AAAI 2025 Papers — Page 27
AAAI Conference on Artificial Intelligence · 3028 papers
STAMPsy: Towards SpatioTemporal-Aware Mixed-Type Dialogues for Psychological Counseling
Jieyi Wang (Peking University), Yu Huang (Peking University)
Recommendation SystemTransformerLarge Language ModelSupervised Fine-TuningTextRetrieval-Augmented Generation
🎯 What it does: This paper proposes and implements a hybrid psychological counseling dialogue system, which first helps users clarify their counseling goals and then interacts based on different dialogue types (diagnosis, knowledge Q&A, recommendations, empathy, and Q&A).
Standing on the Shoulders of Giants: Reprogramming Visual-Language Model for General Deepfake Detection
Kaiqing Lin (Shenzhen University), Bin Li (Tencent)
Anomaly DetectionTransformerPrompt EngineeringVision Language ModelVideo
🎯 What it does: By adding learnable visual perturbations at the input end of the frozen CLIP model and face embedding-driven text prompts (Face2Text Prompt), the task of deepfake detection is reprogrammed;
State Encodings for GNN-Based Lifted Planners
Rostislav Horčik, Tomáš Pevný (Czech Technical University in Prague)
OptimizationComputational EfficiencyGraph Neural NetworkGraph
🎯 What it does: A systematic comparison of state encoding schemes in heuristic learning of graph neural networks (GNN) for classical planning tasks is conducted, evaluating their expressiveness and computational efficiency in offline search, and it is found that the minimal encoding based on Gaifman graphs performs best in the IPC 2023 learning track.
State-Based Disassembly Planning
Chao Lei (University of Melbourne), Krista A. Ehinger (University of Melbourne)
OptimizationComputational EfficiencyRobotic Intelligence
🎯 What it does: This paper proposes a state-based decomposition planning (SBDP) method that prioritizes translational motion using physical simulation, stores intermediate states to avoid redundant simulations, thereby achieving efficient automatic disassembly sequence and path generation.
Statistical Model-driven Similarity Hashing: Bridging Modalities for Efficient Unsupervised Retrieval
Mingjin Kuai (Central South University), Zhan Yang (Central South University)
RetrievalConvolutional Neural NetworkAuto EncoderImageTextMultimodality
🎯 What it does: An unsupervised cross-modal hashing retrieval method SMSH is proposed, which constructs a unified similarity matrix and enhances similarity using statistical models to generate more discriminative binary hash codes.
STD-PLM: Understanding Both Spatial and Temporal Properties of Spatial-Temporal Data with PLM
Yiheng Huang (Beijing Jiaotong University), Huaiyu Wan (Beijing Jiaotong University)
Data SynthesisTransformerLarge Language ModelSupervised Fine-TuningTime Series
🎯 What it does: This paper proposes STD-PLM, a unified pre-trained language model framework for simultaneously performing spatial-temporal prediction and imputation tasks.
STEM-LTS: Integrating Semantic-Temporal Dynamics in LLM-driven Time Series Analysis
Zhe Zhao (University of Science and Technology of China), Yang Wang (University of Science and Technology of China)
TransformerLarge Language ModelPrompt EngineeringTime SeriesFinance Related
🎯 What it does: The STEM-LTS framework is proposed, which combines temporal decomposition, LLM semantic-temporal alignment, and multi-task learning to achieve high-precision forecasting of high-dimensional multi-scale time series.
Step-Calibrated Diffusion for Biomedical Optical Image Restoration
Yiwei Lyu (University of Michigan), Todd C. Hollon (University of Michigan)
RestorationDiffusion modelImageBiomedical Data
🎯 What it does: A method for unpaired optical image restoration called RSCD is proposed, utilizing the reverse diffusion process of diffusion models and dynamic step calibration to achieve denoising and reconstruction of Raman-based optical images.
STGC-NeRF: Spatial-Temporal Geometric Consistency for LiDAR Neural Radiance Fields in Dynamic Scenes
Shangshu Yu (Nanyang Technological University), Cheng Wang
GenerationDepth EstimationAutonomous DrivingNeural Radiance FieldPoint Cloud
🎯 What it does: Proposes STGC-NeRF, which enhances the NeRF reconstruction of dynamic scenes from LiDAR through spatial-temporal geometric consistency.
Stitch, Contrast, and Segment: Learning a Human Action Segmentation Model Using Trimmed Skeleton Videos
Haitao Tian (University of Ottawa), Pierre Payeur (University of Ottawa)
SegmentationDomain AdaptationRepresentation LearningGraph Neural NetworkContrastive LearningVideo
🎯 What it does: This paper proposes a transfer learning framework for unsupervised long-sequence action segmentation using cropped skeleton videos through three steps: stitching, contrastive learning, and segmentation.
STLC-KG:A Social Text Steganalysis Method Combining Large-Scale Language Models and Common-Sense Knowledge Graphs
Zhuang Wang (Beijing University of Posts and Telecommunications), Zhongliang Yang (Guangzhou University)
ClassificationAnomaly DetectionGraph Neural NetworkTransformerLarge Language ModelContrastive LearningText
🎯 What it does: A social text steganography detection framework STLC-KG is proposed and implemented, which combines large language models with common sense knowledge graphs. After expanding the text semantics using the knowledge graph, semantic features are extracted using ChatGLM2-6B, and knowledge graph features are encoded through GAT, ultimately completing the detection of steganographic text.
Stochastic Online Instrumental Variable Regression: Regrets for Endogeneity and Bandit Feedback
Riccardo Della Vecchia (University of Lille), Debabrota Basu (University of Lille)
Reinforcement LearningTabular
🎯 What it does: This paper studies random online instrumental variable regression, particularly addressing endogeneity and feedback in linear regression problems, and proposes the Online Two-Stage Least Squares (O2SLS) method and the OFUL IV algorithm based on O2SLS.
Stop Diverse OOD Attacks: Knowledge Ensemble for Reliable Defense
Zhenbo Shi (University of Science and Technology of China), Liusheng Huang (University of Science and Technology of China)
Domain AdaptationAnomaly DetectionAdversarial AttackTransformerReinforcement LearningImage
🎯 What it does: A Reliable Defense Ensemble (REE) framework is proposed to resist OOD attacks.
Storynizor: Consistent Story Generation via Inter-Frame Synchronized and Shuffled ID Injection
Yuhang Ma (Fuxi AI Lab, Netease Inc.), Zhipeng Hu
GenerationData SynthesisDiffusion modelImage
🎯 What it does: The Storynizor model is proposed to achieve coherent story image generation with multiple characters.
StoryWeaver: A Unified World Model for Knowledge-Enhanced Story Character Customization
Jinlu Zhang (Xiamen University), Xiaoshuai Sun (Xiamen University)
GenerationGraph Neural NetworkTransformerVision Language ModelDiffusion modelImageGraphBenchmark
🎯 What it does: A unified story visualization framework called StoryWeaver is proposed, which utilizes a Character-Graph to encode characters, attributes, and relationships in the story world, generating knowledge-enhanced scene descriptions, and refines cross-attention through Knowledge Enhanced Spatial Guidance (KE-SG) to avoid identity entanglement of multiple characters.
STraj: Self-training for Bridging the Cross-Geography Gap in Trajectory Prediction
Zhanwei Zhang (Zhejiang University), Wenxiao Wang (Zhejiang University)
Domain AdaptationAutonomous DrivingGraph Neural NetworkContrastive LearningTime Series
🎯 What it does: A self-training framework STraj is proposed for cross-geographical domain trajectory prediction.
Strategic Network Creation for Enabling Greedy Routing
Julian Berger (Hasso Plattner Institute), Janosch Ruff (Hasso Plattner Institute)
🎯 What it does: This paper proposes and analyzes a network creation game based on greedy routing, studying how self-interested intelligent agents can balance routing feasibility and connectivity quality when constructing networks in various metric spaces.
Strategyproof Matching of Roommates and Rooms
Hadi Hosseini (Pennsylvania State University), Sanjukta Roy (Indian Statistical Institute)
Optimization
🎯 What it does: This study investigates the matching problem between tenants and rooms, considering the complementary preferences of agents for rooms and tenants (Leontief), and explores how to maximize social welfare while maintaining strategy invariance.
StressPrompt: Does Stress Impact Large Language Models and Human Performance Similarly?
Guobin Shen (BrainCog Lab Institute of Automation Chinese Academy of Sciences), Yi Zeng (BrainCog Lab Institute of Automation Chinese Academy of Sciences)
TransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: This paper systematically evaluates the performance of large language models under different levels of 'stress' by constructing the StressPrompt dataset. It finds that model performance improves under moderate stress but declines under extremely high or low stress, showing a trend similar to the Yerkes-Dodson law in humans.
StructSR: Refuse Spurious Details in Real-World Image Super-Resolution
Yachao Li (Nanjing University of Aeronautics and Astronautics), Sheng-Jun Huang (Microsoft)
RestorationSuper ResolutionDiffusion modelImage
🎯 What it does: This paper proposes a plug-in method called StructSR that does not require fine-tuning, external priors, or semantic knowledge to improve the structural fidelity of diffusion models in real image super-resolution and suppress pseudo-details.
Structural Entropy Guided Probabilistic Coding
Xiang Huang (Beihang University), Philip S. Yu (University of Illinois Chicago)
ClassificationTransformerSupervised Fine-TuningText
🎯 What it does: This paper proposes a structure entropy-guided probabilistic encoding model (SEPC), which combines structure entropy regularization with a probabilistic encoding tree for classification and regression tasks in natural language understanding.
Structural Entropy Guided Unsupervised Graph Out-Of-Distribution Detection
Yue Hou (Beihang University), Ke Xu (Beihang University)
Anomaly DetectionGraph Neural NetworkContrastive LearningGraph
🎯 What it does: We propose SEGO, an unsupervised graph structure OOD detection framework that uses a coding tree obtained through structural entropy minimization as an anchor view, and distinguishes between ID and OOD graphs through multi-granularity (node, graph, tree) contrastive learning.
Structural Pruning via Spatial-aware Information Redundancy for Semantic Segmentation
Dongyue Wu (Huazhong University of Science and Technology), Changxin Gao (Huazhong University of Science and Technology)
ClassificationObject DetectionSegmentationConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a filter pruning method based on spatially aware information redundancy (SIRFP), specifically addressing the pruning needs of semantic segmentation models.
Structure Balance and Gradient Matching-Based Signed Graph Condensation
Rong Li (Shenzhen University), Lijia Ma (Shenzhen Technology University)
CompressionRecommendation SystemGraph Neural NetworkGraph
🎯 What it does: A method is proposed to compress signed graphs into small yet information-rich graphs for link prediction.
Structure-Adaptive Multi-View Graph Clustering for Remote Sensing Data
Renxiang Guan (National University of Defense Technology), Xinwang Liu (China University of Geosciences)
Graph Neural NetworkAuto EncoderContrastive LearningGraph
🎯 What it does: A multi-view graph clustering method SAMVGC based on superpixels, graph autoencoders, and adaptive graph structures is proposed for remote sensing data clustering.
Structured IB: Improving Information Bottleneck with Structured Feature Learning
Hanzhe Yang (ShanghaiTech University), Yuanming Shi (ShanghaiTech University)
ClassificationRepresentation LearningAuto EncoderImage
🎯 What it does: Designed and validated the Structured Information Bottleneck (SIB) framework, which enhances the representation capability of traditional IB methods using auxiliary encoders.
Structured Packing in LLM Training Improves Long Context Utilization
Konrad Staniszewski (University of Warsaw), Piotr Miłoś (University of Warsaw)
RetrievalOptimizationTransformerLarge Language ModelSupervised Fine-TuningTextMultimodalityRetrieval-Augmented Generation
🎯 What it does: This paper proposes and implements SPLICE, a structured packaging method that stitches semantically related documents into long training samples through retrieval, aimed at enhancing the long context utilization capability of large language models.
Style Nursing with Spatial and Semantic Guidance for Zero-Shot Traffic Scene Style Transfer
Zhen Wang (Xi'an Jiaotong University), Chi Zhang (Xi'an Jiaotong University)
Image TranslationGenerationDiffusion modelImageBenchmark
🎯 What it does: This paper proposes an unsupervised zero-shot transfer method that utilizes Style Nursing along with spatial and semantic guidance to achieve diverse style transfer of traffic scene images while preserving the original content.
StyO: Stylize Your Face in Only One-Shot
Bonan Li (University of Chinese Academy of Sciences), Tiande Guo (University of Chinese Academy of Sciences)
Image TranslationGenerationDiffusion modelImage
🎯 What it does: A first-order sample facial stylization model, StyO, is proposed, which can transfer the geometry and texture style of a single target image to the source image while preserving the content of the source image.
Sub-Interest-Aware Representation Uniformity for Recommender System
Ruijia Ma (Nankai University), Chunyao Song (Nankai University)
Recommendation SystemGraph Neural NetworkTabular
🎯 What it does: The SIURec model is proposed, utilizing sub-interest-aware uniformization for hierarchical representation learning in recommendation systems.
Subgraph Aggregation for Out-of-Distribution Generalization on Graphs
Bowen Liu (Harbin Institute of Technology), Shanghang Zhang (Peking University)
Graph Neural NetworkContrastive LearningGraph
🎯 What it does: The SuGAr framework is proposed, which can learn and aggregate multiple causal subgraphs in graph neural networks to enhance prediction performance under out-of-distribution (OOD) conditions.
Subgraph Invariant Learning Towards Large-Scale Graph Node Classification
Leilei Wang (Shenzhen University), Ying Tiffany He (Hong Kong University of Science and Technology Guangzhou)
ClassificationExplainability and InterpretabilityComputational EfficiencyGraph Neural NetworkTransformerContrastive LearningGraph
🎯 What it does: Proposes a Subgraph Invariant Learning framework that utilizes subgraph training with an Invariance Encoder (IRE) and Node Encoder (NRE) to achieve large-scale graph node classification, significantly reducing computational costs and enhancing out-of-distribution (OOD) generalization.
SubjectDrive: Scaling Generative Data in Autonomous Driving via Subject Control
Binyuan Huang (Wuhan University), Xiangyu Zhang (Mach Drive)
Object DetectionObject TrackingData SynthesisAutonomous DrivingDiffusion modelVideo
🎯 What it does: For the task of autonomous driving, a SubjectDrive framework is proposed, which generates scalable and diverse synthetic driving video data through a subject control mechanism;
Suboptimal Search with Dynamic Distribution of Suboptimality
Mohammadreza Hami (University of Alberta), Nathan R. Sturtevant (University of Alberta)
OptimizationReinforcement LearningTabularBenchmark
🎯 What it does: Proposed the DSWA* algorithm, which dynamically allocates search sub-optimality, avoids state reopening, and maintains global sub-optimality;
Sum of Squares Circuits
Lorenzo Loconte (University of Edinburgh), Antonio Vergari (University of Artois)
ImageTabular
🎯 What it does: This paper studies the expressive power of differentiable probabilistic circuits (PCs), proposes a new model called 'sum of squares (SOS) PCs', and constructs an expression hierarchy, proving that it significantly enhances expressiveness compared to single square PCs and monotonic PCs.
SUMI-IFL: An Information-Theoretic Framework for Image Forgery Localization with Sufficiency and Minimality Constraints
Ziqi Sheng (Sun Yat-sen University), Xiaochun Cao (University of Macau)
SegmentationConvolutional Neural NetworkImage
🎯 What it does: An information theory-driven image forgery localization framework SUMI-IFL is proposed, which combines multi-view attention U-Net to extract features and achieves feature sufficiency and simplicity through mutual information and information bottleneck constraints.
SUMO: Search-Based Uncertainty Estimation for Model-Based Offline Reinforcement Learning
Zhongjian Qiao (Tsinghua University), Xiu Li (Tsinghua University)
Reinforcement LearningTabular
🎯 What it does: A search-based uncertainty estimation method called SUMO is proposed, which is integrated into model-based offline reinforcement learning (such as MOPO, AMOReL). It estimates the cross-entropy between model-generated dynamics and real dynamics through k-NN search, thereby enhancing policy performance.
Super-Class Guided Transformer for Zero-Shot Attribute Classification
Sehyung Kim (Korea University), Hyunwoo J. Kim (Korea University)
ClassificationRetrievalTransformerVision Language ModelImage
🎯 What it does: A framework for zero-shot attribute classification using Superclass-guided Transformer (SugaFormer) is proposed.
Supervised Score-Based Modeling by Gradient Boosting
Changyuan Zhao (Nanyang Technological University), Dusit Niyato (Nanyang Technological University)
ClassificationOptimizationComputational EfficiencyScore-based ModelTabular
🎯 What it does: A score-based supervised model SSM is proposed, which combines gradient boosting with denoising score matching for supervised learning tasks.
Support Vector-based Estimation of Multilinear Games for Feature Selection and Explanation
Majid Mohammadi (Vrije Universiteit Amsterdam), Annette Ten Teije (Vrije Universiteit Amsterdam)
Explainability and InterpretabilityTabular
🎯 What it does: A method is proposed to model supervised learning as a multilinear game, utilizing support vector machines (SVM) and dynamic programming to efficiently compute the Shapley values and interaction indices of features.
Supportive Negatives Spectral Augmentation for Source-Free Cross-Domain Segmentation
Kexin Zheng (Southeast University), Zhengming Ding (Tulane University)
SegmentationDomain AdaptationImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A source-free domain adaptation method based on actively discovering hard negative samples and performing spectral mixing enhancement (SNSA) is proposed for medical image segmentation.
Surgical Workflow Recognition and Blocking Effectiveness Detection in Laparoscopic Liver Resection with Pringle Maneuver
Diandian Guo (Chinese University of Hong Kong), Pheng-Ann Heng (Chinese University of Hong Kong)
RecognitionTransformerContrastive LearningVideo
🎯 What it does: This paper proposes a unified framework for online workflow recognition and effectiveness detection of the Pringle maneuver in laparoscopic liver resection, and constructs the corresponding PmLR50 dataset.
SUTrack: Towards Simple and Unified Single Object Tracking
Xin Chen (Dalian University of Technology), Huchuan Lu (Dalian University of Technology)
Object TrackingTransformerMultimodality
🎯 What it does: The SUTrack framework is proposed, which allows a single model to simultaneously handle five types of single-object tracking tasks, including RGB, RGB-Depth, RGB-Thermal, RGB-Event, and RGB-Language with just one training session.
SVasP: Self-Versatility Adversarial Style Perturbation for Cross-Domain Few-Shot Learning
Wenqian Li (Southeast University), Hui Xue (Southeast University)
Domain AdaptationConvolutional Neural NetworkTransformerGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes an adaptive diversity adversarial style perturbation method (SVasP), which stabilizes the global style adversarial updates by utilizing locally cropped style gradients in single-source cross-domain few-shot learning, thereby enhancing the model's domain transfer performance.
SVGBuilder: Component-Based Colored SVG Generation with Text-Guided Autoregressive Transformers
Zehao Chen (Sun Yat-sen University), Rong Pan (Sun Yat-sen University)
GenerationData SynthesisTransformerLarge Language ModelImage
🎯 What it does: A component-based autoregressive model called SVGBuilder has been developed, which can quickly generate colored SVG graphics based on text.
SVRMamba: Slice-to-Volume Reconstruction from Multiple MRI Stacks with Slice Sequence Guided Mamba
Jiangjie Wu (ShanghaiTech University), Yuyao Zhang (Shanghai Jiao Tong University)
RestorationConvolutional Neural NetworkImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: This paper proposes a method for fetal MRI multi-stack 2D slice to 3D reconstruction (SVR) based on the Mamba state space model and convolutional interpolation network.
SVTformer: Spatial-View-Temporal Transformer for Multi-View 3D Human Pose Estimation
Wanruo Zhang (Peking University), Wenhao Li (Nanyang Technological University)
Pose EstimationTransformerImage
🎯 What it does: This paper proposes SVTformer, which utilizes a spatial-view-time three-dimensional transformer for 3D pose recovery from multi-view 2D skeletons.
SWAMamba: A Sliding Window Attention Mamba Framework for Predicting Translation Elongation Rates
Xi Zeng (Northwestern Polytechnical University), Jiajie Peng (Northwestern Polytechnical University)
Supervised Fine-TuningBiomedical Data
🎯 What it does: Proposed the SWAMamba framework, which combines Mamba and sliding window attention to predict mRNA translation elongation rates.
SWEA: Updating Factual Knowledge in Large Language Models via Subject Word Embedding Altering
Xiaopeng Li (National University of Defense Technology), Weimin Zhang (National University of Defense Technology)
TransformerLarge Language ModelText
🎯 What it does: A method for editing factual knowledge in large language models by changing topic word embeddings is proposed—SWEA⊕OS.
Sweeping Heterogeneity with Smart MoPs: Mixture of Prompts for LLM Task Adaptation
Chen Dun (Rice University), Anastasios Kyrillidis (Rice University)
CompressionDomain AdaptationFederated LearningTransformerLarge Language ModelPrompt EngineeringMixture of ExpertsText
🎯 What it does: This paper proposes an efficient prompt tuning method for pre-trained large language models using intelligent gated mixed prompts (MoPs) in multi-source and multi-task scenarios.
SwiftTry: Fast and Consistent Video Virtual Try-On with Diffusion Models
Hung Nguyen (VinAI Research), Rang Nguyen (VinAI Research)
GenerationData SynthesisDiffusion modelVideo
🎯 What it does: A video virtual try-on framework called SwiftTry is proposed, based on a diffusion model, which achieves efficient and temporally consistent video try-on by incorporating temporal attention layers and ShiftCaching technology into UNet.
Symbolic Functional Decomposition: A Reconfiguration Approach
Mateus de Oliveira Oliveira (Stockholm University), Wim Van Den Broeck
Optimization
🎯 What it does: This study explores the issues of function decomposition and function reconfiguration, proposing a symbolic framework to address these problems, utilizing Ordered Binary Decision Diagrams (OBDDs) to represent functions and describing the reconfiguration process through Boolean circuits.
Symbolic Neural Ordinary Differential Equations
Xin Li (National University of Defense Technology), Xiaojun Duan (National University of Defense Technology)
Explainability and InterpretabilityComputational EfficiencyTime SeriesPhysics RelatedOrdinary Differential Equation
🎯 What it does: This paper proposes the Symbolic Neural Ordinary Differential Equations (SNODEs) framework, which utilizes symbolic continuous deep networks to learn the mapping from parameter functions to system states, achieving high-precision modeling of parameterized ODEs/PDEs through a three-stage training process (gradient flow matching pre-training, NODE fine-tuning, and residual networks).
SymmCompletion: High-Fidelity and High-Consistency Point Cloud Completion with Symmetry Guidance
Hongyu Yan (Hong Kong University of Science and Technology), Ping Tan (Hong Kong University of Science and Technology)
RestorationGenerationAutonomous DrivingOptimizationTransformerPoint Cloud
🎯 What it does: This paper proposes the SymmCompletion framework, which achieves high-fidelity and high-consistency point cloud completion through the Local Symmetric Transformation Network (LSTNet) and the Symmetric-guided Transformer (SGFormer).
Synchronization in Learning in Periodic Zero-Sum Games Triggers Divergence from Nash Equilibrium
Yuma Fujimoto (CyberAgent), Kenshi Abe (CyberAgent)
OptimizationReinforcement Learning from Human FeedbackReinforcement LearningTime SeriesSequential
🎯 What it does: This study investigates the behavior of learning dynamics in periodic zero-sum games as the game cycle changes, finding that when the learning rate is synchronized with the game cycle, it leads to strategy divergence and non-convergence of time averages; otherwise, the strategies are cyclic and the time averages converge.
SyncNoise: Geometrically Consistent Noise Prediction for Instruction-based 3D Editing
Ruihuang Li (Hong Kong Polytechnic University), Lei Zhang (Stability AI)
GenerationData SynthesisDepth EstimationDiffusion modelNeural Radiance FieldPoint CloudMesh
🎯 What it does: The SyncNoise method is proposed, achieving high-quality 3D scene editing under text instructions through geometrically guided multi-view synchronous noise prediction.
Synergistic Multi-Agent Framework with Trajectory Learning for Knowledge-Intensive Tasks
Shengbin Yue (Fudan University), Zhongyu Wei (Fudan University)
GenerationRetrievalExplainability and InterpretabilityTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: The SMART multi-agent framework is proposed, where four specialized agents collaborate to complete knowledge-intensive tasks through intent reconstruction, knowledge retrieval, fact localization, and response generation, significantly enhancing the factual consistency and interpretability of the generated results.
Synergy of GFlowNet and Protein Language Model Makes a Diverse Antibody Designer
Mingze Yin (Zhejiang University), Jian Wu (Zhejiang University)
GenerationDrug DiscoveryGraph Neural NetworkFlow-based ModelBiomedical Data
🎯 What it does: A framework for antibody design called PG-AbD is proposed, which combines generative flow networks with pre-trained protein language models to generate highly developable, diverse, and novel antibody CDRs.
Synthetic Tabular Data Generation for Imbalanced Classification: The Surprising Effectiveness of an Overlap Class
Annie D'souza (Indian Institute of Technology Bombay), Sunita Sarawagi (Indian Institute of Technology Bombay)
ClassificationData SynthesisDiffusion modelTabular
🎯 What it does: This study focuses on synthetic data augmentation in extremely imbalanced tabular data, proposing the Overlap Region Detection (ORD) method. It changes the binary classification labels to three classes by introducing an overlapping class, using a conditional generator to produce higher quality minority class samples, and training the classifier after removing overlapping majority class samples.
T-MDML: Triplet-based Multiple Distance Metric Learning for Multi-Instance Multi-Label Classification with Label Correlation
Dongyeon Kim (Dongguk University), Gangman Yi (Dongguk University)
ClassificationTabular
🎯 What it does: This paper proposes a method called T-MDML, a triplet multi-distance metric learning approach for multi-instance multi-label (MIML) classification.
Tab-Shapley: Identifying Top-k Tabular Data Quality Insights
Manisha Padala (Indian Institute of Technology Gandhinagar), Shiv Kumar Saini (Adobe Research)
Anomaly DetectionOptimizationAuto EncoderTabular
🎯 What it does: Proposes an unsupervised Tab-Shapley method that uses Shapley values to aggregate anomalies in tabular data, generating focused Top-k data quality insight blocks;
TabGLM: Tabular Graph Language Model for Learning Transferable Representations Through Multi-Modal Consistency Minimization
Anay Majee (University of Texas at Dallas), Wei-Peng Chen (Fujitsu Research of America)
Representation LearningGraph Neural NetworkSupervised Fine-TuningContrastive LearningMultimodalityTabular
🎯 What it does: This paper proposes a method that simultaneously converts each record into a fully connected graph and serialized text, and then performs joint table representation learning using graph neural networks and a pre-trained text encoder.
TableBench: A Comprehensive and Complex Benchmark for Table Question Answering
Xianjie Wu (Beihang University), Guanglin Niu (Beihang University)
TransformerLarge Language ModelSupervised Fine-TuningTabularBenchmarkChain-of-Thought
🎯 What it does: A new complex table question-answering benchmark named TableBench has been constructed, and the TableInstruct instruction corpus has been created to train the TABLELLM model, which is used to evaluate the table reasoning capabilities of LLMs in industrial scenarios.
Tackling Intertwined Data and Device Heterogeneities in Federated Learning with Unlimited Staleness
Haoming Wang (University of Pittsburgh), Wei Gao (University of Pittsburgh)
Federated LearningImage
🎯 What it does: A gradient inversion-based federated learning framework is proposed to convert delayed model updates into non-delayed updates, thereby addressing the issue of intertwined data and device heterogeneity.
TAIL-MIL: Time-Aware and Instance-Learnable Multiple Instance Learning for Multivariate Time Series Anomaly Detection
Jaeseok Jang (Seoul National University of Science and Technology), Hyuk-Yoon Kwon (Seoul National University of Science and Technology)
Anomaly DetectionTransformerTime Series
🎯 What it does: The TAIL-MIL model is proposed for anomaly detection at the time point (VI-TPIM), time period (TPIM), and overall (TAB) levels of multivariate time series with only bag-level labels.
TAMER: Tree-Aware Transformer for Handwritten Mathematical Expression Recognition
Jianhua Zhu (Peking University), Liangcai Gao (Peking University)
RecognitionTransformerImage
🎯 What it does: The TAMER model is proposed, which learns sequences and tree structures in parallel through a Tree-Aware Module and Transformer to recognize handwritten mathematical expressions.
Target Scanpath-Guided 360-Degree Image Enhancement
Yujia Wang (Victoria University of Wellington), Neil A. Dodgson (Victoria University of Wellington)
GenerationTransformerDiffusion modelImage
🎯 What it does: This paper proposes a target scanpath-guided 360° image enhancement task and designs a three-stage Progressive Scanpath-Guided Enhancement Model (PSEM). It first generates representative source scanpaths using TASSC, then predicts differences and generates editing masks using DSDE and SAM, and finally fine-tunes Stable Diffusion with LoRA to generate enhanced images consistent with the target scanpath.
Target Semantics Clustering via Text Representations for Robust Universal Domain Adaptation
Weinan He (University of Science and Technology of China), Yixin Zhang (University of Science and Technology of China)
Domain AdaptationContrastive LearningText
🎯 What it does: This paper proposes a Text-based Target Semantic Clustering (TASC) framework for achieving semantic clustering and domain alignment in Unified Domain Adaptation (UniDA).
Target-Driven Distillation: Consistency Distillation with Target Timestep Selection and Decoupled Guidance
Cunzheng Wang (Zhejiang University), Yao Hu (Xiaohongshu)
GenerationKnowledge DistillationDiffusion modelImage
🎯 What it does: A target-driven distillation (TDD) method is proposed for consistency distillation to improve the quality of few-shot generation.
Task-Agnostic Language Model Watermarking via High Entropy Passthrough Layers
Vaden Masrani (Huawei Technologies Canada), Yong Zhang (Huawei Technologies Canada)
ClassificationTransformerLarge Language ModelTextSequential
🎯 What it does: This paper proposes a task-agnostic, black-box retrievable watermarking method by inserting self-supervised passthrough layers into pre-trained language models.
Task-level Distributionally Robust Optimization for Large Language Model-based Dense Retrieval
Guangyuan Ma (Institute of Information Engineering Chinese Academy of Sciences), Songlin Hu (Langboat Technology)
RetrievalOptimizationTransformerLarge Language ModelContrastive LearningText
🎯 What it does: In multi-task multi-source training data, an end-to-end data distribution optimization method called tDRO is proposed for large language model dense retrieval (LLM-DR) to enhance general domain generalization capabilities.
Task-Specific Preconditioner for Cross-Domain Few-Shot Learning
Suhyun Kang (Samsung Research), Wonjong Rhee (Seoul National University)
Domain AdaptationOptimizationMeta LearningConvolutional Neural NetworkImage
🎯 What it does: An adaptive optimization mechanism based on Task-Specific Preconditioned Gradient Descent (TSP) is proposed to enhance the performance of Cross-Domain Few-Shot Learning (CDFSL).
Taylor Series-Inspired Local Structure Fitting Network for Few-shot Point Cloud Semantic Segmentation
Changshuo Wang (Nanyang Technological University), Prayag Tiwari (Halmstad University)
SegmentationConvolutional Neural NetworkPoint Cloud
🎯 What it does: A pre-training-free TaylorSeg network is proposed, which uses Taylor series-inspired TaylorConv to extract local structures of point clouds, and aligns query and prototype features through the APP module to achieve few-shot point cloud semantic segmentation.
TB-HSU: Hierarchical 3D Scene Understanding with Contextual Affordances
Wenting Xu (University of Sydney), Craig T. Jin (University of Sydney)
ClassificationSegmentationTransformerPoint Cloud
🎯 What it does: Construct a 3D Hierarchical Scene Graph (3DHSG) and utilize the Transformer multi-task learning framework TB-HSU to automatically learn room classification and area segmentation, while introducing context-variable functional (affordance) annotations.
TC-Diffuser: Bi-Condition Multi-Modal Diffusion for Tropical Cyclone Forecasting
Shiqi Zhang (Zhejiang University of Technology), Cong Bai (Shandong University)
GenerationData SynthesisRecurrent Neural NetworkTransformerDiffusion modelMultimodalityTime Series
🎯 What it does: A dual-condition multimodal diffusion model (TC-Diffuser) is proposed for predicting the trajectory, pressure, and wind speed of tropical cyclones.
TC-LLaVA: Rethinking the Transfer of LLava from Image to Video Understanding with Temporal Considerations
Mingze Gao (Hong Kong University of Science and Technology), Hui Xiong (Hong Kong University of Science and Technology)
RecognitionGenerationTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelVideoMultimodality
🎯 What it does: This paper proposes TC-LLaVA, which utilizes Temporal-Aware Dual RoPE and Frame-wise Block Causal Attention Mask to enhance the temporal awareness and visual token interaction of video language models, achieving end-to-end fine-tuning for video tasks.
TCAM-Diff: Triplane-Aware Cross-Attention Medical Diffusion Model
Zhenkai Zhang (University of Melbourne), Tom Drummond (University of Melbourne)
RestorationGenerationCompressionDiffusion modelAuto EncoderGenerative Adversarial NetworkMultimodalityBiomedical DataMagnetic Resonance Imaging
🎯 What it does: The TCAM-Diff model is proposed, which uses a decoder-only autoencoder to compress 3D medical images into a three-plane representation, and then generates high-resolution 3D volumes through a cross-plane attention diffusion model.
TCAQ-DM: Timestep-Channel Adaptive Quantization for Diffusion Models
Haocheng Huang (Beihang University), Yunhong Wang (Beihang University)
GenerationCompressionConvolutional Neural NetworkDiffusion modelImage
🎯 What it does: This paper proposes TCAQ-DM, a post-training quantization method for diffusion models;
TCPFormer: Learning Temporal Correlation with Implicit Pose Proxy for 3D Human Pose Estimation
Jiajie Liu (Peking University), Wenhao Li (Nanyang Technological University)
Pose EstimationTransformerSequential
🎯 What it does: Proposes TCPFormer, which learns the temporal correlation of 3D human poses through an implicit pose proxy.
TdAttenMix: Top-Down Attention Guided Mixup
Zhiming Wang (Beihang University), Feng Lu (Beihang University)
ClassificationSegmentationTransformerContrastive LearningImage
🎯 What it does: A task-oriented Top-Down Attention Guided Mixup (TdAttenMix) data augmentation method is proposed, which utilizes human eye gaze to guide attention balance between top-down and bottom-up attention for selecting mixing regions and dynamically adjusting label ratios.
Teacher-guided Edge Discriminator for Personalized Graph Masked Autoencoder
Qiqi Zhang (Shandong University of Science and Technology), Zhongying Zhao (Shandong University of Science and Technology)
ClassificationRepresentation LearningGraph Neural NetworkAuto EncoderContrastive LearningGraph
🎯 What it does: This paper proposes a teacher-guided edge discriminator and a personalized graph mask autoencoder (TEDMAE) for self-supervised learning of node representations in graphs with a mix of homogeneous and heterogeneous edges.
Teaching Models to Improve on Tape
Liat Bezalel (Tel Aviv University), Amir Globerson (Tel Aviv University)
GenerationMeta LearningReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningText
🎯 What it does: This paper designs and implements CORGI, a framework for LLM trained through interactive critical feedback and reinforcement learning (PPO) to improve the accuracy of constrained text generation.
TechSinger: Technique Controllable Multilingual Singing Voice Synthesis via Flow Matching
Wenxiang Guo (Zhejiang University), Zhou Zhao (Zhejiang University)
GenerationData SynthesisTransformerPrompt EngineeringFlow-based ModelAudio
🎯 What it does: A controllable singing synthesis system called TechSinger is proposed, which is based on flow matching and allows for fine control of singing techniques through natural language prompts.
Template-Driven LLM-Paraphrased Framework for Tabular Math Word Problem Generation
Xiaoqiang Kang (Xi'an Jiaotong-Liverpool University), Qiufeng Wang (Xi'an Jiaotong-Liverpool University)
GenerationData SynthesisExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningTabular
🎯 What it does: A framework based on template-driven and LLM rephrasing (TeLL) is proposed for the automatic generation of table mathematics problems with high correctness and diversity, and based on this, the TabMWP-TeLL dataset is constructed.
Temporal Action Localization with Cross Layer Task Decoupling and Refinement
Qiang Li (Northeast Normal University), Jianzhong Wang (Northeast Normal University)
ClassificationRecognitionObject DetectionConvolutional Neural NetworkVideoBenchmark
🎯 What it does: A temporal action localization framework based on Cross-Layer Task Decoupling and Refinement (CLTDR) and Gated Multi-Granularity Features (GMG) is proposed, which can decouple action classification and localization tasks at the feature level and further align the predictions of both through a refinement head.
Temporal Causal Reasoning with (Non-Recursive) Structural Equation Models
Maksim Gladyshev (Utrecht University), Brian Logan (University of Aberdeen)
🎯 What it does: A framework that combines Structural Equation Modeling (SEM) with temporal logic is proposed, defining the temporal causal logic CPLTL and providing its model checking algorithm.
Temporal Coherent Object Flow for Multi-Object Tracking
Zikai Song (Huazhong University of Science and Technology), Wei Yang (Huazhong University of Science and Technology)
Object TrackingAutonomous DrivingOptical FlowVideo
🎯 What it does: A multi-object tracking framework OFTrack based on temporal consistent target flow is proposed, integrating the detector and flow head to achieve one-time multi-frame tracking.
Temporal Conjunctive Query Answering via Rewriting
Lukas Westhofen (German Aerospace Center), Daniel Neider (TU Dortmund University)
Time SeriesBenchmark
🎯 What it does: A rewriting rule based on temporal logic expansion and distribution laws is proposed to transform time series queries into a form that can be directly executed as ordinary CQ solving;
Temporal Fair Division
Benjamin Cookson (University of Toronto), Nisarg Shah (University of Toronto)
Optimization
🎯 What it does: This paper studies a model for fair distribution of goods among a group of agents over time, providing optimal fairness guarantees under various constraints.
Temporal Numeric Planning with Patterns
Matteo Cardellini (University of Genoa), Enrico Giunchiglia (University of Genoa)
Optimization
🎯 What it does: This paper extends Symbolic Pattern Planning (SPP) to temporal numeric planning, providing a set of SMT-based encodings and proving their correctness and completeness.
Temporal Specification Optimisation for the Event Calculus
Periklis Mantenoglou (NCSR Demokritos), Alexander Artikis (University of Piraeus)
OptimizationTabularTime Series
🎯 What it does: A compiler has been constructed to convert simple fluents in RTEC into statically determined fluents, thereby optimizing event descriptions.
Temporal Task and Motion Planning with Metric Time for Multiple Object Navigation
Elisa Tosello (Fondazione Bruno Kessler), Andrea Micheli (Fondazione Bruno Kessler)
OptimizationRobotic IntelligenceSimultaneous Localization and MappingTabularBenchmark
🎯 What it does: This paper studies the time task and motion planning problem for multi-object simultaneous navigation (TTAMP) and proposes a novel interactive solver T-TAMPEST.
Temporal Triadic Closure: Finding Dense Substructures in Social Networks That Evolve over Time
Tom Davot (University of Glasgow), Kitty Meeks (University of Glasgow)
Graph Neural NetworkGraphTime Series
🎯 What it does: A new definition of temporal triadic closure graphs is proposed, studying the dense substructures that evolve over time in social networks.
Temporal-Aware Evaluation and Learning for Temporal Graph Neural Networks
Junwei Su (University of Hong Kong), Shan Wu (Hefei University of Technology)
Graph Neural NetworkGraphTime Series
🎯 What it does: This paper addresses the inadequacy of evaluation metrics for temporal graph neural networks by proposing a new evaluation method based on Volatility Cluster Statistics (VCS) and a corresponding learning regularization (VCA).
TEncDM: Understanding the Properties of the Diffusion Model in the Space of Language Model Encodings
Alexander Shabalin (HSE University), Dmitry Vetrov (Constructor University)
GenerationTransformerDiffusion modelText
🎯 What it does: Train diffusion models in the encoding space of pre-trained language models and propose the TEncDM framework, which combines Transformer decoders with techniques such as self-conditioning and noise scheduling.
Tensor Decomposition Meets Knowledge Compilation: A Study Comparing Tensor Trains with OBDDs
Ryoma Onaka (NTT Corporation), Norihito Yasuda (NTT Corporation)
🎯 What it does: Evaluates the succinctness and query/transformation handling of tensor decomposition (tensor training) as a representation of Boolean functions, and makes a theoretical comparison with OBDD.
Tensorized Attention for Understanding Multi-Object Relationships
Makoto Nakatsuji (NTT Human Informatics Laboratories), Yoshihide Sato (NTT Human Informatics Laboratories)
TransformerText
🎯 What it does: Proposes the Tensorized Attention Model (TAM), which introduces three-dimensional attention with Tucker decomposition into the Transformer, incorporating semantic objects and aligning them with queries to capture multi-object relationships.
Tensorized Label Learning Based Fast Fuzzy Clustering
Xingyu Xue (Xidian University), Qianqian Wang (Xidian University)
OptimizationMultimodality
🎯 What it does: A fast fuzzy clustering method based on tensor label learning (TLLFFC) is proposed, which avoids additional parameters by balancing regularization and achieves large-scale multi-view clustering using anchor graphs and label propagation.
Test-Time Adaptation on Noisy Data via Model-Pruning-Based Filtering and Flatness-Aware Entropy Minimization
Xingzhi Zhou (Hong Kong University of Science and Technology), Nevin L. Zhang (National University of Defense Technology)
Domain AdaptationImage
🎯 What it does: MoTTA is proposed, which filters noise samples through the output difference after pruning during testing and updates model parameters using entropy minimization with flatness constraints.
Testing Causal Models with Hidden Variables in Polynomial Delay via Conditional Independencies
Hyunchai Jeong (Columbia University), Elias Bareinboim (Mohamed bin Zayed University of Artificial Intelligence)
GraphTabular
🎯 What it does: This paper proposes the use of the C-component Local Markov Property (C-LMP) to test models in causal graphs with hidden variables, and presents an algorithm LISTCI that can list all non-empty conditional independence relations with polynomial delay.