AAAI 2024 Papers — Page 23
AAAI Conference on Artificial Intelligence · 2331 papers
Unify Named Entity Recognition Scenarios via Contrastive Real-Time Updating Prototype
Yanhe Liu (Southeast University), Ziyu Shang (Nanjing University of Finance and Economics)
RecognitionTransformerContrastive LearningText
🎯 What it does: A unified continuous named entity recognition framework CRUP is proposed, capable of handling supervised, class-incremental, and online scenarios simultaneously.
Unifying Decision and Function Queries in Stochastic Boolean Satisfiability
Yu-Wei Fan (National Taiwan University), Jie-Hong R. Jiang (National Taiwan University)
🎯 What it does: A new formalization of SSAT(Θ) is proposed, which incorporates threshold quantifiers into the prefix of the original SSAT to unify decision-making and function queries, along with the corresponding solver ClauSSat(Θ).
Unifying Multi-Modal Uncertainty Modeling and Semantic Alignment for Text-to-Image Person Re-identification
Zhiwei Zhao (University of Science and Technology of China), Nenghai Yu (University of Science and Technology of China)
RecognitionRetrievalTransformerContrastive LearningImageTextMultimodality
🎯 What it does: This paper models image and text features as Gaussian distributions and jointly estimates multi-granularity uncertainty through batch-level and identity-level variance, enhancing feature sampling. It then proposes a bidirectional cross-modal circular loss and a cross-modal global semantic recovery task, forming a unified end-to-end framework to achieve better semantic alignment between images and texts.
Unifying Visual and Vision-Language Tracking via Contrastive Learning
Yinchao Ma (University of Science and Technology of China), Mengxue Kang (Intelligent Science Technology Academy of CASIC)
Object TrackingTransformerContrastive LearningVideoMultimodality
🎯 What it does: A unified visual and visual-language tracking framework UVLTrack is proposed, which can simultaneously support three target reference methods: BBOX, NL, and NL+BBOX.
UniGen: A Unified Generative Framework for Retrieval and Question Answering with Large Language Models
Xiaoxi Li (Renmin University of China), Zhicheng Dou (Renmin University of China)
GenerationRetrievalTransformerLarge Language ModelSupervised Fine-TuningTextRetrieval-Augmented Generation
🎯 What it does: A unified generative framework called UniGen is proposed, capable of simultaneously performing retrieval (Generative Document Retrieval) and question-answering (Grounded Answer Generation) tasks.
Union Subgraph Neural Networks
Jiaxing Xu (Nanyang Technological University), Yiping Ke (Nanyang Technological University)
ClassificationRepresentation LearningGraph Neural NetworkGraphBenchmark
🎯 What it does: This paper proposes the Union Subgraph Neural Network (UnionSNN), which enhances the expressive power of GNNs by combining the unified subgraph of 1-hop neighbors with the shortest path matrix.
Unit Selection with Nonbinary Treatment and Effect
Ang Li (Florida State University), Judea Pearl (University of California)
TabularBiomedical Data
🎯 What it does: A unit selection benefit function for non-binary treatments and outcomes is proposed, along with algorithms for identifiability testing and boundary calculation.
United We Stand: Accelerating Privacy-Preserving Neural Inference by Conjunctive Optimization with Interleaved Nexus
Qiao Zhang (Chongqing University), Hongyi Wu (The University of Arizona)
OptimizationSafty and PrivacyComputational EfficiencyConvolutional Neural NetworkImage
🎯 What it does: The COIN framework is proposed, which significantly improves the efficiency of privacy-preserving neural network inference by jointly optimizing adjacent nonlinear and linear functions.
United We Stand: Using Epoch-Wise Agreement of Ensembles to Combat Overfit
Uri Stern (Hebrew University of Jerusalem), Daphna Weinshall (Hebrew University of Jerusalem)
ClassificationOptimizationConvolutional Neural NetworkImageText
🎯 What it does: A MAP (Max Agreement Prediction) method based on deep network ensemble is proposed, which utilizes the prediction consistency of models across different training epochs to mitigate overfitting, showing excellent performance especially on datasets with label noise.
Universal Weak Coreset
Ragesh Jaiswal (Indian Institute of Technology Delhi), Amit Kumar (Indian Institute of Technology Delhi)
CompressionOptimization
🎯 What it does: This study constructs a universal weak core set for data compression in constrained k-means/mean clustering, with the size of the core set independent of the data scale.
Unknown-Aware Graph Regularization for Robust Semi-supervised Learning from Uncurated Data
Heejo Kong (Korea University), Seong-Whan Lee (Korea University)
ClassificationAnomaly DetectionGraph Neural NetworkContrastive LearningImage
🎯 What it does: Proposes Unknown Awareness Graph Regularization (UAG) for open semi-supervised learning, which can robustly learn from unlabeled data containing unknown classes.
Unlocking the Power of Open Set: A New Perspective for Open-Set Noisy Label Learning
Wenhai Wan (Nanjing University of Aeronautics and Astronautics), Songcan Chen (Nanjing University of Aeronautics and Astronautics)
ClassificationData-Centric LearningContrastive LearningImage
🎯 What it does: Proposes the CECL method, which combines class extension and contrastive learning to handle open set noisy label learning.
Unraveling Batch Normalization for Realistic Test-Time Adaptation
Zixian Su (Xi'an Jiaotong-Liverpool University), Kaizhu Huang (Duke Kunshan University)
Domain AdaptationImageBenchmark
🎯 What it does: A test-time adaptive method based on batch normalization is proposed, utilizing Test-time Exponential Moving Average (TEMA) and hierarchical normalization correction to enhance the model's robustness under different batch sizes.
Unravelling Expressive Delegations: Complexity and Normative Analysis
Giannis Tyrovolas (Independent), Edith Elkind (University of Oxford)
Optimization
🎯 What it does: This paper proposes and analyzes the binary issue decision-making using two 'unrolling' strategies, MINSUM and MINMAX, under a hierarchical flow democracy model with available Boolean functions, and provides a complete complexity dichotomy and efficient algorithms.
Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms
Zheng Xing (Chinese University of Hong Kong), Weibing Zhao (Shenzhen MSU-BIT University)
RecognitionSegmentationVideo
🎯 What it does: This paper proposes a completely unsupervised action segmentation framework called Split-and-Merge (SaM), which identifies local minima by calculating the subspace similarity between adjacent frames to partition them into semantically consistent actoms, and then merges actoms of the same action using spatio-temporal similarity, ultimately achieving complete video action segmentation results.
Unsupervised Continual Anomaly Detection with Contrastively-Learned Prompt
Jiaqi Liu (Southern University of Science and Technology), Feng Zheng (Tencent Youtu Lab)
Anomaly DetectionTransformerContrastive LearningImage
🎯 What it does: This paper proposes an unsupervised continuous anomaly detection framework called UCAD, which utilizes a Continuous Prompt Module (CPM) and Structural Contrastive Learning (SCL) to achieve anomaly detection and segmentation in a single model that is task-agnostic.
Unsupervised Cross-Domain Image Retrieval via Prototypical Optimal Transport
Bin Li (ShanghaiTech University), Jingya Wang (ShanghaiTech University)
RetrievalDomain AdaptationContrastive LearningImage
🎯 What it does: This paper proposes a prototype-based optimal transport framework, ProtoOT, to address the unsupervised cross-domain image retrieval task, balancing intra-domain feature learning and inter-domain alignment.
Unsupervised Domain Adaptative Temporal Sentence Localization with Mutual Information Maximization
Daizong Liu (Peking University), Yu Cheng (Chinese University of Hong Kong)
Domain AdaptationTransformerContrastive LearningVideoTextMultimodality
🎯 What it does: The paper proposes an unsupervised domain adaptation method for video sentence localization, which can utilize labeled knowledge from the source domain to perform localization in an unlabeled target domain.
Unsupervised Extractive Summarization with Learnable Length Control Strategies
Renlong Jie (Northwestern Polytechnical University), Qun Liu (Huawei)
TransformerContrastive LearningText
🎯 What it does: Proposes an unsupervised extractive summarization model based on Siamese networks, achieving end-to-end training by combining bidirectional prediction objectives, and designs a differentiable 0-1 knapsack Transformer for length-controllable extraction;
Unsupervised Gene-Cell Collective Representation Learning with Optimal Transport
Jixiang Yu (City University of Hong Kong), Ka-Chun Wong
Representation LearningGraph Neural NetworkTransformerAuto EncoderBiomedical Data
🎯 What it does: An end-to-end unsupervised gene-cell ensemble representation learning and optimal transport (OT) framework called scGCOT is proposed for cell type clustering of single-cell RNA sequencing data.
Unsupervised Group Re-identification via Adaptive Clustering-Driven Progressive Learning
Hongxu Chen (Sun Yat-Sen University), Xiaohua Xie (Sun Yat-Sen University)
RecognitionRetrievalConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: This paper proposes an unsupervised group re-identification method called ACPL, which utilizes adaptive clustering and dynamic prototype updates to generate high-quality pseudo-labels and enhance model robustness.
Unsupervised Layer-Wise Score Aggregation for Textual OOD Detection
Maxime Darrin (International Laboratory on Learning Systems), Pierre Colombo (Universite Paris-Saclay)
Anomaly DetectionTransformerTextBenchmark
🎯 What it does: An unsupervised hierarchical score aggregation method is proposed, which automatically combines the anomaly scores from all encoder layers to enhance text OOD detection performance.
Unsupervised Neighborhood Propagation Kernel Layers for Semi-supervised Node Classification
Sonny Achten (KU Leuven), Johan A.K. Suykens
ClassificationRepresentation LearningGraph Neural NetworkAuto EncoderGraph
🎯 What it does: A deep graph convolution kernel machine (GCKM) is proposed, which achieves recursive aggregation and classification of node features through multi-layer unsupervised kernel PCA and semi-supervised kernel spectral clustering.
Unsupervised Object Interaction Learning with Counterfactual Dynamics Models
Jongwook Choi (University of Michigan), Honglak Lee (University of Michigan)
Robotic IntelligenceReinforcement LearningSequential
🎯 What it does: The COIL method is proposed, which uses counterfactual reasoning to provide intrinsic rewards for object states, enabling the learning of unsupervised skills for object interactions without external rewards.
Unsupervised Pan-Sharpening via Mutually Guided Detail Restoration
Huangxing Lin (National University of Defense Technology), Yongxiang Liu (National University of Defense Technology)
RestorationSuper ResolutionConvolutional Neural NetworkImage
🎯 What it does: This paper proposes an unsupervised mutual guidance detail recovery framework PAN-MGDR for the super-resolution fusion (pan-sharpening) of spectral images.
Unsupervised Training Sequence Design: Efficient and Generalizable Agent Training
Wenjun Li (Singapore Management University), Pradeep Varakantham (Singapore Management University)
Robotic IntelligenceMeta LearningReinforcement LearningAgentic AISequential
🎯 What it does: The Unsupervised Training Sequence Design (UTSD) framework is proposed, treating the teacher as a finite time-step MDP. It utilizes Quality Diversity (QD) methods to generate diverse validation environments and encodes student policies as teacher states; then, through context-aware meta reinforcement learning (PEARL), enables the teacher to quickly adapt to different students.
Unveiling Details in the Dark: Simultaneous Brightening and Zooming for Low-Light Image Enhancement
Ziyu Yue (Dalian University of Technology), Zhixun Su (Dalian University of Technology)
RestorationSuper ResolutionConvolutional Neural NetworkImage
🎯 What it does: This paper proposes BrZoNet, which aims to enhance brightness and perform super-resolution reconstruction on low-light low-resolution images simultaneously.
Unveiling Implicit Deceptive Patterns in Multi-Modal Fake News via Neuro-Symbolic Reasoning
Yiqi Dong (Tianjin University), Di Jin (Tianjin University)
ClassificationExplainability and InterpretabilityKnowledge DistillationConvolutional Neural NetworkAuto EncoderMultimodality
🎯 What it does: This paper proposes a Neuro-Symbolic Latent Model (NSLM) that can automatically infer the authenticity of multimodal fake news and reveal underlying forgery patterns (image tampering, cross-modal inconsistencies, image reuse) as interpretable outputs.
Unveiling the Significance of Toddler-Inspired Reward Transition in Goal-Oriented Reinforcement Learning
Junseok Park (Seoul National University), Byoung-Tak Zhang (Seoul National University)
Reinforcement Learning
🎯 What it does: Proposed a heuristic sparse-to-dense reward transfer (S2D) strategy for children and validated its effectiveness in various goal-oriented RL environments.
UPDP: A Unified Progressive Depth Pruner for CNN and Vision Transformer
Ji Liu (Advanced Micro Devices), Ashish Sirasao (Advanced Micro Devices)
Convolutional Neural NetworkTransformerImage
🎯 What it does: A unified deep pruning framework is proposed for hierarchical pruning of CNNs and vision Transformers.
Upper Bounding Barlow Twins: A Novel Filter for Multi-Relational Clustering
Xiaowei Qian (University of Electronic Science and Technology of China), Zhao Kang (University of Electronic Science and Technology of China)
Representation LearningGraph Neural NetworkAuto EncoderContrastive LearningGraph
🎯 What it does: This paper proposes a Barlow Twins-based Graph Filter (BTGF) for multi-relational graph clustering.
Urban Region Embedding via Multi-View Contrastive Prediction
Zechen Li (Shandong University), Meng Chen (Shandong University)
Representation LearningAuto EncoderContrastive LearningTabular
🎯 What it does: This paper proposes a multi-view urban area embedding model called ReCP based on a consistency learning paradigm, which jointly learns area representations using POI attribute views and human mobility views, emphasizing consistency between different views rather than just late-stage fusion.
Using Artificial Populations to Study Psychological Phenomena in Neural Models
Jesse Roberts (Vanderbilt University), Douglas Fisher (Cornell University)
TransformerLarge Language ModelText
🎯 What it does: This paper proposes PopulationLM, which constructs an artificial model family by applying hierarchical MC dropout on the Transformer, thereby systematically studying the cognitive behavior of language models in terms of typicality and structural derivation effects.
Using Clustering to Strengthen Decision Diagram Bounds for Discrete Optimization
Mohsen Nafar (Bielefeld University), Michael Römer (Bielefeld University)
OptimizationTabular
🎯 What it does: This paper proposes a clustering-based node selection method to determine which nodes need to be merged or discarded when constructing approximate decision diagrams (DD), thereby obtaining stronger upper and lower bounds.
Using Stratified Sampling to Improve LIME Image Explanations
Muhammad Rashid (University of Torino), Damiano Verda (Rulex Innovation Labs)
Object DetectionExplainability and InterpretabilityConvolutional Neural NetworkImage
🎯 What it does: This study investigates the use of stratified sampling in LIME Image as a replacement for Monte Carlo sampling to reduce the inaccuracies in explanations caused by downsampling.
Using Symmetries to Lift Satisfiability Checking
Pierre Carbonnelle (KU Leuven), Marc Denecker (KU Leuven)
OptimizationComputational EfficiencyTabular
🎯 What it does: This paper proposes a two-step method for 'lifting' structures using symmetry: ① Automatically translate the sentence to be satisfied into a sentence that is equivalent and satisfiable on a 'lifting vocabulary', thereby achieving domain compression; ② Use SAT/SMT solvers to solve the lifted sentence on the compressed smaller domain, gradually expanding the domain until a satisfying model is obtained; the expanded structure can be restored to the original structure.
UVAGaze: Unsupervised 1-to-2 Views Adaptation for Gaze Estimation
Ruicong Liu (Beihang University), Feng Lu (Beihang University)
Pose EstimationDomain AdaptationConvolutional Neural NetworkImage
🎯 What it does: An unsupervised perspective adaptation framework UVAGaze is proposed, which can adapt and infer traditional single-view gaze estimation models under any placement of dual-camera configurations.
V2A-Mapper: A Lightweight Solution for Vision-to-Audio Generation by Connecting Foundation Models
Heng Wang (University of Sydney), Weidong Cai (Dolby Laboratories)
GenerationData SynthesisTransformerDiffusion modelImageVideoMultimodalityAudio
🎯 What it does: A lightweight V2A-Mapper is proposed, which maps CLIP visual representations to CLAP audio space, and then uses AudioLDM to generate high-quality sounds that are consistent with the visuals.
V2Meow: Meowing to the Visual Beat via Video-to-Music Generation
Kun Su (University of Washington), Timo Denk (Google DeepMind)
GenerationData SynthesisTransformerVideoMultimodalityAudio
🎯 What it does: Proposes the V2Meow video-to-music generation system, which can synthesize high-quality music audio from various video inputs;
VadCLIP: Adapting Vision-Language Models for Weakly Supervised Video Anomaly Detection
Peng Wu (Northwestern Polytechnical University), Yanning Zhang (Northwestern Polytechnical University)
Anomaly DetectionTransformerVision Language ModelContrastive LearningVideo
🎯 What it does: This paper proposes VadCLIP, a dual-branch weakly supervised video anomaly detection framework that utilizes a frozen CLIP model to directly perform video feature extraction and anomaly detection without the need for pre-training or fine-tuning.
Value at Adversarial Risk: A Graph Defense Strategy against Cost-Aware Attacks
Junlong Liao (Fudan University), Jiarong Xu (Fudan University)
OptimizationAdversarial AttackGraph Neural NetworkGraph
🎯 What it does: In response to the vulnerability of graph neural networks when facing attacks, a cost-based graph defense framework called RisKeeper is proposed, which enhances robustness against structural attacks by learning node costs.
Value Kaleidoscope: Engaging AI with Pluralistic Human Values, Rights, and Duties
Taylor Sorensen (University of Washington), Yejin Choi (University of Washington)
GenerationExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper first constructs a large-scale multi-value dataset called ValuePrism, which contains 218k entries of values, rights, and obligations for 31k real-life scenarios. Based on this dataset, the Value Kaleidoscope (Kaleido) model is trained to generate, explain, and evaluate the relevance and positive/negative tendencies of values, rights, and obligations in given contexts.
Variable Importance in High-Dimensional Settings Requires Grouping
Ahmad Chamma (Inria), Denis Engemann (Roche Pharma Research and Early Development)
Explainability and InterpretabilityComputational EfficiencyBiomedical DataMagnetic Resonance Imaging
🎯 What it does: This study investigates variable importance interpretation in high-dimensional environments, proposing the Block-Based Conditional Permutation Importance (BCPI) framework and introducing internal stacking to enhance computational efficiency.
Variance-Insensitive and Target-Preserving Mask Refinement for Interactive Image Segmentation
Chaowei Fang (Xidian University), Guanbin Li (Sun Yat-sen University)
SegmentationTransformerSupervised Fine-TuningImage
🎯 What it does: A variance-insensitive and target-preserving mask refinement method is proposed, achieving click-based interactive image segmentation through mask matching regularization and target-aware scaling techniques.
Variational Hybrid-Attention Framework for Multi-Label Few-Shot Aspect Category Detection
Cheng Peng (Zhejiang University), Gang Chen (Zhejiang University)
ClassificationTransformerSupervised Fine-TuningText
🎯 What it does: A Variational Hybrid Attention Framework (VHAF) is proposed to address the category detection problem in multi-label few-shot learning.
VELMA: Verbalization Embodiment of LLM Agents for Vision and Language Navigation in Street View
Raphael Schumann (Heidelberg University), William Yang Wang (University of California)
TransformerLarge Language ModelAgentic AIPrompt EngineeringText
🎯 What it does: A LLM-based entity agent named VELMA is proposed, capable of completing visual language navigation tasks in real Street View environments through natural language descriptions of environmental observations and trajectories.
Video Event Extraction with Multi-View Interaction Knowledge Distillation
Kaiwen Wei (Chongqing University), Zhi Guo (Kuaishou Technology Inc.)
Object DetectionKnowledge DistillationTransformerVideo
🎯 What it does: Proposes the MID framework, which utilizes self-relation knowledge distillation (self-RKD) and layer-wise knowledge distillation (LKD) to enhance inter-object interactions and inter-modal interactions in video event extraction, improving verb classification and semantic role prediction;
Video Frame Prediction from a Single Image and Events
Juanjuan Zhu (Northwestern Polytechnical University), Yuchao Dai (Northwestern Polytechnical University)
GenerationData SynthesisOptical FlowImageVideo
🎯 What it does: This paper proposes a video frame prediction model based on a single RGB image and subsequent event streams, capable of generating high-quality future frames in real-time over any time interval;
Video-Context Aligned Transformer for Video Question Answering
Linlin Zong (Dalian University of Technology), Bo Xu (Dalian University of Technology)
TransformerContrastive LearningVideoText
🎯 What it does: This paper proposes a Video-Context Aligned Transformer (V-CAT), which introduces video-related questions as context to first achieve semantic and content alignment between video and text, and then performs fine-grained interaction to generate answers.
VIGC: Visual Instruction Generation and Correction
Bin Wang (Shanghai Artificial Intelligence Laboratory), Conghui He (Sun Yat-sen University)
GenerationData SynthesisTransformerLarge Language ModelVision Language ModelTextMultimodality
🎯 What it does: A visual instruction generation and correction framework (VIGC) based on a multimodal large language model is proposed, utilizing two submodules: Visual Instruction Generation (VIG) and Visual Instruction Correction (VIC) to automatically generate high-quality visual-text instruction fine-tuning data.
ViLT-CLIP: Video and Language Tuning CLIP with Multimodal Prompt Learning and Scenario-Guided Optimization
Hao Wang (Xidian University), Xu Liu (Xidian University)
RecognitionRetrievalOptimizationTransformerPrompt EngineeringContrastive LearningVideoMultimodality
🎯 What it does: A method called ViLT-CLIP is proposed, which adapts the image-based CLIP model to video-specific tasks, including video recognition and video-text retrieval, through multimodal prompt learning and scene-guided optimization.
Vision Transformer Off-the-Shelf: A Surprising Baseline for Few-Shot Class-Agnostic Counting
Zhicheng Wang (Huazhong University of Science and Technology), Hao Lu (Huazhong University of Science and Technology)
Object DetectionRepresentation LearningTransformerImage
🎯 What it does: A single network CACViT based on Vision Transformer (ViT) is proposed, which directly performs feature extraction and matching in self-attention, constructing a simple and efficient few-shot class-agnostic counting baseline.
Vision-Language Pre-training with Object Contrastive Learning for 3D Scene Understanding
Taolin Zhang (Tsinghua University), Shu-Tao Xia (Harbin Institute of Technology)
RecognitionObject DetectionTransformerVision Language ModelContrastive LearningMultimodalityPoint Cloud
🎯 What it does: A 3DVLP pre-training framework is proposed, achieving unified representation and transfer of 3D vision-language through object-level contrastive learning.
ViSTec: Video Modeling for Sports Technique Recognition and Tactical Analysis
Yuchen He (Zhejiang University), Yingcai Wu (Zhejiang University)
ClassificationRecognitionTransformerVideo
🎯 What it does: This study proposes the ViSTec model for recognizing tennis/ping pong techniques and tactical analysis in broadcast videos, combining sparse visual features with contextual knowledge through a two-stage segmentation-classification framework.
Visual Chain-of-Thought Prompting for Knowledge-Based Visual Reasoning
Zhenfang Chen (MIT-IBM Watson Artificial Intelligence Lab), Chuang Gan (MIT-IBM Watson Artificial Intelligence Lab)
Object DetectionComputational EfficiencyKnowledge DistillationTransformerLarge Language ModelPrompt EngineeringVision Language ModelImageTextChain-of-Thought
🎯 What it does: A Visual Chain Thinking Prompt (VCTP) framework is proposed, which implements knowledge-based visual reasoning through an iterative see-think-confirm three-stage interaction.
Visual Hallucination Elevates Speech Recognition
Fang Zhang (University of Science and Technology of China), Linli Xu (University of Science and Technology of China)
RecognitionGenerationTransformerMultimodalityAudio
🎯 What it does: This paper proposes a discrete feature visual generation model (DFVGM) to generate visual illusions based on audio in the absence of visual input, aiming to enhance the robustness of multimodal speech recognition.
Visual Instruction Tuning with Polite Flamingo
Delong Chen (Xiaobing AI), Baoyuan Wang (Xiaobing AI)
GenerationData SynthesisTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodality
🎯 What it does: This paper proposes Polite Flamingo, a multimodal response rewriter that transforms short, unformatted answers from the original visual-language dataset into polite, structured, high-quality replies, thereby generating the PF-1M dataset for training multimodal LLMs (Clever Flamingo).
Visual Redundancy Removal for Composite Images: A Benchmark Dataset and a Multi-Visual-Effects Driven Incremental Method
Miaohui Wang (Shenzhen University), Yanshan Li (Shenzhen University)
CompressionConvolutional Neural NetworkImageBenchmark
🎯 What it does: This paper constructs the first composite image visual redundancy prediction database ciVRP-Set and proposes an incremental learning framework mveVRP driven by multiple visual effects to accurately predict and remove visual redundancy in composite images.
ViT-Calibrator: Decision Stream Calibration for Vision Transformer
Lin Chen (Zhejiang University), Zunlei Feng (Zhejiang University)
ClassificationExplainability and InterpretabilityKnowledge DistillationTransformerImage
🎯 What it does: A ViT-Calibrator is designed to enhance the classification performance of Vision Transformers through decision flow calibration (token-level feedback + dimensional constraints).
VITA: ‘Carefully Chosen and Weighted Less’ Is Better in Medication Recommendation
Taeri Kim (Hanyang University), Sang-Wook Kim (Hanyang University)
Recommendation SystemDrug DiscoveryGraph Neural NetworkTransformerBiomedical DataElectronic Health Records
🎯 What it does: This paper proposes a framework for drug recommendation called VITA, which automatically selects the most relevant historical visit records related to the current visit and weights them to more accurately represent the patient's current health status and recommend medications.
ViTEraser: Harnessing the Power of Vision Transformers for Scene Text Removal with SegMIM Pretraining
Dezhi Peng (South China University of Technology), Lianwen Jin (Huazhong University of Science and Technology)
RestorationSegmentationTransformerGenerative Adversarial NetworkImage
🎯 What it does: We propose ViTEraser, a single-stage scene text removal model based on Vision Transformer, and design the SegMIM pre-training method.
ViTree: Single-Path Neural Tree for Step-Wise Interpretable Fine-Grained Visual Categorization
Danning Lao (Shanghai Jiao Tong University), Wei Shen (Shanghai Jiao Tong University)
ClassificationExplainability and InterpretabilityRepresentation LearningTransformerImage
🎯 What it does: A single-path neural tree called ViTree is designed and implemented for fine-grained visual classification, combining visual Transformers for feature extraction, achieving progressive representation learning through hard patch selection in the tree, and ultimately outputting classification results via a single path.
VIXEN: Visual Text Comparison Network for Image Difference Captioning
Alexander Black (University of Surrey), John Collomosse (University of Surrey)
RecognitionGenerationTransformerLarge Language ModelContrastive LearningImageText
🎯 What it does: VIXEN is an image difference subtitle generation model that can describe the visual changes between two images with brief text, helping to identify forged content.
VLCounter: Text-Aware Visual Representation for Zero-Shot Object Counting
Seunggu Kang (Sungkyunkwan University), Jae-Pil Heo (Sungkyunkwan University)
Object DetectionVision Language ModelContrastive LearningImage
🎯 What it does: A one-stage zero-shot object counting framework called VLBase is proposed, and based on it, semantic condition prompt tuning (SPT), learnable affine transformation (LAT), and segmented skip connections (SaSC) are added to construct VLCounter, eliminating the traditional two-stage example discovery process.
VLM2Scene: Self-Supervised Image-Text-LiDAR Learning with Foundation Models for Autonomous Driving Scene Understanding
Guibiao Liao (Peking University), Xiaoqing Ye (Baidu Inc.)
SegmentationAutonomous DrivingTransformerVision Language ModelContrastive LearningImagePoint Cloud
🎯 What it does: This paper proposes VLM2Scene, which achieves unsupervised training of 3D scene representation through self-supervised learning by combining CLIP, BLIP-2, and SAM in image-text-lidar contrastive learning.
VLN-Video: Utilizing Driving Videos for Outdoor Vision-and-Language Navigation
Jialu Li (University of North Carolina), Mohit Bansal (University of North Carolina)
Data SynthesisAutonomous DrivingTransformerVision Language ModelVideoText
🎯 What it does: Utilize driving videos to generate a large amount of outdoor visual-language navigation (VLN) training data, and after pre-training the model on this data, fine-tune it on the Touchdown dataset to achieve better navigation performance.
VMT-Adapter: Parameter-Efficient Transfer Learning for Multi-Task Dense Scene Understanding
Yi Xin (Nanjing University), Ke Yan (Nanjing University)
SegmentationTransformerSupervised Fine-TuningImage
🎯 What it does: A parameter-efficient multi-task adapter named VMT-Adapter is proposed, specifically designed for transfer learning in dense scene understanding tasks.
Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection
Yuhao Huang (Xi'an Jiaotong University), Nanning Zheng (Xi'an Jiaotong University)
Object DetectionAutonomous DrivingPoint Cloud
🎯 What it does: A Voxel-Pillar Fusion (VPF) hybrid voxel-pillar encoding framework is proposed, which utilizes sparse convolution to simultaneously extract voxel and pillar features, and achieves bidirectional interaction through a sparse fusion layer to enhance vertical representation.
VPDETR: End-to-End Vanishing Point DEtection TRansformers
Taiyan Chen (Peking University), Ji Shi (Peking University)
Object DetectionDepth EstimationTransformerImage
🎯 What it does: An end-to-end transformer framework VPDETR has been developed for detecting vanishing points in images without the need for line segment extraction or candidate point sampling.
VQ-FONT: Few-Shot Font Generation with Structure-Aware Enhancement and Quantization
Mingshuai Yao (Harbin Institute of Technology), Wangmeng Zuo (Harbin Institute of Technology)
GenerationTransformerAuto EncoderGenerative Adversarial NetworkImage
🎯 What it does: A few-shot font generation framework called VQ-Font has been developed, which is based on VQGAN codebook and structure-level style enhancement, capable of transferring fine-grained strokes and structural styles from a small number of reference glyphs to the target glyph.
VQAttack: Transferable Adversarial Attacks on Visual Question Answering via Pre-trained Models
Ziyi Yin (Pennsylvania State University), Fenglong Ma (Pennsylvania State University)
Adversarial AttackTransformerLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: A transferable adversarial attack method based on pre-trained vision-language models, VQAttack, is proposed to generate adversarial samples that perturb both images and text without accessing the target VQA model.
VQCNIR: Clearer Night Image Restoration with Vector-Quantized Codebook
Wenbin Zou (South China University of Technology), Sixiang Chen (Hong Kong University of Science and Technology)
RestorationGenerative Adversarial NetworkImage
🎯 What it does: A night scene image restoration framework based on vector quantization codebook, VQCNIR, is proposed, along with the design of an Adaptive Illumination Enhancement Module and a Deformable Bi-directional Cross-Attention module to enhance the details and lighting consistency of low-light and blurred images.
VSFormer: Visual-Spatial Fusion Transformer for Correspondence Pruning
Tangfei Liao (Wenzhou University), Guobao Xiao (Tongji University)
Pose EstimationTransformerImage
🎯 What it does: A visual-spatial fusion Transformer (VSFormer) is proposed for two-view correspondence pruning and camera pose estimation, significantly improving performance in high outlier rate scenarios.
VVS: Video-to-Video Retrieval with Irrelevant Frame Suppression
Won Jo (Sejong University), Yukyung Choi (Sejong University)
RetrievalTransformerVideo
🎯 What it does: A VVS framework is proposed, which first removes easily recognizable interference frames, then generates suppression weights based on temporal saliency and thematic relevance, ultimately producing video-level features to enhance content video retrieval performance.
W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation
Fangyuan Zhang (Tsinghua University), Bin Wang (Tsinghua University)
SegmentationImage
🎯 What it does: A new Weak to Partial Supervision (W2P) framework is proposed, which utilizes pseudo-labels generated from image-level labels to separate reliable 'clean' labels from unreliable 'noise' labels through Boundary Preserving Noise Detection (BPND). Based on this, a Partial Supervised Learning (PSL) module combined with a teacher-student model is used to gradually correct the noise, further enhancing boundary segmentation quality through Boundary Preserving Noise Correction (BPNC) and boundary generation strategies.
Wasserstein Differential Privacy
Chengyi Yang (Shanghai Institute of AI for Education), Aimin Zhou (Shanghai Institute of AI for Education)
Safty and PrivacyConvolutional Neural NetworkImage
🎯 What it does: A differential privacy framework WDP based on Wasserstein distance is proposed, and a Wasserstein accountant is implemented for privacy budget calculation in deep learning.
Watch Your Head: Assembling Projection Heads to Save the Reliability of Federated Models
Jinqian Chen (Xi'an Jiaotong University), Zhiqiang Tian (Xi'an Jiaotong University)
Federated LearningImage
🎯 What it does: This study investigates the reliability issues of federated learning under non-IID data and proposes an Assembly of Projection Heads (APH) to enhance model calibration and uncertainty estimation.
Watermarking Conditional Text Generation for AI Detection: Unveiling Challenges and a Semantic-Aware Watermark Remedy
Yu Fu (University of California Riverside), Yue Dong (Tianjin University)
GenerationData SynthesisTransformerLarge Language ModelText
🎯 What it does: A semantic-aware watermarking algorithm is proposed, which constructs a green word list by combining semantically similar words from the input context in conditional text generation tasks, achieving a balance between watermark injection and generation quality across multiple models.
WaveFormer: Wavelet Transformer for Noise-Robust Video Inpainting
Zhiliang Wu (Zhejiang University), Yan Yan (Cisco Research)
RestorationTransformerVideo
🎯 What it does: A Transformer network called WaveFormer based on discrete wavelet transform is proposed for video restoration in noisy environments.
Wavelet Dynamic Selection Network for Inertial Sensor Signal Enhancement
Yifeng Wang (Harbin Institute of Technology), Yi Zhao (Harbin Institute of Technology)
Pose EstimationConvolutional Neural NetworkGenerative Adversarial NetworkTime Series
🎯 What it does: A waveform dynamic selection network (WDSNet) is designed to adaptively select wavelet bases based on inertial sensor signals, significantly improving signal quality and subsequent task performance.
Wavelet-Driven Spatiotemporal Predictive Learning: Bridging Frequency and Time Variations
Xuesong Nie (Zhejiang University), Donglian Qi (Westlake University)
Autonomous DrivingOptimizationConvolutional Neural NetworkVideo
🎯 What it does: A WaST framework based on 3D wavelet transform is proposed, utilizing a time-frequency aware decoder to learn spatial-temporal dependencies in multi-scale wavelet space, thereby achieving non-recursive video prediction.
WaveNet: Tackling Non-stationary Graph Signals via Graph Spectral Wavelets
Zhirui Yang (Renmin University of China), Yong Liu (Renmin University of China)
Graph Neural NetworkImageGraph
🎯 What it does: This paper proposes a spectral domain graph neural network named WaveNet, which constructs non-polynomial filters using multi-resolution analysis (MRA) and Haar wavelet scaling functions to effectively capture the high-frequency components of graph spectral signals and overcome the limitations of traditional polynomial filters.
Weak Distribution Detectors Lead to Stronger Generalizability of Vision-Language Prompt Tuning
Kun Ding (Chinese Academy of Sciences), Chunhong Pan (Chinese Academy of Sciences)
ClassificationDomain AdaptationTransformerPrompt EngineeringVision Language ModelImage
🎯 What it does: This paper proposes a method that utilizes OOD detection to dynamically fuse zero-shot and few-shot prompt learning classifiers during the testing phase, thereby enhancing the ability of visual-language models to generalize to base-new categories.
Weakly Supervised Few-Shot Object Detection with DETR
Chenbo Zhang (Fudan University), Shuigeng Zhou (Fudan University)
Object DetectionKnowledge DistillationTransformerImage
🎯 What it does: Under the condition of no bounding box annotations, a weakly supervised few-shot object detection method based on DETR, called WFS-DETR, is proposed, enabling the model to achieve object localization and detection using only image-level labels.
Weakly Supervised Multimodal Affordance Grounding for Egocentric Images
Lingjing Xu (Beihang University), Aimin Hao (Beihang University)
RecognitionObject DetectionSegmentationKnowledge DistillationTransformerContrastive LearningImageTextMultimodality
🎯 What it does: A weakly supervised multimodal method WSMA is proposed, which transfers appearance human-computer interaction images and textual knowledge to first-person perspective images to locate functional areas.
Weakly Supervised Open-Vocabulary Object Detection
Jianghang Lin (Xiamen University), Liujuan Cao (Xiamen University)
Object DetectionVision Language ModelImage
🎯 What it does: A weakly supervised open vocabulary object detection framework WSOVOD is proposed, which can identify and locate unseen categories using only image-level labels and supports cross-dataset training.
Weakly Supervised Semantic Segmentation for Driving Scenes
Dongseob Kim (Yonsei University), Hyunjung Shim (Korea Advanced Institute of Science and Technology)
SegmentationAutonomous DrivingConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: This paper proposes a weakly supervised semantic segmentation framework for driving scenarios, utilizing CLIP to generate pseudo-masks and enhancing segmentation performance through global-local perspective training and Consistency-Aware Region Balancing (CARB).
Weakly-Supervised Mirror Detection via Scribble Annotations
Mingfeng Zha (University of Electronic Science and Technology of China), Heng Tao Shen (University of Electronic Science and Technology of China)
Object DetectionSegmentationTransformerImage
🎯 What it does: This paper proposes a weakly supervised mirror detection framework based on sparse graffiti annotations and constructs the first graffiti-style mirror dataset.
Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature
Wulian Yun (Beijing University of Posts and Telecommunications), Huadong Ma (Beijing University of Posts and Telecommunications)
RecognitionObject DetectionKnowledge DistillationVideo
🎯 What it does: A weakly supervised temporal action localization framework is proposed, which generates high-quality pseudo-labels by inferring significant segment features and uses these pseudo-labels to train the action localization network.
WeakPCSOD: Overcoming the Bias of Box Annotations for Weakly Supervised Point Cloud Salient Object Detection
Jun Wei (Chinese University of Hong Kong Shenzhen), Zhen Li (Chinese University of Hong Kong Shenzhen)
Object DetectionPoint Cloud
🎯 What it does: The first weakly supervised point cloud salient object detection model, WeakPCSOD, is proposed, significantly reducing annotation costs;
WebVLN: Vision-and-Language Navigation on Websites
Qi Chen (Australian Institute for Machine Learning), Qi Wu (Australian Institute for Machine Learning)
TransformerLarge Language ModelVision Language ModelTextMultimodality
🎯 What it does: A web-based visual and language navigation task called WebVLN is proposed, along with the corresponding WebVLN-v1 dataset and baseline model WebVLNNet.
WeditGAN: Few-Shot Image Generation via Latent Space Relocation
Yuxuan Duan (Shanghai Jiao Tong University), Liqing Zhang (Ant Group)
GenerationData SynthesisGenerative Adversarial NetworkContrastive LearningImage
🎯 What it does: We propose WeditGAN, a model transfer method for cross-domain few-shot image generation that learns a fixed latent space offset Δw.
Weighted Envy-Freeness for Submodular Valuations
Luisa Montanari (Technische Universitat Berlin), Nicholas Teh (National University of Singapore)
🎯 What it does: This paper proposes two new concepts of weighted fair division—TWEF(x,1−x) and WMEF(x,1−x), and proves that under submodular (especially matrix rank) preferences, these fairness criteria can be achieved through modified selection sequences, maximum weighted Nash social welfare, and new harmonic welfare rules.
Weisfeiler and Lehman Go Paths: Learning Topological Features via Path Complexes
Quang Truong (Dartmouth), Peter Chin (Dartmouth)
Graph Neural NetworkGraph
🎯 What it does: This paper proposes a graph isomorphism testing method based on path complexes (Path Weisfeiler–Lehman, PWL) and its implementation in graph networks called Path Complex Networks (PCN). By elevating graphs to a topological domain that only contains simple paths, it eliminates the assumptions of traditional high-order GNNs regarding substructures like cliques and cycles.
Well, Now We Know! Unveiling Sarcasm: Initiating and Exploring Multimodal Conversations with Reasoning
Gopendra Vikram Singh (Indian Institute of Technology Bombay), Pushpak Bhattacharyya (Indian Institute of Technology Bombay)
RecognitionGenerationTransformerLarge Language ModelVideoTextMultimodalityAudio
🎯 What it does: This paper proposes the task of identifying the starting point of sarcasm and generating reasons in multimodal dialogue (SIRC), and constructs the corresponding SIRD dataset;
What Are the Rules? Discovering Constraints from Data
Boris Wiegand (Stahl-Holding-Saar), Jilles Vreeken (CISPA Helmholtz Center for Information Security)
OptimizationTabular
🎯 What it does: Learn and mine constraint models that satisfy example solutions, treating it as a Minimum Description Length (MDL) problem, and propose a greedy algorithm URPILS to generate concise and noise-robust constraint sets.
What Do Hebbian Learners Learn? Reduction Axioms for Iterated Hebbian Learning
Caleb Schultz Kisby (Indiana University), Lawrence S. Moss (Indiana University)
🎯 What it does: This paper proposes a framework that maps dynamic logic operators [φ] to iterative Hebbian learning, providing corresponding reduction axioms and proving the completeness of this learning strategy.
What Does a Query Answer Tell You? Informativeness of Query Answers for Knowledge Bases
Luca Andolfi (Sapienza University of Rome), Maurizio Lenzerini (Sapienza University of Rome)
🎯 What it does: This paper proposes a formal framework to characterize the informativeness of query answers in knowledge bases, and based on this, evaluates existing methods and introduces a new n-connected answer form.
What Effects the Generalization in Visual Reinforcement Learning: Policy Consistency with Truncated Return Prediction
Shuo Wang (Beijing Jiaotong University), Kai Lv (Beijing Jiaotong University)
Robotic IntelligenceTransformerReinforcement LearningImage
🎯 What it does: This paper studies the generalization problem of visual reinforcement learning in unknown environments, proposing a Truncated Return Prediction (TRP) auxiliary task and implementing multi-policy cross-domain consistency using a Transformer network.
What Makes Good Collaborative Views? Contrastive Mutual Information Maximization for Multi-Agent Perception
Wanfang Su (Shanghai Jiao Tong University), Pan Zhou (Central South University)
Object DetectionCompressionAutonomous DrivingContrastive LearningMultimodalityPoint Cloud
🎯 What it does: A collaborative framework CMiMC is proposed to construct high-quality collaborative views by maximizing mutual information between intermediate collaborative views and individual views.