Generating Risky Samples with Conformity Constraints via Diffusion Models
Han Yu (Tsinghua University), Peng Cui (Tsinghua University)
ClassificationGenerationData SynthesisDiffusion modelContrastive LearningImageMultimodality
🎯 What it does: This paper generates risky samples using diffusion models and introduces category consistency constraints to ensure the generated samples align with the desired categories.
Generating Sketches in a Hierarchical Auto-Regressive Process for Flexible Sketch Drawing Manipulation at Stroke-Level
Sicong Zang (Donghua University), Zhijun Fang (Donghua University)
GenerationRecurrent Neural NetworkTransformerImage
🎯 What it does: Propose a hierarchical autoregressive sketch generation process called Sketch-HARP, which allows flexible editing, deletion, or insertion of individual strokes during the drawing process, enabling fine-grained control over sketches.
Generating-Filtering-Ranking: A Three-Stage MultiModal Data Augmentation Framework Under Partial Modality Missing
Zhirui Kuai (Central South University), Li Kuang (Central South University)
Data SynthesisTransformerLarge Language ModelVision Language ModelDiffusion modelContrastive LearningMultimodality
🎯 What it does: Propose a three-stage multimodal data augmentation framework GFR (Generation-Filtering-Ranking) to complete missing modalities and enhance model performance under missing modalities.
Generative Branching for Mixed-Integer Linear Programming
Ruobing Wang, Mingzhong Wang (Beijing Institute Of Technology)
OptimizationComputational EfficiencyGraph Neural NetworkDiffusion modelTabular
🎯 What it does: This paper transforms the branch variable selection problem into a conditional generation task, generating branch score distributions using diffusion models, and achieving high-quality inference in one step through consistency learning, significantly improving branching efficiency and generality.
Generic Adversarial Attack Framework Against Graph-based Vertical Federated Learning
Yimin Liu (Beijing Institute of Technology), Liehuang Zhu (Beijing Institute of Technology)
Domain AdaptationFederated LearningAdversarial AttackGraph Neural NetworkGraph
🎯 What it does: Proposed a general adversarial attack framework SGAC for graph structure vertical federated learning, which can generate high-fidelity proxy models using cross-domain auxiliary graphs without acquiring server models or in-domain labeled data, and finally generate adversarial embeddings that dominate joint inference by constructing diverse shadow inputs through node attribute and edge significance.
GENMAC: Compositional Text-to-Video Generation with Multi-Agent Collaboration
Kaiyi Huang (University of Hong Kong), Xihui Liu (University of Hong Kong)
GenerationLarge Language ModelAgentic AIDiffusion modelVideoTextMultimodalityBenchmark
🎯 What it does: Proposed a multi-agent collaborative framework called GENMAC, supporting hierarchical generation, iterative optimization, and layout control for text-to-video tasks.
GenPRM: Scaling Test-Time Compute of Process Reward Models via Generative Reasoning
Jian Zhao (Tsinghua University), Bowen Zhou (Tsinghua University)
Computational EfficiencyReinforcement Learning from Human FeedbackTransformerSupervised Fine-TuningReinforcement LearningTextChain-of-Thought
🎯 What it does: This paper proposes GenPRM, a process reward model that enhances the reasoning quality of large language models through generative chain reasoning and code verification.
GenPTW: Latent Image Watermarking for Provenance Tracing and Tamper Localization
Zhenliang Gan (Fudan University), Xinpeng Zhang (Fudan University)
GenerationConvolutional Neural NetworkDiffusion modelAuto EncoderImage
🎯 What it does: Proposes the GenPTW framework, embedding watermarks in the latent space of generative models to enable copyright tracing and tamper localization during generation and post-processing.
Gentle Manipulation Policy Learning via Demonstrations from VLM Planned Atomic Skills
Jiayu Zhou (Hong Kong University of Science and Technology), Renjing Xu (Harbin Institute of Technology)
Knowledge DistillationTransformerReinforcement LearningVision Language ModelDiffusion modelTextMultimodalityPoint Cloud
🎯 What it does: This paper proposes a complete framework based on hierarchical semantic decomposition, reinforcement learning, vision-language models, and knowledge distillation to automatically generate demonstrations and learn long-term gentle manipulation strategies.
GenVidBench: A 6-Million Benchmark for AI-Generated Video Detection
Zhenliang Ni (Huawei), Yunhe Wang (Huawei)
ClassificationAnomaly DetectionConvolutional Neural NetworkTransformerVideoBenchmark
🎯 What it does: Created a video dataset named GenVidBench with a scale of 6.78 million videos, and conducted cross-source and cross-generator detection experiments on multiple video classification models.
Geo2Vec: Shape- and Distance-Aware Neural Representation of Geospatial Entities
Chen Chu (University of Southern California), Cyrus Shahabi (University of Southern California)
ClassificationRepresentation LearningImage
🎯 What it does: Proposes a unified spatial representation learning method called Geo2Vec, which directly encodes the shape and positional information of various geographic entities (points, lines, polygons, etc.) in the coordinate space by learning signed distance fields (SDF);
GeoBayes: Probabilistic Image Geo-Localization Inference via Sequential Bayesian Updating
Weimin Shi (Beihang University), Zhong Zhou (Beihang University)
RetrievalTransformerLarge Language ModelImageTextMultimodalityRetrieval-Augmented Generation
🎯 What it does: Propose the GeoBayes framework, which achieves image geolocation through a multi-round hypothesis-verification-update loop using sequential Bayesian inference;
GeoCoBox: Box-supervised 3D Tumor Segmentation via Geometric Co-embedding
Tianzhong Lan (Sichuan University), Min Zhu (Sichuan University)
SegmentationConvolutional Neural NetworkContrastive LearningBiomedical DataComputed Tomography
🎯 What it does: Propose GeoCoBox, which utilizes box supervision combined with anatomical priors and geometric contrastive embedding to achieve 3D tumor segmentation.
GeoGen: A Two-stage Coarse-to-Fine Framework for Fine-grained Synthetic Location-based Social Network Trajectory Generation
Rongchao Xu (Florida State University), Guang Wang (Florida State University)
Data SynthesisSafty and PrivacyTransformerDiffusion modelTime SeriesSequential
🎯 What it does: Proposed the GeoGen two-stage coarse-to-fine framework for generating high-fidelity, privacy-safe fine-grained LBSN check-in trajectories.
Geometric Correspondence Constrained Pseudo-Label Alignment for Source-Free Domain Adaptive Fundus Image Segmentation
Zhouhongyuan Hu (Sichuan University), Zhenbin Wang (Sichuan University)
SegmentationDomain AdaptationConvolutional Neural NetworkGenerative Adversarial NetworkBiomedical Data
🎯 What it does: In the retinal fundus image segmentation task under source-free unsupervised domain adaptation (SF-UDA), the Geometric Correspondence Constrained (GCC) framework is proposed. It first stratifies pseudo-labels by entropy for quality assessment, then aligns low-quality samples using geometric correspondence information from high-quality samples, and further corrects high-confidence noise through adaptive Gaussian perturbation (SAPE).
Geometry Meets Light: Leveraging Geometric Priors for Universal Photometric Stereo Under Limited Multi-Illumination Cues
King-Man Tam (Institute of Science Tokyo), Rei Kawakami (National Institute of Informatics)
Depth EstimationTransformerImage
🎯 What it does: Designed and implemented the GeoUniPS network by incorporating pre-trained 3D reconstruction models (e.g., VGGT) as geometric priors into a general photometric stereo framework, and constructed a synthetic dataset PS-Perp with perspective projection.
Geometry-Aware Noisy Correspondence Mitigation for Cross-Modal Text-Based Person Retrieval
Xinpan Yuan (Hunan University of Technology), Lin Yuanbo Wu (University of Warwick)
RetrievalVision Language ModelContrastive LearningMultimodality
🎯 What it does: Studied the noise correspondence problem in text-image retrieval, and proposed two modules, GSCA and SRAM, to enhance the model's robustness to noise.
Geometry-Aware Stereo Matching via Monocular Disparity Distribution Prior and Gradient Enhancement
Junze Zhang (Academy of Military Science), Chunping Qiu (Intelligent Game and Decision Lab (IGDL))
Depth EstimationAutonomous DrivingConvolutional Neural NetworkRecurrent Neural NetworkImage
🎯 What it does: Propose a stereo matching framework named GEAStereo, which combines monocular disparity distribution prior (MDPV) and gradient enhancement to construct Mono-Stereo Fusion Volume (MSFV) and Detail-Aware Volume (DAV), and iteratively optimizes the final disparity using GRU.
Geometry-Aware Variational Information Maximization for Deep Incomplete Multi-view Clustering
Wenlan Chen (Central South University), Cheng Liang (Shandong Normal University)
Representation LearningAuto EncoderContrastive LearningMultimodalityBenchmark
🎯 What it does: Propose a GAVIM framework that is imputation-free and based on variational autoencoders to complete incomplete multi-view clustering tasks.
GeoMoE: Divide-and-Conquer Motion Field Modeling with Mixture-of-Experts for Two-View Geometry
Jiajun Le (Wuhan University), Jiayi Ma (Wuhan University)
Pose EstimationGraph Neural NetworkMixture of ExpertsImagePoint Cloud
🎯 What it does: Propose GeoMoE, which utilizes Mixture-of-Experts to perform probabilistic prior-driven decomposition and subfield linearization of two-view motion fields, thereby achieving more accurate and robust motion field estimation in tasks such as relative pose, homography, and point cloud registration.
GeoNum: Bridging Numerical Continuity and Language Semantics via Geometric Embedding
Shengkai Jin, Jun Han (Beihang University)
Representation LearningLarge Language ModelSupervised Fine-TuningBenchmark
🎯 What it does: This paper proposes GeoNum, a continuous numerical embedding based on polar coordinate decomposition, to address discretization distortion when LLMs process numerical values.
GeoPTH: A Lightweight Approach to Category-Based Trajectory Retrieval via Geometric Prototype Trajectory Hashing
Yang Xu (Nanjing University), Kai Ming Ting (Nanjing University)
RetrievalComputational EfficiencyTime SeriesSequential
🎯 What it does: Proposed a lightweight, learning-free GeoPTH framework that achieves category-based trajectory retrieval using geometric prototype hashing.
GeoShield: Safeguarding Geolocation Privacy from Vision-Language Models via Adversarial Perturbations
Xinwei Liu (Institute of Information Engineering, Chinese Academy of Sciences), Xiaochun Cao (Northeastern University)
Safty and PrivacyAdversarial AttackVision Language ModelImage
🎯 What it does: Studies how to prevent VLMs from accurately predicting location information by adding adversarial perturbations to images, thereby protecting geographic privacy.
GeoX-Bench: Benchmarking Cross-View Geo-Localization and Pose Estimation Capabilities of Large Multimodal Models
Yushuo Zheng (Shanghai Jiao Tong University), Guangtao Zhai (Shanghai Jiao Tong University)
Pose EstimationTransformerSupervised Fine-TuningPrompt EngineeringVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Propose the GeoX-Bench benchmark to evaluate the capabilities of large multimodal models in cross-perspective geolocation and pose estimation
GEWDiff: Geometric Enhanced Wavelet-based Diffusion Model for Hyperspectral Image Super-resolution
Sirui Wang (Technical University of Munich), Xiao Xiang Zhu (Universitat Aut'noma de Barcelona)
Super ResolutionConvolutional Neural NetworkDiffusion modelImage
🎯 What it does: Propose GEWDiff, a diffusion model based on wavelet encoding and geometric enhancement, achieving four times super-resolution for hyperspectral image reconstruction.
GeWu: A Culturally-Grounded Chinese Benchmark for Multi-Stage Social Bias Evaluation in Large Language Models
Yi Lin (Southern University of Science and Technology), Xuetao Wei (Lingnan University)
TransformerLarge Language ModelTextBenchmark
🎯 What it does: Constructed the GeWu Chinese social bias assessment benchmark, containing 60,192 questions, and selected a high-bias subset of 1,000 questions named GeWu-1K.
Ghost in the Transformer: Detecting Model Reuse with Invariant Spectral Signatures
Suqing Wang (Wuhan University), Zuchao Li (Wuhan University)
Anomaly DetectionTransformerText
🎯 What it does: Propose the GhostSpec method, utilizing the singular value spectrum of the attention weight matrix within Transformers as an immutable fingerprint of the model's origin, achieving lightweight, data-free verification for large language models (LLMs).
GHOST: Solving the Traveling Salesman Problem on Graphs of Convex Sets
Jingtao Tang (Simon Fraser University), Hang Ma (Simon Fraser University)
OptimizationGraphBenchmark
🎯 What it does: Proposed and implemented the GHOST framework for optimally solving the Traveling Salesman Problem (GCS-TSP) on graphical convex sets (GCS), integrating combinatorial path search with convex trajectory optimization.
GIER: Addressing Class Imbalance in GNNs Through Experience Replay
Liu Yang (Central South University), Hongyu Zhang (Central South University)
ClassificationData-Centric LearningGraph Neural NetworkGraphBenchmark
🎯 What it does: Propose the GIER framework to compensate for and correct the forgetting phenomenon in GNNs on imbalanced graphs.
GigaMoE: Sparsity-Guided Mixture of Experts for Efficient Gigapixel Object Detection
Xiang Li (Tsinghua University), Yuchen Guo (Tsinghua University)
Object DetectionComputational EfficiencyTransformerMixture of ExpertsImage
🎯 What it does: This paper proposes GigaMoE, which replaces the FFN in Transformer with Mixture-of-Experts and combines sparse window selection to achieve an adaptive computing detection framework for gigapixel images.
GIIM: Graph-based Learning of Inter- and Intra-view Dependencies for Multi-view Medical Image Diagnosis
Tran Bao Sam (NVIDIA), Steven Truong (NVIDIA)
ClassificationConvolutional Neural NetworkGraph Neural NetworkMultimodalityBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: Propose the GIIM framework, which utilizes multi-heterogeneous graphs to simultaneously model intra-view interactions and inter-view dynamics, while providing robust handling for missing views.
GINO-Q: Learning an Asymptotically Optimal Index Policy for Restless Multi-armed Bandits
Gongpu Chen (Imperial College London), Deniz Gündüz (Chinese University of Hong Kong)
OptimizationReinforcement Learning
🎯 What it does: This paper proposes the GINO-Q algorithm, which can learn and implement the asymptotically optimal index policy for Restless Multi-Armed Bandits (RMAB) without relying on indexability.
GitTaskBench: A Benchmark for Code Agents Solving Real-World Tasks Through Code Repository Leveraging
Ziyi Ni (Institute of Automation, Chinese Academy of Sciences), Pin Lyu (Institute of Automation, Chinese Academy of Sciences)
AI Code AssistantTransformerLarge Language ModelAgentic AITextMultimodalityBenchmark
🎯 What it does: Proposed GitTaskBench, a specialized benchmark to evaluate the ability of code agents to complete end-to-end real-world tasks using real GitHub repositories.
GLANCE: Global Actions in a Nutshell for Counterfactual Explainability
Loukas Kavouras (Information Management Systems Institute, Athena Research Center), Ioannis Emiris (National and Kapodistrian University of Athens)
OptimizationExplainability and InterpretabilityTabularFinance Related
🎯 What it does: Propose a Global Explainable Adversarial Explanation (GCE) algorithm named GLANCE, which can generate efficient, low-cost, and interpretable action sets under a given threshold s;
GlitchCleaner: Lightweight Glitch Tokens Repairing by Lossless Gated LoRA in Large Language Models
Yibo Fan (Nankai University), Huan Li (Nankai University)
Anomaly DetectionComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: Propose a lightweight, lossless method called GlitchCleaner, which automatically repairs glitch tokens by introducing a gated LoRA branch in the key MLP layers of large language models;
GlitchMiner: Mining Glitch Tokens in Large Language Models via Gradient-based Discrete Optimization
Zihui Wu (Xidian University), Shiguo Lian (China Unicom)
Anomaly DetectionOptimizationTransformerLarge Language ModelText
🎯 What it does: Designed a behavior-driven gradient-guided local search framework called GlitchMiner to discover glitch tokens that cause abnormal behavior in large language models.
GLOBA: Rethinking Parameter Conflicts in Model Merging
Zehao Liu (Chinese Academy of Sciences), Wei Zhou (Chinese Academy of Sciences)
Representation LearningTransformerLarge Language ModelText
🎯 What it does: Investigated parameter conflicts in multi-task model merging, analyzed the row-column space relationships of task vectors from a geometric perspective, and proposed the GLOBA framework to extract fully orthogonal parameters, classify overlapping parameters, and perform selective fusion based on different types.
Global Compression Commander: Plug-and-Play Inference Acceleration for High-Resolution Large Vision-Language Models
Xuyang Liu (Sichuan University), Honggang Chen (Zhejiang University)
CompressionComputational EfficiencyTransformerMultimodalityBenchmark
🎯 What it does: Aiming at the dynamic cropping strategies in high-resolution vision-language models (HR-LVLM), this paper proposes GlobalCom 2, a zero-training overhead, plug-and-play global-local guided visual token compression framework, which can significantly compress the number of tokens while retaining most of the visual information.
Global-Lens Transformers: Adaptive Token Mixing for Dynamic Link Prediction
Tao Zou (Beihang University), Bowen Du (Beihang University)
TransformerGraph
🎯 What it does: This paper proposes GLFormer, a Transformer framework without self-attention for link prediction on dynamic graphs.
Global-Local Confidence Fusion for Hallucination Detection in Mathematical Reasoning Task
Bo Zhang (Pla Rocket Force University Of Engineering), Zhong Wang (Pla Rocket Force University Of Engineering)
Anomaly DetectionReinforcement Learning from Human FeedbackTransformerLarge Language ModelText
🎯 What it does: Built a hallucination detection framework named ConfFuse that integrates global and local confidence, aimed at identifying hallucinations in mathematical reasoning tasks.
GloCTM: Cross-Lingual Topic Modeling via a Global Context Space
Nguyen Tien Phat (Hanoi University of Science and Technology), Thien Huu Nguyen (University of Oregon)
Representation LearningAuto EncoderText
🎯 What it does: Proposes GloCTM, leveraging a global context space and a dual-channel VAE architecture to achieve unified learning and alignment of cross-lingual topics.
GLoMOT: Efficient Online GNN-based Low-Frame-Rate Multi-Object Tracker
Yaxuan Hu, Zhongyuan Wang (Wuhan University)
Object TrackingComputational EfficiencyGraph Neural NetworkVideoGraph
🎯 What it does: Proposed an online low-frame-rate multi-object tracking framework called GLoMOT, which realizes real-time association using graph neural networks.
GloTok: Global Perspective Tokenizer for Image Reconstruction and Generation
Xuan Zhao (Fudan University), Shuigeng Zhou (Tencent)
GenerationTransformerAuto EncoderImage
🎯 What it does: Proposes GloTok image tokenizer, leveraging global relation learning and residual learning to achieve a more uniform semantic distribution, thereby enhancing image reconstruction and generation quality.
GlyphShield: Document Watermarking for the Physical World via Vector Typeface Synthesis
Nan Sun (Huazhong University Of Science And Technology), Chengxin Zhao (Huazhong University Of Science And Technology)
GenerationTransformerImageText
🎯 What it does: Proposed an end-to-end vector font watermark framework called GlyphShield, which can embed invisible watermarks into text and accurately extract them even after physical world disturbances.
GMAI-VL & GMAI-VL-5.5M: A Large Vision-Language Model and a Comprehensive Multimodal Dataset Towards General Medical AI
Tianbin Li (Shanghai Artificial Intelligence Laboratory), Junjun He (Shanghai Artificial Intelligence Laboratory)
TransformerLarge Language ModelSupervised Fine-TuningVision Language ModelMultimodalityBiomedical Data
🎯 What it does: Constructed a general-purpose medical vision-language model named GMAI-VL with 7B parameters, and developed a high-quality multimodal medical dataset named GMAI-VL-5.5M containing 5.5M samples based on over 200 specialized medical datasets, supporting cross-modal learning and medical question answering from images to text.
Goal-Oriented Time-Series Forecasting: Foundation Framework Design
Luca-Andrei Fechete (École Polytechnique), Tareq Si Salem (Huawei Technologies)
Convolutional Neural NetworkTransformerTime Series
🎯 What it does: This paper proposes a time series prediction framework that performs fine-grained partitioning and dynamic reweighting of the prediction space during training, enabling the model to flexibly focus on target intervals during inference according to different application requirements without needing retraining.