AAAI 2025 Papers — Page 12
AAAI Conference on Artificial Intelligence · 3028 papers
Fast Omni-Directional Image Super-Resolution: Adapting the Implicit Image Function with Pixel and Semantic-Wise Spherical Geometric Priors
Xuelin Shen (Shenzhen University), Xu Wang (Shenzhen University)
RestorationSuper ResolutionImage
🎯 What it does: A fast arbitrary scale panoramic image super-resolution method FAOR is proposed.
Fast Think-on-Graph: Wider, Deeper and Faster Reasoning of Large Language Model on Knowledge Graph
Xujian Liang (Beijing University Of Posts And Telecommunications), Zhaoquan Gu (Harbin Institute of Technology)
TransformerLarge Language ModelPrompt EngineeringTextGraphRetrieval-Augmented Generation
🎯 What it does: Proposes the FastThink-on-Graph (FastToG) framework, allowing large language models to perform multi-step reasoning on knowledge graphs at the community level.
Fast Track to Winning Tickets: Repowering One-Shot Pruning for Graph Neural Networks
Yanwei Yue (Tongji University), Dawei Cheng (Tongji University)
Graph Neural NetworkGraph
🎯 What it does: This study proposes a FastGLT framework based on one-shot pruning and denoising for the rapid discovery of Graph Lottery Tickets (GLT) in Graph Neural Networks (GNN), aiming to obtain sparse subgraphs and sub-networks while maintaining performance.
Faster Double Adaptive Gradient Methods
Feihu Huang (Nanjing University of Aeronautics and Astronautics), Yuning Luo (Nanjing University of Aeronautics and Astronautics)
OptimizationImageText
🎯 What it does: This paper proposes a dual adaptive gradient method (2AdaSGD and 2AdaSPIDER) that simultaneously adapts the learning rate and batch size in non-convex finite-sum optimization.
FastLGS: Speeding Up Language Embedded Gaussians with Feature Grid Mapping
Yuzhou Ji (East China Normal University), Yuan Xie (East China Normal University)
RetrievalComputational EfficiencyGaussian SplattingPoint Cloud
🎯 What it does: FastLGS proposes a method for the rapid construction and real-time querying of a 3D high-resolution open vocabulary semantic field based on semantic feature grid mapping.
FastPERT: Towards Fast Microservice Application Latency Prediction via Structural Inductive Bias over PERT Networks
Da Sun Handason Tam (Chinese University of Hong Kong), Wing Cheong Lau (Chinese University of Hong Kong)
OptimizationComputational EfficiencyGraph
🎯 What it does: This paper proposes the FastPERT model, which constructs a PERT graph on the execution trajectory of microservices to first predict the delay of each microservice task, and then quickly obtain the end-to-end delay using computational and structural induced bias.
FATE: Feature-Adapted Parameter Tuning for Vision-Language Models
Zhengqin Xu (Shanghai Jiao Tong University), Wei Shen (Shanghai Jiao Tong University)
ClassificationDomain AdaptationTransformerSupervised Fine-TuningVision Language ModelImageMultimodality
🎯 What it does: This paper achieves efficient fine-tuning of visual-language models by projecting the features extracted by the CLIP visual encoder as learnable parameters of the language encoder.
FatesGS: Fast and Accurate Sparse-View Surface Reconstruction Using Gaussian Splatting with Depth-Feature Consistency
Han Huang (Tsinghua University), Yu-Shen Liu (Tsinghua University)
RestorationOptimizationGaussian SplattingImage
🎯 What it does: This paper proposes a sparse view surface reconstruction framework named FatesGS, which utilizes efficient Gaussian Splatting and optimizes surface geometry through in-view depth consistency and multi-view feature alignment.
FBRT-YOLO: Faster and Better for Real-Time Aerial Image Detection
Yao Xiao (Beijing Institute of Technology), Jianan Li (Beijing Institute of Technology)
Object DetectionConvolutional Neural NetworkImage
🎯 What it does: A real-time object detection framework FBRT-YOLO specifically designed for aerial images is proposed, incorporating lightweight modules Feature Complementary Mapping Module (FCM) and Multi-Kernel Perception Unit (MKP) to achieve efficient small object detection.
FCOM: A Federated Collaborative Online Monitoring Framework via Representation Learning
Tanapol Kosolwattana (University of Houston), Ying Lin (University of Michigan)
Federated LearningRepresentation LearningBiomedical DataAlzheimer's Disease
🎯 What it does: A Federated Collaborative Online Monitoring framework (FCOM) is proposed, which utilizes representation learning to mine the latent structure of heterogeneous groups and implements resource allocation using an event-triggered UCB algorithm in a distributed environment.
FD2-Net: Frequency-Driven Feature Decomposition Network for Infrared-Visible Object Detection
Ke Li (Xidian University), Quan Wang (Northwest Institute of Nuclear Technology)
Object DetectionConvolutional Neural NetworkImageMultimodality
🎯 What it does: This paper proposes a Frequency-Driven Feature Decomposition Network (FD2-Net), which separates the features of infrared images and visible light images into high-frequency and low-frequency components in the frequency domain, and utilizes a multimodal reconstruction unit to cross-enhance the extracted frequency domain features, thereby improving the performance of infrared-visible object detection.
FEAST-Mamba: FEAture and SpaTial Aware Mamba Network with Bidirectional Orthogonal Fusion for Cross-Modal Point Cloud Segmentation
Chade Li (Chinese Academy of Sciences), Yihong Wu (Chinese Academy of Sciences)
SegmentationAutonomous DrivingMultimodalityPoint Cloud
🎯 What it does: A multi-modal point cloud semantic segmentation framework FEAST-Mamba based on Mamba is proposed, combining bidirectional orthogonal fusion and feature-space dual perception rearrangement.
Feature Clipping for Uncertainty Calibration
Linwei Tao (University of Sydney), Chang Xu (City University of Hong Kong)
TransformerImage
🎯 What it does: A post-hoc feature clipping method is proposed to enhance the model calibration performance of deep neural networks.
Feature Denoising Diffusion Model for Blind Image Quality Assessment
Xudong Li (Xiamen University), Sicheng Zhao (Tsinghua University)
TransformerDiffusion modelImage
🎯 What it does: This paper proposes a blind image quality assessment framework based on diffusion models, PFD-IQA, which enhances the perception of low-level distortion information by denoising semantic features transferred from high-level tasks, and utilizes visual-language prompts to condition the denoising process, ultimately achieving efficient and accurate quality scoring.
Feature-Structure Adaptive Completion Graph Neural Network for Cold-start Recommendation
Songyuan Lei (Tianjin University), Di Jin (Tianjin University)
Recommendation SystemGraph Neural NetworkLarge Language ModelAuto EncoderTabular
🎯 What it does: This paper proposes a feature and structure completion graph neural network, FS-GNN, specifically designed to address the issues of feature and structural missing data in cold-start recommendations.
Fed-DFA: Federated Distillation for Heterogeneous Model Fusion Through the Adversarial Lens
Zichen Wang (Zhejiang University), Jiming Chen (Zhejiang University)
Federated LearningKnowledge DistillationAdversarial AttackImage
🎯 What it does: This study investigates the performance bottlenecks of federated distillation in heterogeneous model fusion and proposes a decision boundary estimation based on adversarial attacks to improve the distillation process.
FedAA: A Reinforcement Learning Perspective on Adaptive Aggregation for Fair and Robust Federated Learning
Jialuo He (Huazhong University of Science and Technology), Xiaojin Zhang (Huazhong University of Science and Technology)
Federated LearningReinforcement LearningImageText
🎯 What it does: Proposes the FedAA method, which utilizes reinforcement learning to dynamically adjust client aggregation weights to enhance the robustness and fairness of federated learning.
FedCFA: Alleviating Simpson’s Paradox in Model Aggregation with Counterfactual Federated Learning
Zhonghua Jiang (Zhejiang University), Fei Wu (Zhejiang University)
Federated LearningContrastive LearningImage
🎯 What it does: The FedCFA framework is proposed in federated learning, utilizing counterfactual learning to generate adversarial samples to mitigate the local and global model bias caused by Simpson's paradox.
FedCross: Intertemporal Federated Learning Under Evolutionary Games
Jianfeng Lu (Wuhan University of Science and Technology), Hao Fu (Wuhan University of Science and Technology)
Federated LearningImage
🎯 What it does: Proposes the FedCross framework, which ensures the continuity of federated learning tasks through task migration and incentive mechanisms in mobile networks.
Federated Assemblies
Daniel Halpern (Harvard University), Nimrod Talmon (Ben-Gurion University)
OptimizationFederated LearningGraph
🎯 What it does: Proposes a federal-style citizen assembly framework and designs a random election algorithm to meet the requirements of descriptive representation under a hierarchical structure and the pre/post-representative requirements of sub-assemblies.
Federated Binary Matrix Factorization Using Proximal Optimization
Sebastian Dalleiger (KTH Royal Institute of Technology), Michael Kamp (Institute for AI in Medicine)
Recommendation SystemFederated LearningSafty and PrivacyTabular
🎯 What it does: A federated Boolean matrix factorization algorithm (FELB) based on proximal gradient descent is proposed, which can complete distributed binary matrix factorization without leaking the original binary data, and provides guarantees for convergence and differential privacy.
Federated Causally Invariant Feature Learning
Xianjie Guo (Hefei University of Technology), Xiaoxiao Li (University of British Columbia)
Federated LearningAuto EncoderTabular
🎯 What it does: A framework for learning causal invariant features in federated learning, called FedCIFL, is proposed, which addresses data heterogeneity (Non-IID) and out-of-distribution (OOD) issues, enabling feature selection without sharing raw data.
Federated Foundation Models on Heterogeneous Time Series
Shengchao Chen (Australian Artificial Intelligence Institute), Chengqi Zhang (Hong Kong Polytechnic University)
Anomaly DetectionFederated LearningTransformerMixture of ExpertsTime Series
🎯 What it does: Train a temporal foundation model using the federated learning framework FFTS to avoid centralized data fusion, thereby enhancing the generalization ability of cross-domain tasks.
Federated Graph Anomaly Detection Through Contrastive Learning with Global Negative Pairs
Nannan Wu (Tianjin University), Wenjun Wang (Tianjin University)
Anomaly DetectionFederated LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: A FedCLGN framework is constructed in a federated learning environment to detect anomalous nodes in attribute graphs using contrastive learning.
Federated Graph Condensation with Information Bottleneck Principles
Bo Yan (Beijing University of Posts and Telecommunications), Chuan Shi (China University of Mining and Technology)
Federated LearningSafty and PrivacyGraph Neural NetworkGraph
🎯 What it does: This paper proposes the Federated Graph Condensation (FGC) framework, which generates small synthetic graphs for subsequent tasks such as node classification using gradient matching and information bottleneck techniques without sharing raw data.
Federated Graph-Level Clustering Network
Jingxin Liu (Hainan University), Xin Peng (National University of Defense Technology)
Federated LearningSafty and PrivacyGraph Neural NetworkGraph
🎯 What it does: A federated learning-based unsupervised graph-level clustering framework, FedGCN, is proposed, which can achieve clustering on cross-domain, non-IID graph data.
Federated Learning with Sample-level Client Drift Mitigation
Haoran Xu (Zhejiang University), Hao Ren (Shandong University)
Federated LearningImage
🎯 What it does: This paper proposes FedBSS—a two-stage federated learning framework that alleviates client drift by first collecting diverse knowledge during a warm-up phase through bias identification based on sample loss and uncertainty, and then gradually adding samples from low bias to high bias during an evolution phase.
Federated Recommendation with Explicitly Encoding Item Bias
Zhihao Wang (Wuhan University), Bing Li (Hubei Luojia Laboratory)
Recommendation SystemFederated LearningTabular
🎯 What it does: This study addresses the issue of label bias in platform-level federated recommendation, proposing the FREIB method which explicitly encodes item bias and combines global knowledge guidance with feature prototype alignment.
Federated t-SNE and UMAP for Distributed Data Visualization
Dong Qiao (Chinese University of Hong Kong), Jicong Fan (Chinese University of Hong Kong)
Federated LearningSafty and PrivacyImage
🎯 What it does: A high-dimensional data visualization framework based on federated learning, Fed-tSNE and Fed-UMAP, is proposed, which can estimate the distance/similarity matrix of t-SNE/UMAP and generate low-dimensional embeddings without exchanging raw data.
Federated Unlearning with Gradient Descent and Conflict Mitigation
Zibin Pan (Chinese University of Hong Kong), Junhua Zhao (Chinese University of Hong Kong)
OptimizationFederated LearningConvolutional Neural NetworkImage
🎯 What it does: A novel Federated Unlearning framework, FedOSD, is proposed to efficiently remove specified client data from the global model without retraining, while maintaining model performance.
Federated Unsupervised Domain Generalization Using Global and Local Alignment of Gradients
Farhad Pourpanah (Queen's University), Ali Etemad (Queen's University)
Domain AdaptationFederated LearningContrastive LearningImage
🎯 What it does: Addressed the problem of federated unsupervised domain generalization, proposing the FedGaLA method that achieves domain-invariant feature learning through local and global gradient alignment.
Federated Weakly Supervised Video Anomaly Detection with Multimodal Prompt
Benfeng Wang (Sun Yat-sen University), Yong Xu (Harbin Institute of Technology)
Anomaly DetectionFederated LearningTransformerPrompt EngineeringContrastive LearningVideoMultimodality
🎯 What it does: A federated weakly supervised video anomaly detection framework based on CLIP is proposed, utilizing a context-driven prompt generator for global and local contexts to achieve anomaly frame localization and detection.
FedFSL-CFRD: Personalized Federated Few-Shot Learning with Collaborative Feature Representation Disentanglement
Shanfeng Wang (Xidian University), Maoguo Gong (Inner Mongolia Normal University)
Federated LearningMeta LearningImage
🎯 What it does: The FedFSL-CFRD method is proposed, achieving personalized models in federated few-shot learning through feature decoupling at the client and category levels.
FedGOG: Federated Graph Out-of-Distribution Generalization with Diffusion Data Exploration and Latent Embedding Decorrelation
Pengyang Zhou (Zhejiang University), Xiaolin Zheng (Zhejiang University)
Federated LearningDrug DiscoveryGraph Neural NetworkDiffusion modelScore-based ModelGraphStochastic Differential Equation
🎯 What it does: The FedGOG framework is proposed, combining Diffusion Data Exploration (DDE) and Latent Embedding Decorrelation (LED) to achieve OOD generalization in federated graph learning.
FedMSGL: A Self-Expressive Hypergraph Based Federated Multi-View Learning
Daoyuan Li (Guangdong University of Technology), Shengli Xie (Guangdong University of Technology)
Federated LearningMultimodality
🎯 What it does: Designed and implemented FedMSGL, a self-expressive hypergraph multi-view learning framework in a vertical federated learning environment, addressing the issues of unbalanced feature dimensions and multi-view information fusion.
FedPIA – Permuting and Integrating Adapters Leveraging Wasserstein Barycenters for Finetuning Foundation Models in Multi-Modal Federated Learning
Pramit Saha (University of Oxford), J. Alison Noble (University of Oxford)
Federated LearningSupervised Fine-TuningVision Language ModelMultimodalityBiomedical DataMagnetic Resonance ImagingComputed TomographyUltrasound
🎯 What it does: This paper proposes the FedPIA framework for parameter-efficient fine-tuning of large-scale vision-language models in multimodal federated learning.
FedPop: Federated Population-based Hyperparameter Tuning
Haokun Chen (Siemens Technology), Volker Tresp (Ludwig Maximilian University of Munich)
Federated LearningHyperparameter SearchImage
🎯 What it does: Designed and validated FedPop, a method for online population evolutionary hyperparameter tuning in federated learning.
FedSA: A Unified Representation Learning via Semantic Anchors for Prototype-based Federated Learning
Yanbing Zhou (Chongqing University), Yingbo Wu (Chongqing University)
Federated LearningRepresentation LearningContrastive LearningImage
🎯 What it does: The FedSA framework is proposed in federated learning, utilizing semantic anchors to decouple prototype generation, helping each client learn consistent feature representations and calibrate classifiers, thereby improving performance in heterogeneous environments.
FedSPU: Personalized Federated Learning for Resource-Constrained Devices with Stochastic Parameter Update
Ziru Niu (RMIT University), A. K. Qin (Swinburne University of Technology)
Federated LearningConvolutional Neural NetworkImageStochastic Differential EquationAudio
🎯 What it does: The FedSPU framework is proposed, which maintains the integrity of local models in personalized federated learning by randomly freezing some neurons instead of pruning them.
FedSum: Data-Efficient Federated Learning Under Data Scarcity Scenario for Text Summarization
Zhiyong Ma (South China University of Technology), Jian Chen (Washington State University)
Federated LearningTransformerSupervised Fine-TuningTextBiomedical Data
🎯 What it does: In the task of text summarization with data scarcity under federated learning, the FedSum framework is proposed, which combines depth (sample level) and breadth (knowledge level) expansion to enhance model performance.
FedVCK: Non-IID Robust and Communication-Efficient Federated Learning via Valuable Condensed Knowledge for Medical Image Analysis
Guochen Yan (Peking University), Zhonghai Wu (Peking University)
Federated LearningContrastive LearningImageBiomedical DataComputed Tomography
🎯 What it does: A federated learning framework called FedVCK based on value-compressed knowledge is proposed to address the issues of non-IID and communication costs in medical image analysis.
Few-Shot Audio-Visual Class-Incremental Learning with Temporal Prompting and Regularization
Yawen Cui (Hong Kong Polytechnic University), Xiaopeng Hong (Hong Kong University of Science and Technology)
ClassificationRecognitionTransformerPrompt EngineeringVideoMultimodalityAudio
🎯 What it does: In the audio-visual classification task with limited labeled samples, the FS-AVCIL framework is proposed to achieve continuous learning from base categories to incremental categories, balancing multi-modal fusion and catastrophic forgetting suppression.
Few-Shot Domain Adaptation for Learned Image Compression
Tianyu Zhang (University of Science and Technology of China), Dong Liu (University of Science and Technology of China)
CompressionDomain AdaptationConvolutional Neural NetworkImage
🎯 What it does: A general few-shot domain adaptation method is proposed, utilizing lightweight adapters for channel-level redistribution on a pre-trained learning-based image compression model, significantly improving the model's compression performance across various domains.
Few-Shot Fine-Grained Image Classification with Progressively Feature Refinement and Continuous Relationship Modeling
Zhen-Xiang Ma (Shandong University), Xin-Shun Xu (Shandong University)
ClassificationConvolutional Neural NetworkGraph Neural NetworkImage
🎯 What it does: This paper proposes a Few-Shot fine-grained image classification method called SUITED, which combines advanced feature refinement and continuous relationship modeling to improve category discrimination and relationship modeling accuracy under few samples.
Few-Shot Incremental Learning via Foreground Aggregation and Knowledge Transfer for Audio-Visual Semantic Segmentation
Jingqiao Xiu (National University of Singapore), Roger Zimmermann (Institute for Infocomm Research, A*STAR)
SegmentationTransformerVideoMultimodalityAudio
🎯 What it does: This study explores the application of Few-Shot Incremental Learning (FSIL) in the Audio-Video Semantic Segmentation (AVSS) task, proposing the FINGER framework and implementing a baseline model.
Few-Shot, No Problem: Descriptive Continual Relation Extraction
Nguyen Xuan Thanh (Oraichain Labs), Thien Huu Nguyen (University of Oregon)
RetrievalRepresentation LearningTransformerLarge Language ModelContrastive LearningText
🎯 What it does: This paper proposes a retrieval-based few-shot continual relation extraction method that enhances class representation by generating diverse relation descriptions and trains and infers within a retrieval framework, significantly alleviating catastrophic forgetting.
FFCG: Effective and Fast Family Column Generation for Solving Large-Scale Linear Program
Yi-Xiang Hu (University of Science and Technology of China), Xiang-Yang Li (University of Science and Technology of China)
OptimizationGraph Neural NetworkReinforcement LearningTabularBenchmark
🎯 What it does: This paper proposes Fast Family Column Generation (FFCG), a multi-column selection reinforcement learning framework designed to accelerate the column generation of large linear programming problems.
FigStep: Jailbreaking Large Vision-Language Models via Typographic Visual Prompts
Yichen Gong (Tsinghua University), Xiaoyun Wang (Tsinghua University)
Adversarial AttackTransformerLarge Language ModelPrompt EngineeringVision Language ModelImageTextMultimodality
🎯 What it does: This paper studies a black-box jailbreak method called FigStep that converts text instructions into formatted images, successfully bypassing the safety alignment of multimodal language models.
Filling Memory Gaps: Enhancing Continual Semantic Parsing via SQL Syntax Variance-Guided LLMs Without Real Data Replay
Ruiheng Liu (Harbin Institute of Technology), Bailong Yang (Xi'an Research Institute of High-Tech)
Knowledge DistillationTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This work proposes a continuous semantic parsing framework LECSP that does not rely on historical data replay or ideal scene constraints.
Filter or Compensate: Towards Invariant Representation from Distribution Shift for Anomaly Detection
Zining Chen (Beijing University of Posts and Telecommunications), Aidong Men (Beijing University of Posts and Telecommunications)
Anomaly DetectionKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: A new anomaly detection method called FiCo is proposed, which addresses the performance degradation caused by distribution shifts through two steps: distribution-specific compensation and distribution-invariant filtering.
FilterTS: Comprehensive Frequency Filtering for Multivariate Time Series Forecasting
Yulong Wang (Nankai University), Kai Wang (Nankai University)
Time Series
🎯 What it does: The FilterTS model is designed to extract stable and varying frequency components from multivariate time series using frequency domain filtering techniques, improving prediction accuracy.
FIND: A Framework for Discovering Formulas in Data
Tingxiong Xiao (Tsinghua University), Jinli Suo (Tsinghua University)
TabularPhysics Related
🎯 What it does: Proposed the FIND framework, which utilizes hidden layers to generate dimensionless latent variables and directly discovers physical formulas from observational data through polynomial expression layers;
Fine-grained Adaptive Visual Prompt for Generative Medical Visual Question Answering
Ting Yu (Hangzhou Normal University), Ke Zhang (Hangzhou Dianzi University)
GenerationRetrievalTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringImageMultimodalityBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: Proposes the FAVP framework, utilizing fine-grained adaptive visual prompts (AVPC) and a hierarchical answer generator (HAG) to enhance the precise localization and answer quality of generative medical visual question answering.
Fine-Grained Graph Representation Learning for Heterogeneous Mobile Networks with Attentive Fusion and Contrastive Learning
Shengheng Liu (National Mobile Communications Research Laboratory), Yongming Huang (National Mobile Communications Research Laboratory)
Autonomous DrivingRepresentation LearningGraph Neural NetworkContrastive LearningGraphTime Series
🎯 What it does: An unsupervised data and model-driven graph structure learning framework DMGSL is proposed for the automated construction and fine-grained optimization of the network topology of Wireless Data Knowledge Graphs (WDKG), and its effectiveness is validated in node classification tasks.
Fine-Grained Perception in Panoramic Scenes: A Novel Task, Dataset, and Method for Object Importance Ranking
Jia Song (China University of Petroleum), Shanchen Pang (China University of Petroleum)
Object DetectionSegmentationImage
🎯 What it does: The FOIR-360 task is proposed, which predicts the fine-grained importance ranking of all instances in a 360-degree image, and the first 360Rank dataset is constructed.
Fine-Tuning Language Models with Collaborative and Semantic Experts
Jiaxi Yang (Shenzhen Institutes of Advanced Technology), Junyang Lin (Shenzhen Institutes of Advanced Technology)
Explainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningMixture of ExpertsText
🎯 What it does: A two-stage supervised fine-tuning (SFT) process is proposed, where experts in general knowledge, mathematics, and programming are first trained on a feed-forward network, and then these experts are integrated into a single model through a semantically guided Mixture-of-Experts (MoE) routing.
FIRM: Flexible Interactive Reflection ReMoval
Xiao Chen (Hong Kong Polytechnic University), Zhaoxiang Zhang (Institute of Automation, Chinese Academy of Sciences)
RestorationSegmentationImageMultimodalityBenchmark
🎯 What it does: A multi-modal interaction framework for reflection removal, FIRM, has been developed, capable of handling various user prompts such as points, boxes, strokes, and text.
First Line of Defense: A Robust First Layer Mitigates Adversarial Attacks
Janani Suresh (Indian Institute Of Technology Madras), Sheetal Kalyani (Indian Institute Of Technology Madras)
Adversarial AttackConvolutional Neural NetworkImage
🎯 What it does: A first-layer network called the Anti-Interference Noise Filter (ANF) is proposed, which achieves inherent robustness against adversarial attacks through large convolutional kernels, more filters, and max pooling.
First-Order Automata
Luca Geatti (University of Udine), Nicola Gigante (Free University of Bozen-Bolzano)
🎯 What it does: A first-order automaton model is defined to capture first-order linear temporal logic (FOLTL) over finite words and to study its regular languages and closure properties.
First-Order Federated Bilevel Learning
Yifan Yang (University at Buffalo), Kaiyi Ji (Rice University)
OptimizationFederated LearningImage
🎯 What it does: This paper proposes the MemFBO algorithm for Federated Bi-level Optimization (FBO), which utilizes first-order gradients to achieve global and local updates within a single-loop structure, significantly reducing computational and memory consumption.
Fit and Prune: Fast and Training-free Visual Token Pruning for Multi-modal Large Language Models
Weihao Ye (Xiamen University), Yiyi Zhou (Xiamen University)
OptimizationComputational EfficiencyTransformerVision Language ModelMultimodality
🎯 What it does: Proposes FitPrune, a training-free visual token pruning method that utilizes attention distribution to quickly generate pruning schemes;
FLAME: Learning to Navigate with Multimodal LLM in Urban Environments
Yunzhe Xu (Shanghai Jiao Tong University), Hesheng Wang (Shanghai Jiao Tong University)
Data SynthesisAutonomous DrivingTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelTextMultimodality
🎯 What it does: Designed and trained a city visual-language navigation agent FLAME based on a multimodal LLM (Flamingo), employing a three-stage fine-tuning process and automatic data synthesis to achieve efficient street scene understanding and path planning.
Flash Diffusion: Accelerating Any Conditional Diffusion Model for Few Steps Image Generation
Clément Chadebec (Jasper Research), Benjamin Aubin (Jasper Research)
GenerationComputational EfficiencyKnowledge DistillationDiffusion modelGenerative Adversarial NetworkImage
🎯 What it does: An efficient, fast, and general diffusion model distillation method named Flash Diffusion is proposed for generating high-quality images with only two steps (or fewer) of sampling.
FlexDataset: Crafting Annotated Dataset Generation for Diverse Applications
Ellen Yi-Ge (Carnegie Mellon University), Leo Shawn (University of the Chinese Academy of Sciences)
Object DetectionSegmentationData SynthesisDepth EstimationTransformerDiffusion modelImage
🎯 What it does: The FlexDataset framework is proposed, achieving composition-to-data (C2D) generation based on scene composition, capable of synthesizing multiple instances and multiple categories (MIMC) scenes at once and automatically generating pixel-level annotations for tasks such as salient object detection, depth estimation, and semantic/instance segmentation.
Flexible Sharpness-Aware Personalized Federated Learning
Xinda Xing (University of Electronic Science and Technology of China), Guisong Liu (Southwest University of Finance and Economics)
Federated LearningConvolutional Neural NetworkImage
🎯 What it does: A framework based on Hierarchical Flexible Sharpness-Aware Minimization (FedFSA) is designed in personalized federated learning, applying larger perturbations to the layers with the highest sharpness during local training to enhance the model's generalization performance.
FlexiTex: Enhancing Texture Generation via Visual Guidance
Dadong Jiang (Tianjin University), Zhihui Ke (Tianjin University)
GenerationData SynthesisDiffusion modelMesh
🎯 What it does: We propose FlexiTex, a multi-view texture generation framework that supports text and image conditions, achieving high-quality, globally consistent textures through visual guidance enhancement and direction-aware adaptation.
FlexPose: Pose Distribution Adaptation with Limited Guidance
Zixiao Wang (Chinese University of Hong Kong), Bei Yu (Chinese University of Hong Kong)
GenerationPose EstimationDomain AdaptationGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes FlexPose, a method for transferring a pre-trained pose generator from the source domain to the target domain using a small number of target pose annotations;
FloNa: Floor Plan Guided Embodied Visual Navigation
Jiaxin Li (Beijing Institute of Technology), Feng Liu (Beijing Racobit Electronic Information Technology Company)
Robotic IntelligenceTransformerDiffusion modelImage
🎯 What it does: This paper proposes a new task called Floor Plan Guided Embodied Visual Navigation (FloNa), which requires agents to complete target navigation based solely on RGB observations and floor plans.
Flow Factorization for Efficient Generative Flow Networks
Jiashun Liu (Hong Kong University of Science and Technology), Ling Pan (Mila-Quebec AI Institute, McGill University)
GenerationData SynthesisComputational EfficiencyReinforcement LearningFlow-based ModelGraph
🎯 What it does: This paper proposes Bifurcated Generative Flow Networks (BN), which decomposes edge flow into state flow and edge allocation to improve the data efficiency and scalability of GFlowNet.
FlowMamba: Learning Point Cloud Scene Flow with Global Motion Propagation
Min Lin (Huazhong University of Science and Technology), Xin Yang (Huazhong University of Science and Technology)
Autonomous DrivingOptimizationOptical FlowPoint Cloud
🎯 What it does: A point cloud scene flow estimation network called FlowMamba is proposed, which achieves high-precision scene flow estimation through the ISU module and Feature Induced Ordering (FIO).
FlowPolicy: Enabling Fast and Robust 3D Flow-Based Policy via Consistency Flow Matching for Robot Manipulation
Qinglun Zhang (University of Electronic Science and Technology of China), Shuaicheng Liu (Megvii Technology)
Robotic IntelligenceReinforcement LearningFlow-based ModelPoint Cloud
🎯 What it does: This paper proposes FlowPolicy, a framework for generating single-step real-time robot policies using 3D point clouds as conditions, achieved through consistency flow matching.
FLUE: Streamlined Uncertainty Estimation for Large Language Models
Shiqi Gao (Beihang University), Jianxin Li (Beihang University)
GenerationOptimizationComputational EfficiencyRecurrent Neural NetworkTransformerLarge Language ModelText
🎯 What it does: This paper proposes FLUE, a method for estimating token and sequence uncertainty in large language models through hidden layer entropy sampling combined with state transition post-processing during a single forward inference.
FMPM-DNet: Hyperspectral Pansharpening Dynamic Network Based on Feature Modulation and Probability Mask
Xiaozheng Wang (Tiangong University), Aoqi Zhao (Jinhua University of Vocational Technology)
RestorationSuper ResolutionConvolutional Neural NetworkImage
🎯 What it does: A dual-stage dynamic fusion network FMPM-DNet is proposed for the fusion of high-resolution PAN images and low-resolution hyperspectral images;
FNIN: A Fourier Neural Operator-based Numerical Integration Network for Surface-from-gradients
Jiaqi Leng (Ocean University of China), Hao Fan (Ocean University of China)
RestorationOptimizationImage
🎯 What it does: A two-stage numerical integration network (FNIN) based on the Fourier Neural Operator is proposed for recovering three-dimensional surfaces from gradient fields.
Focus on Local: Finding Reliable Discriminative Regions for Visual Place Recognition
Changwei Wang (Qilu University of Technology), Shibiao Xu (Beijing University of Posts and Telecommunications)
RecognitionRetrievalTransformerImage
🎯 What it does: This paper proposes a two-stage visual place recognition framework called FoL, which explicitly models reliable discrimination of local regions and combines pseudo-correspondence weakly supervised training to achieve improvements in the overall retrieval and re-ranking performance.
FOCUS: Towards Universal Foreground Segmentation
Zuyao You (Fudan University), Zuxuan Wu (Fudan University)
Object DetectionSegmentationTransformerContrastive LearningImageMultimodality
🎯 What it does: A unified foreground segmentation framework called FOCUS is proposed, which can simultaneously handle various foreground segmentation tasks (SOD, COD, SD, DBD, FD) and generate boundary-aware masks.
FoldToken: Learning Protein Language via Vector Quantization and Beyond
Zhangyang Gao (Westlake University), Stan Z. Li
GenerationProtein Structure PredictionTransformerLarge Language ModelBiomedical Data
🎯 What it does: This paper maps protein sequences and three-dimensional structures to discrete symbols through vector quantization, constructing FoldTokenizer and FoldGPT based on this discrete language to achieve joint generation of protein sequences and structures.
Follow-Your-Click: Open-domain Regional Image Animation via Motion Prompts
Yue Ma (Hong Kong University of Science and Technology), Qifeng Chen (Hong Kong University of Science and Technology)
SegmentationGenerationDiffusion modelOptical FlowImageVideo
🎯 What it does: Proposed the Follow-Your-Click framework, which utilizes single clicks and motion prompts to achieve open-domain region image animation.
Forecasting Competitions with Correlated Events
Rafael Frongillo (University of Colorado Boulder), Bo Waggoner (University of Colorado Boulder)
🎯 What it does: This paper studies how to design an effective prediction competition mechanism when events are correlated;
Foreground-Covering Prototype Generation and Matching for SAM-Aided Few-Shot Segmentation
Suho Park (Sungkyunkwan University), Jae-Pil Heo (Sungkyunkwan University)
SegmentationConvolutional Neural NetworkImage
🎯 What it does: A foreground coverage prototype generation and matching (FGPM) method is proposed for few-shot segmentation, which generates visual reference prompts for the SAM parser by constructing prototypes of support images and query images and performing prototype-to-prototype matching.
Forget to Flourish: Leveraging Machine-Unlearning on Pretrained Language Models for Privacy Leakage
Md Rafi Ur Rashid (Pennsylvania State University), Shagufta Mehnaz (Mitsubishi Electric Research Laboratories)
Safty and PrivacyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Targeted machine unlearning is applied to pre-trained language models to poison the model, making it more prone to privacy leaks during downstream fine-tuning, with a focus on experimental validation against membership inference and data extraction attacks.
Formal Quality Measures for Predictors in Markov Decision Processes
Christel Baier (Technische Universitat Dresden), Robin Ziemek (Technische Universitat Dresden)
OptimizationReinforcement LearningTabular
🎯 What it does: Research on the quality metrics of predictors in Markov Decision Processes
Formal Synthesis of Barrier Certificates Using Fourier Kolmogorov-Arnold Network
Xiongqi Zhang (Zhejiang Sci-Tech University), Zuohua Ding (Zhejiang Sci-Tech University)
Safty and PrivacyComputational EfficiencyTabularOrdinary Differential Equation
🎯 What it does: This study investigates the method of using Fourier KAN to generate and verify barrier certificates to improve the efficiency of safety verification for continuous dynamical systems.
Formally Verified Approximate Policy Iteration
Maximilian Schäffeler, Mohammad Abdulaziz (King's College London)
OptimizationReinforcement Learning
🎯 What it does: This paper achieves the formal verification of Approximate Policy Iteration (API) for factored Markov Decision Processes (MDP) through interactive theorem proving, and provides an executable, proven correct implementation.
Forward KL Regularized Preference Optimization for Aligning Diffusion Policies
Zhao Shan (China Telecom), Chenjia Bai (Hong Kong University of Science and Technology)
OptimizationRobotic IntelligenceReinforcement LearningDiffusion modelMultimodality
🎯 What it does: A diffusion model-based policy alignment framework FKPD is proposed in the offline reinforcement learning scenario, where a baseline diffusion policy is first trained using behavior cloning, and then the policy is aligned to human preferences through direct preference optimization (DPO) combined with forward KL regularization.
Foundation Model Driven Appearance Extraction for Robust Multiple Object Tracking
Teng Fu (Fudan University), Xiangyang Xue (Fudan University)
Object TrackingTransformerVision Language ModelVideoMultimodality
🎯 What it does: This paper proposes FDTracker, a multi-target tracking framework based on multimodal foundational models. It extracts semantic appearance features through a two-stage training TAE module and integrates IOU and appearance similarity in trajectory management to achieve long-term re-identification and suppress identity switching.
Fourier Guided Adaptive Adversarial Augmentation for Generalization in Visual Reinforcement Learning
Jeong Woon Lee (Kyung Hee University), Hyoseok Hwang (Kyung Hee University)
Reinforcement LearningImageVideoBenchmark
🎯 What it does: A visual reinforcement learning method for adaptive adversarial enhancement of Fourier magnitudes in the frequency domain (FGA3) is proposed, which maintains image semantic consistency during training and significantly improves robustness in unseen environments.
FR²Seg: Continual Segmentation Across Multiple Sites via Fourier Style Replay and Adaptive Consistency Regularization
Cheng Xu (South China University of Technology), Jing Qin (The Hong Kong Polytechnic University)
SegmentationDomain AdaptationConvolutional Neural NetworkContrastive LearningBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A continuous segmentation framework FR 2 Seg based on Fourier transform is designed, utilizing low-frequency information to achieve domain-specific and domain-invariant knowledge learning;
Frame Order Matters: A Temporal Sequence-Aware Model for Few-Shot Action Recognition
Bozheng Li (Zhejiang University), Yunlong Yu (Zhejiang University)
RecognitionTransformerLarge Language ModelContrastive LearningVideoText
🎯 What it does: This paper proposes a Temporal Sequence-Aware Model (TSAM) that achieves few-shot action recognition by incorporating a sequence-aware Perceiver Adapter into a pre-trained CLIP visual encoder, along with an expanded text corpus and unbalanced optimal transport matching.
Free Lunch in the Forest: Functionally-Identical Pruning of Boosted Tree Ensembles
Youssouf Emine (Polytechnique Montreal), Thibaut Vidal (Polytechnique Montreal)
CompressionOptimizationSupervised Fine-TuningTabular
🎯 What it does: This paper proposes a complete method for faithful pruning of additive tree ensemble models while ensuring the invariance of the prediction function, and presents an iterative FIPE algorithm.
Free-Moving Object Reconstruction and Pose Estimation with Virtual Camera
Haixin Shi (École Polytechnique Fédérale de Lausanne), David Ferstl (Magic Leap)
Object DetectionPose EstimationSimultaneous Localization and MappingVideo
🎯 What it does: This paper studies how to simultaneously reconstruct and estimate the pose of freely moving objects in monocular RGB videos without any prior knowledge of object categories or gestures.
FreeCap: Hybrid Calibration-Free Motion Capture in Open Environments
Aoru Xue (ShanghaiTech University), Yuexin Ma (ShanghaiTech University)
Pose EstimationMultimodalityPoint Cloud
🎯 What it does: We propose FreeCap, a calibration-free hybrid motion capture system that utilizes a single LiDAR and scalable mobile cameras to accurately capture global human motion of multiple people in open environments.
FreeGen: Bridging Visual-Linguistic Discrepancies Towards Diffusion-based Pixel-level Data Synthesis
Wenzhuang Wang (Beihang University), Jia Li (Beihang University)
SegmentationGenerationData SynthesisDiffusion modelImageText
🎯 What it does: This paper presents FreeGen, a self-driven text-to-pixel data generation framework based on diffusion models, which corrects visual-language inconsistencies through two-stage training and automatically generates open vocabulary segmentation data.
FreeMask: Rethinking the Importance of Attention Masks for Zero-Shot Video Editing
Lingling Cai (Zhejiang University), Kejie Huang (Tongyi Lab)
SegmentationGenerationTransformerDiffusion modelVideo
🎯 What it does: The FreeMask method is proposed, which selects the best cross-attention mask through adaptive Mask Matching Cost, achieving more precise structural control in zero-shot video editing.
FreeNet: Liberating Depth-Wise Separable Operations for Building Faster Mobile Vision Architectures
Hao Yu (University of Oulu), Guoying Zhao (Aalto University)
ClassificationObject DetectionSegmentationComputational EfficiencyKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: A novel and efficient visual backbone network called FreeNet is proposed, which completely removes segmented operations such as depthwise separable convolutions. It uses Sparse Continuous Convolution (SWConv) to construct the S2-Mixer for token mixing and employs a Shift Feed-Forward Network (ShiftFFN) to replace the traditional FFN, achieving faster channel mixing.
FreqTS: Frequency-Aware Token Selection for Accelerating Diffusion Models
Xinye Yang (Newcastle University), Luking Li (Independent Researcher)
GenerationComputational EfficiencyHyperparameter SearchDiffusion modelImage
🎯 What it does: A frequency-domain based token selection method (FreqTS) is proposed to accelerate the inference of diffusion models without the need for retraining.
Frequency-Aware Density Control via Reparameterization for High-Quality Rendering of 3D Gaussian Splatting
Zhaojie Zeng (Huazhong University of Science and Technology), Tao Guan (University of South Carolina)
GenerationData SynthesisOptimizationGaussian SplattingPoint Cloud
🎯 What it does: This paper proposes FDS-GS, which achieves frequency-aware density control in 3D Gaussian Splatting by reparameterizing scale and associating it with density.
Frequency-Masked Embedding Inference: A Non-Contrastive Approach for Time Series Representation Learning
En Fu (University of Science and Technology Beijing), Yanyan Hu (University of Science and Technology Beijing)
ClassificationRepresentation LearningContrastive LearningTime Series
🎯 What it does: A novel non-contrastive learning framework called Frequency-masked Embedding Inference (FEI) is proposed, which performs embedding inference through frequency masking prompts, eliminating the need for positive and negative sample pairs.
Friends-MMC: A Dataset for Multi-modal Multi-party Conversation Understanding
Yueqian Wang (Peking University), Dongyan Zhao (Peking University)
RecognitionOptimizationConvolutional Neural NetworkTransformerSupervised Fine-TuningVideoTextMultimodality
🎯 What it does: This paper studies multimodal multi-party conversation (MMC) understanding, constructs the Friends-MMC dataset, and proposes two fundamental tasks: speaker identification and response prediction.
FriendsQA: A New Large-Scale Deep Video Understanding Dataset with Fine-grained Topic Categorization for Story Videos
Zhengqian Wu (Wuhan University), Chao Liang (Wuhan University)
Large Language ModelAgentic AIVideoText
🎯 What it does: This study proposes the StoryMind framework, which utilizes multi-agent LLMs to automatically generate and review a large-scale deep video understanding dataset called FriendsQA, and evaluates 10 top VideoQA models on it.