AAAI 2026 Papers — Page 37
AAAI Conference on Artificial Intelligence · 4149 papers
SVD-NO: Learning PDE Solution Operators with SVD Integral Kernels
Noam Koren (Technion Israel Institute of Technology), Daniel Freedman (Tel Aviv University)
Computational EfficiencyPhysics Related
🎯 What it does: This paper proposes a new neural operator, SVD-NO, which directly parameterizes and learns the integral kernel using singular value decomposition (SVD), thereby efficiently approximating the solver of PDEs.
SVGL: Scale-Variable Graph Learning in Model Space for Multivariate Time Series Classification
Shikang Liu (University of Science and Technology of China), Huanhuan Chen (University of Science and Technology of China)
ClassificationGraph Neural NetworkTime Series
🎯 What it does: A multi-variable time series classification framework named SVGL based on multi-scale dynamic representation and scale-variable graph learning was developed.
SWIFT：A General Sensitive Weight Identification Framework for Fast Sensor-Transfer Pansharpening
Zeyu Xia (University of Electronic Science and Technology of China), Liang-Jian Deng (University of Electronic Science and Technology of China)
Domain AdaptationImage
🎯 What it does: Proposes the SWIFT framework to achieve low-cost, rapid adaptation for cross-sensor high-resolution hyperspectral fusion models.
SwiftVideo: A Unified Framework for Few-Step Video Generation Through Trajectory-Distribution Alignment
Yanxiao Sun (Fudan University), Yanwei Fu (Tencent Youtu Lab)
GenerationKnowledge DistillationTransformerDiffusion modelGenerative Adversarial NetworkVideoOrdinary Differential Equation
🎯 What it does: Propose the SwiftVideo framework, achieving few-step video generation through a unified approach combining Continuous Time Consistency Distillation (CCD) + Distribution Alignment (DA) + Trajectory Alignment (TA)
Syllogism-Inspired TableQA: Evidentialization Makes Decomposition Reasoning and Answer Verification More Reliable
Zhe Zhang (Northeastern University), Jie Song (Northeastern University)
TransformerLarge Language ModelPrompt EngineeringTabularBenchmarkChain-of-Thought
🎯 What it does: This paper proposes a syllogism-based TableQA method called SIRV, which can utilize factual evidence from tables to enhance the reliability of reasoning and answer verification in LLMs.
Symbolic Planning and Multi-Agent Path Finding in Extremely Dense Environments with Unassigned Agents
Bo Fu (Amazon Robotics), Federico Pecora (Amazon Robotics)
Optimization
🎯 What it does: This paper proposes and addresses the Block Rearrangement Problem (BRaP) in dense warehouse grids, aiming to move specified blocks to target positions;
SymGS: Leveraging Reflective Symmetries for 3DGS Compression
Keshav Gupta (IIIT Hyderabad), Avinash Sharma (IIT Jodhpur)
CompressionGaussian Splatting
🎯 What it does: Propose the SymGS framework, which leverages the reflection symmetry of 3D Gaussian splatting for compression, recursively detecting and optimizing mirror planes to achieve efficient geometric and attribute redundancy elimination.
Symmetric Aggregation of Conformity Scores for Efficient Uncertainty Sets
Nabil Alami (Mohamed Bin Zayed University of Artificial Intelligence), Souhaib Ben Taieb (Mohamed Bin Zayed University of Artificial Intelligence)
Computational EfficiencyHyperparameter SearchImageTabular
🎯 What it does: This paper proposes a synthetic prediction method called SACP based on symmetric aggregation of non-consistency scores, which can effectively aggregate prediction uncertainties from multiple models while maintaining precise coverage.
Symmetrical Flow Matching: Unified Image Generation, Segmentation, and Classification with Score-Based Generative Models
Francisco Caetano (Eindhoven University of Technology), Fons van der Sommen (Eindhoven University of Technology)
ClassificationSegmentationGenerationScore-based ModelFlow-based ModelImageOrdinary Differential Equation
🎯 What it does: Propose Symmetrical Flow Matching (SymmFlow), a unified flow matching framework that can accomplish three tasks—semantic segmentation, classification, and image generation—in the same model.
Symmetries and Other Variations of “End-Recursive” HTN Problems: Mapping the Border Between Decidable and Undecidable Restrictions
Hadyn Tang (Australian National University), Pascal Bercher (Australian National University)
🎯 What it does: Systematically transform regularity and tail recursion constraints in HTN planning, analyze complexity changes caused by different symmetries and relaxations, and prove that partial mild relaxation maintains PSPACE decidability, but combined relaxation leads to undecidability, while clarifying differences between early and modern definitions.
Symmetry Breaking for Inductive Logic Programming
Andrew Cropper (ELLIS Institute Finland), Matti Järvisalo
Computational EfficiencyBenchmark
🎯 What it does: Proposed and implemented a method in inductive logic programming (ILP) that breaks rule space symmetries by constraining safe variables.
Symmetry-Aware Transformer Training for Automated Planning
Markus Fritzsche (Linköping University), Jendrik Seipp (Linköping University)
OptimizationTransformerContrastive LearningBenchmark
🎯 What it does: Studied and proposed a symmetry-aware Transformer training method to improve plan generation and heuristic prediction in automated planning tasks.
Symphony-MoE: Harmonizing Disparate Pre-trained Models into a Coherent Mixture-of-Experts
Qi Wang (Chinese Academy of Sciences), Yue Yu (Peng Cheng Laboratory)
Computational EfficiencyKnowledge DistillationRepresentation LearningSupervised Fine-TuningMixture of ExpertsText
🎯 What it does: A hybrid expert model framework (Symphony-MoE) based on two-stage untrained functional alignment and post-finetuning was constructed, enabling the integration of experts from multi-source pre-trained models into a unified MoE system;
SyncBrain: Exploring Brain Functional Dynamics Through Neural Oscillatory Synchronization
Jiaqi Ding (University of North Carolina at Chapel Hill), Guorong Wu (University of North Carolina at Chapel Hill)
Explainability and InterpretabilityComputational EfficiencyRepresentation LearningGraphBiomedical DataMagnetic Resonance ImagingAlzheimer's Disease
🎯 What it does: Proposed SyncBrain, a physics-informed deep model based on Kuramoto coupled oscillators, to simulate brain function dynamics and decode brain states.
SynerDetect: Hierarchical Synergistic Learning for Generalizable AI-Generated Image Detection
Shuaibo Li (Hong Kong University Of Science And Technology Guangzhou), Lei Zhu (Sun Yat Sen University)
ClassificationAnomaly DetectionKnowledge DistillationPrompt EngineeringDiffusion modelContrastive LearningImage
🎯 What it does: Propose a hierarchical collaborative learning framework named SynerDetect to unify the capabilities of two fundamental models: perception (CLIP) and generation (diffusion models), achieving more robust AI-generated image detection.
Synergizing Multigrid Algorithms with Vision Transformer: A Novel Approach to Enhance the Seismic Foundation Model
Huiwen Wu (Zhejiang Laboratory), Hongbin Ye (Zhejiang Laboratory)
OptimizationRepresentation LearningTransformerAuto EncoderImagePhysics Related
🎯 What it does: Developed an Adaptive Two-Grid Transformer (ADATG) that efficiently pretrains on seismic waveform data through spectral decomposition, high-low frequency Hilbert encoding, and adaptive loss.
Synthetic Forgetting Without Access: A Few-Shot Zero-Glance Framework for Machine Unlearning
Qipeng Song (Royal Melbourne Institute Of Technology), Feng Xia (Royal Melbourne Institute Of Technology)
ClassificationSafty and PrivacySupervised Fine-TuningGenerative Adversarial NetworkImage
🎯 What it does: Designed and implemented the GFOES framework to achieve machine unlearning under the premise of having only a small amount of retained data and being unable to access forgotten data.
SynWeather: Weather Observation Data Synthesis Across Multiple Regions and Variables via a General Diffusion Transformer
Kaiyi Xu (University of Science and Technology of China), Lei Bai (Shanghai Artificial Intelligence Laboratory)
Data SynthesisTransformerPrompt EngineeringDiffusion modelAuto EncoderImageMultimodality
🎯 What it does: This paper constructs the first unified multi-region, multi-variable synthetic weather observation dataset, SynWeather, and proposes a text-prompt-based general diffusion transformer, SynWeatherDiff, for generating various weather variables;
T-APT: Text-Guided Modality-Aware Prompt Tuning for Arbitrary Multimodal Remote Sensing Data Joint Classification
Qinghao Gao (Xidian University), Wenqian Dong (Xidian University)
ClassificationTransformerPrompt EngineeringVision Language ModelMultimodality
🎯 What it does: Proposed a unified text-guided arbitrary modality remote sensing image joint classification framework called T-APT.
T-GVC: Trajectory-Guided Generative Video Coding at Ultra-Low Bitrates
Zhitao Wang, Debin Zhao (Harbin Institute Of Technology)
CompressionVision Language ModelDiffusion modelAuto EncoderVideo
🎯 What it does: Proposed a trajectory-guided generative video coding framework called T-GVC, achieving better video reconstruction performance at extremely low bitrates.
T-LoRA: Single Image Diffusion Model Customization Without Overfitting
Vera Soboleva (FusionBrain Lab), Konstantin Sobolev (FusionBrain Lab)
GenerationDiffusion modelImageText
🎯 What it does: Study custom diffusion models for single images, proposing the T-LoRA framework to alleviate overfitting.
T-Retriever: Tree-based Hierarchical Retrieval Augmented Generation for Textual Graphs
Chunyu Wei (Renmin University of China), Yueguo Chen (Renmin University of China)
RetrievalGraph Neural NetworkTransformerLarge Language ModelGraphRetrieval-Augmented Generation
🎯 What it does: Propose the T-Retriever framework, transforming attribute graph retrieval into tree structure retrieval and constructing an information-theoretically optimal encoding tree.
T-Rex-Omni: Integrating Negative Visual Prompt in Generic Object Detection
Jiazhou Zhou (Hong Kong University of Science and Technology), Lei Zhang (Hong Kong University of Science and Technology)
Object DetectionTransformerPrompt EngineeringContrastive LearningImage
🎯 What it does: Propose the T-Rex-Omni framework, integrating negative visual prompts into open-set object detection;
T-SKM-Net: Trainable Neural Network Framework for Linear Constraint Satisfaction via Sampling Kaczmarz-Motzkin Method
Haoyu Zhu (Zhejiang University), Qingchun Hou (Zhejiang University)
OptimizationTabularFinance Related
🎯 What it does: Embed the Sampling Kaczmarz-Motzkin (SKM) method into a neural network to construct a trainable linear constraint satisfaction network, T-SKM-Net, achieving input-dependent dynamic linear constraint hard constraint satisfaction.
T2Agent: A Tool-augmented Multimodal Misinformation Detection Agent with Monte Carlo Tree Search
Xing Cui, Huaibo Huang (Beijing University Of Posts And Telecommunications)
ClassificationComputational EfficiencyAgentic AIMultimodalityBenchmark
🎯 What it does: Propose a multi-source misinformation detection agent T2 Agent based on tool expansion, which identifies hybrid forgeries through dynamic programming and multi-modal verification.
T2I-RiskyPrompt: A Benchmark for Safety Evaluation, Attack, and Defense on Text-to-Image Model
Chenyu Zhang (Tianjin University), Anan Liu (Tianjin University)
Explainability and InterpretabilityAdversarial AttackLarge Language ModelPrompt EngineeringVision Language ModelContrastive LearningImageTextMultimodalityBenchmark
🎯 What it does: Constructed the T2I-RiskyPrompt benchmark for evaluating the safety of text-to-image models.
T3former: Temporal Graph Classification with Topological Machine Learning
Md Joshem Uddin (University of Texas at Dallas), Baris Coskunuzer (University of Texas at Dallas)
ClassificationGraph Neural NetworkTransformerGraph
🎯 What it does: Propose a temporal graph classification framework named T3FORMER, which extracts temporal information via a sliding window, combines topological (persistent homology) and spectral (Laplacian DOS) descriptors, and employs GraphSAGE to encode global structure; subsequently, it fuses multi-modal features through Transformer and Descriptor-Attention, ultimately achieving temporal graph classification.
T3Time: Tri-Modal Time Series Forecasting via Adaptive Multi-Head Alignment and Residual Fusion
Abdul Monaf Chowdhury (University of Dhaka), Safaeid Hossain Arib (University of Dhaka)
TransformerLarge Language ModelPrompt EngineeringTime Series
🎯 What it does: Proposed a tri-modal (temporal, spectral, and prompt) fusion framework called T3Time for multivariate time series prediction, integrating time-domain and frequency-domain encoding, LLM prompt embeddings, as well as adaptive multi-head cross-modal alignment and channel residual fusion;
T4NMTD: Transition-Centric Reinforcement Learning for Non-Markovian Task Decomposition
Ruixuan Miao (Xidian University), Zhenhua Duan (Xidian University)
Reinforcement LearningBenchmark
🎯 What it does: Proposes the T4NMTD framework, achieving task decomposition and parallel training through DFA conversion based on LTLf, addressing sparse rewards and long-term dependencies in non-Markov tasks.
Tab-PET: Graph-Based Positional Encodings for Tabular Transformers
Yunze Leng (National University of Singapore), Mehul Motani (National University of Singapore)
Representation LearningData-Centric LearningGraph Neural NetworkTransformerTabularBenchmark
🎯 What it does: Proposed the Tab-PET framework, which generates position encodings for Tabular Transformers using graph-based spectral methods, thereby enhancing the model's representation capability for tabular data.
TabFlash: Efficient Table Understanding with Progressive Question Conditioning and Token Focusing
Jongha Kim, Hyunwoo J. Kim (Google Cloud AI)
RecognitionTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelMultimodalityTabularBenchmark
🎯 What it does: Designed and implemented an efficient table image understanding model called TabFlash, combining progressive question conditioning, background token pruning, and token focusing training strategies.
TabGeoFlow: A Geometric Flow Matching Model for Tabular Data Synthesis
Jong In Choi (Korea Credit Information Services)
Data SynthesisSafty and PrivacyTransformerFlow-based ModelTabular
🎯 What it does: Proposed a geometric flow matching model called TabGeoFlow for generating high-quality tabular data while reducing memorization of training samples.
Tabular Learnwares Can Be Repurposed for Seemingly Irrelevant New Tasks
Peng Tan (Nanjing University), Zhi-Hua Zhou (Nanjing University)
Domain AdaptationRepresentation LearningTabularBenchmark
🎯 What it does: Under the learnware framework, this study investigates how to repurpose existing table learning models (without sharing raw data) for new tasks with significantly different feature and label spaces from the original task, and proposes a cross-task adaptation mechanism based on model prediction probabilities.
Tackling Dual-stage Missing Modalities in Brain Tumor Segmentation via Robust Modality Reconstruction and Prompt-guided Modality Adaptation
Yunpeng Zhao (National University Of Singapore), Yueming Jin (National University Of Singapore)
SegmentationKnowledge DistillationConvolutional Neural NetworkPrompt EngineeringMultimodalityBiomedical DataMagnetic Resonance Imaging
🎯 What it does: Proposed a generic model to simultaneously address missing modality issues during both training and inference phases for brain tumor segmentation, incorporating modules such as mask-reconstruction pre-training, distribution approximation, complete-then-distill, and prompt-guided modality adaptation.
Tackling Resource-Constrained and Data-Heterogeneity in Federated Learning with Double-Weight Sparse Pack
Qiantao Yang (Southeast University), Songze Li (East China Normal University)
Federated LearningComputational EfficiencyData-Centric LearningConvolutional Neural NetworkImageBenchmark
🎯 What it does: Designed and implemented a personalized federated learning method called FedCSPACK based on parameter package levels, which selects important parameter packages using cosine similarity and aggregates them through a dual-weight mask, significantly reducing communication overhead and enhancing model robustness.
TAdaRAG: Task Adaptive Retrieval-Augmented Generation via On-the-Fly Knowledge Graph Construction
Jie Zhang (Fuzhou University), Yanchao Tan (MemTensor (Shanghai) Technology Co., Ltd.)
GenerationRetrievalTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextGraphBenchmarkRetrieval-Augmented Generation
🎯 What it does: Propose the TAdaRAG framework, which utilizes intent-driven extraction templates and two-phase training (supervised fine-tuning + reinforcement learning) to instantaneously construct task-adapted knowledge graphs during inference, thereby improving the generation quality of RAG.
Talk2Code: A Multi-Turn Interaction Benchmark with Dual-Track Evaluation for Code Generation
Weibin Yang (Xidian University), Hao Wang (Xidian University)
AI Code AssistantLarge Language ModelTextBenchmark
🎯 What it does: This paper proposes the Talk2Code benchmark, which constructs a code generation task involving six types of users (two each for beginners, intermediates, and experts) with multi-round interactions, aiming to evaluate the ability of LLMs to generate code based on user proficiency levels in multi-round dialogues.
Talk2Image: A Multi-Agent System for Multi-Turn Image Generation and Editing
Shichao Ma (University of Science and Technology of China), Yang Wang (University of Science and Technology of China)
GenerationLarge Language ModelAgentic AIDiffusion modelImageTextMultimodality
🎯 What it does: Proposed the Talk2Image multi-agent system, supporting image generation and editing in multi-round dialogues.
Talon: Breaking the Synchronization Barrier in Speculative Decoding with Hybrid Model-based and Retrieve-based Drafting
Xiangxiang Gao (Bestpay AI Lab), Xiaolong Xu (Bestpay AI Lab)
OptimizationComputational EfficiencyTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Proposed the Talon architecture, leveraging asynchronous parallel inference and adaptive hybrid draft strategies to address synchronization bottlenecks and low acceptance rates during inference in large language models.
Taming Cascaded Mixture-of-Experts for Modality-missing Multi-modal Salient Object Detection
Kunpeng Wang (Anhui University), Keke Chen (Anhui University)
Object DetectionTransformerMixture of ExpertsMultimodality
🎯 What it does: Propose a Cascaded Mixture-of-Experts (CMoE) framework for simultaneously handling missing and complete input scenarios in multi-modal salient object detection, which includes Missing-aware MoE (MaMoE) and Multi-modal MoE (MmMoE).
Taming the Phantom: Token-Asymmetric Filtering for Hallucination Mitigation in Large Vision-Language Models
Shuyi Ouyang (Zhejiang University), Xinchao Wang (National University of Singapore)
Explainability and InterpretabilityComputational EfficiencyTransformerVision Language ModelMultimodality
🎯 What it does: To address hallucinations in large vision-language models, this paper proposes a training-agnostic and plug-and-play method called Token-Asymmetric Filtering (TAF), which reduces hallucinations by suppressing the influence of 'phantom' text tokens and amplifying the influence of 'anchor' visual tokens in the visual active layer.
Tapas Are Free! Training-Free Adaptation of Programmatic Agents via LLM-Guided Program Synthesis in Dynamic Environments
Jinwei Hu (University of Liverpool), Xiaowei Huang (University of Liverpool)
Autonomous DrivingAI Code AssistantTransformerLarge Language ModelReinforcement LearningTabularTime SeriesRetrieval-Augmented Generation
🎯 What it does: Proposed the TAPA framework, which utilizes LLM as an action space regulator, achieving training-agnostic procedural agent adaptation in dynamic environments through online symbolic program synthesis and verification.
TARA: Token-Aware LoRA for Composable Personalization in Diffusion Models
Yuqi Peng (Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences), Shifeng Chen (Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences)
GenerationDiffusion modelContrastive LearningImageText
🎯 What it does: Proposed the Token-Aware LoRA (TARA) framework, which enhances LoRA's performance in multi-concept personalized generation through Token Focus Masking and Token Alignment Loss, achieving high-fidelity image synthesis without additional training or conditions.
TaREx: Reinforcement Learning for Code-Driven Table Reasoning
Fangyu Lei (Key Laboratory of Cognition and Decision Intelligence for Complex Systems, Institute of Automation, Chinese Academy of Sciences), Kang Liu (Key Laboratory of Cognition and Decision Intelligence for Complex Systems, Institute of Automation, Chinese Academy of Sciences)
AI Code AssistantTransformerLarge Language ModelReinforcement LearningTabular
🎯 What it does: Proposed the TAREX framework, combining a unified executable table representation, code-driven reinforcement learning, and multi-step interactive reasoning to address structural diversity, reasoning distortion, and multi-step challenges in table reasoning.
Target Refocusing via Attention Redistribution for Open-Vocabulary Semantic Segmentation: An Explainability Perspective
Jiahao Li (Xiamen University), Yanyun Qu (Hanjiang National Laboratory)
SegmentationExplainability and InterpretabilityTransformerVision Language ModelImage
🎯 What it does: Conduct interpretability analysis on CLIP, identify and eliminate attention dispersion issues, and enhance open-vocabulary semantic segmentation performance through untrained attention and embedding reallocation.
Target-Balanced Score Distillation
Zhou Xu (Tsinghua University), Yang Li (Tsinghua University)
GenerationOptimizationPrompt EngineeringDiffusion modelScore-based ModelNeural Radiance FieldImageText
🎯 What it does: Propose Target-Balanced Score Distillation (TBSD), dynamically balancing shape and texture optimization within Score Distillation Sampling (SDS), addressing the trade-off between shape distortion and suboptimal texture caused by negative prompts.
Targeted Data Protection for Diffusion Model by Matching Training Trajectory
Hojun Lee (Xperty Corp.), Nojun Kwak (Seoul National University)
GenerationSafty and PrivacySupervised Fine-TuningDiffusion modelImage
🎯 What it does: Proposes a target data protection method called TAFAP based on training trajectory matching, which can align the learning trajectory of protected data with the target concept during the fine-tuning process of text-to-image diffusion models, achieving precise transformation of identity and visual patterns;
Targeted Pathway Inference for Biological Knowledge Bases via Graph Learning and Explanation
Rikuto Kotoge (University of Osaka), Yasushi Sakurai (University of Osaka)
Explainability and InterpretabilityRepresentation LearningDrug DiscoveryGraph Neural NetworkLarge Language ModelGraphBiomedical Data
🎯 What it does: This paper proposes the EXPATH framework, which leverages graph learning and interpretable mechanisms to directly infer targeted path subgraphs from complete biological networks using experimental data;
Targeting Borderline Fraudsters: Multi-View Hypergraph Fraud Detection with LLM-Guided Contrastive Learning
Rui Ou (Tongji University), Changjun Jiang (Donghua University)
Anomaly DetectionGraph Neural NetworkLarge Language ModelContrastive LearningGraphTabularFinance Related
🎯 What it does: This paper proposes a multi-perspective hypergraph fraud detection model called MH-LGC, focusing on identifying boundary fraudsters by combining hypergraph attention networks with LLM-guided contrastive learning.
Targeting in Multi-Criteria Decision Making
Nicolas Schwind (National Institute of Advanced Industrial Science and Technology), Emmanuel Lonca (Université d'Artois)
Optimization
🎯 What it does: Propose and formalize the 'Targeting' multi-criteria decision problem, which involves finding the optimal alternative relative to given target options, and define its fundamental properties through axiomatization methods;
TargetVAU: Multimodal Anomaly-Aware Reasoning for Target Behavior Understanding in Videos
Lingru Zhou (Northwestern Polytechnical University), Peng Wang (Northwestern Polytechnical University)
Anomaly DetectionExplainability and InterpretabilityGraph Neural NetworkTransformerSupervised Fine-TuningVision Language ModelMultimodality
🎯 What it does: Propose the TargetVAU framework to achieve individual-level abnormal behavior recognition and explanation in videos.
TarPro: Targeted Protection Against Malicious Image Editing
Kaixin Shen, Jun Xiao (Zhejiang University)
GenerationSafty and PrivacyPrompt EngineeringDiffusion modelImageText
🎯 What it does: Proposed a targeted protection framework called TarPro for diffusion model image editing, which can shield against malicious NSFW edits without affecting normal editing effects.
TASE: Token Awareness and Structured Evaluation for Multilingual Language Models
Chenzhuo Zhao (Peking University), Ziqian Liu (Peking University)
Data SynthesisTransformerLarge Language ModelSupervised Fine-TuningTextBenchmarkChain-of-Thought
🎯 What it does: Proposed and implemented the TASE benchmark to evaluate the fine-grained token-level perception and structural reasoning capabilities of large language models across three languages (English, Chinese, and Korean).
Task Prototype-Based Knowledge Retrieval for Multi-Task Learning from Partially Annotated Data
Youngmin Oh (Kyung Hee University), Jung Uk Kim (Chonnam National University)
Object DetectionSegmentationTransformerImage
🎯 What it does: This paper proposes a prototype-based knowledge retrieval framework for achieving robust multi-task learning under partially labeled data.
Task-Aware 3D Affordance Segmentation via 2D Guidance and Geometric Refinement
Lian He (Beijing Institute of Technology), Gangyi Ding (Harbin Institute of Technology)
SegmentationTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextPoint Cloud
🎯 What it does: A framework named TASA is proposed for scene-level 3D functional segmentation based on natural language instructions, achieving precise segmentation of high-resolution point clouds through task-aware 2D semantic guidance and 3D geometric refinement.
Task-Aware Meta-Learning on Heterogeneous Knowledge Graph for POI Recommendation
Jingyuan Wang, Yiming Guan (Beijing Normal University)
Recommendation SystemMeta LearningRecurrent Neural NetworkGraph Neural NetworkGraphSequential
🎯 What it does: Proposes the Task-aware Meta-learning on Heterogeneous Knowledge Graph (TMHKG) framework, treating POI recommendation as a meta-learning task for each user, and dynamically capturing user interest evolution by fusing geographic proximity and category transfer relationships through dual-view knowledge graphs and interaction graphs.
Task-Aware Retrieval Augmentation for Dynamic Recommendation
Zhen Tao (Great Bay University), Qingqiang Sun (University Of Electronic Science And Technology Of China)
Recommendation SystemGraph Neural NetworkTransformerSupervised Fine-TuningContrastive LearningGraphRetrieval-Augmented Generation
🎯 What it does: This paper proposes a task-aware retrieval-enhanced framework called TarDGR to improve the generalization performance of dynamic graph recommendation systems under temporal distribution drift.
Task-Specific Distance Correlation Matching for Few-Shot Action Recognition
Fei Long (Dalian University of Technology), Peihua Li (Dalian University of Technology)
RecognitionTransformerSupervised Fine-TuningVision Language ModelVideo
🎯 What it does: Propose a new framework TS-FSAR for few-shot action recognition, including efficient fine-tuning of CLIP with Ladder Side Network (LSN), task-specific matching with Task-Specific Distance Correlation Matching (TS-DCM), and a guidance mechanism with Guiding LSN with Adapted CLIP (GLAC).
TawPipe: Topology-Aware Weight Pipeline Parallelism for Accelerating Long-Context Large Models Training
Houming Wu (Zhejiang University), Ling Chen (Zhejiang University)
Computational EfficiencyTransformerLarge Language ModelText
🎯 What it does: To address memory and communication bottlenecks in training large models with long contexts, we propose TawPipe, a topology-based weight pipeline parallel framework.
TaxReasoning: Benchmarking Knowledge-Intensive Mathematical Reasoning with Evolving Tax Laws
Nan Hu (Southeast University), Jeff Z. Pan (Monash University)
TextGraphTabularBenchmarkFinance RelatedRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Constructed and evaluated the TaxReasoning benchmark for tax knowledge reasoning, testing the performance of 13 LLMs on dynamic tax law reasoning tasks
TCoT: Trajectory Chain-of-Thoughts for Robotic Manipulation with Failure Recovery in Vision-Language-Action Model
Xiang Li (Tsinghua University), Shengjin Wang (Tsinghua University)
Robotic IntelligenceSupervised Fine-TuningVision-Language-Action ModelImageVideoTextMultimodalityBenchmarkChain-of-Thought
🎯 What it does: Designed and implemented the TCoT framework, integrating vision-language-action models with global/local trajectory planning and fault detection and recovery mechanisms to enhance the robot's execution capability in long-sequence tasks.
TDSNNs: Competitive Topographic Deep Spiking Neural Networks for Visual Cortex Modeling
Deming Zhou (Hong Kong University of Science and Technology (Guangzhou)), Renjing Xu (Hong Kong University of Science and Technology (Guangzhou))
ClassificationRepresentation LearningRecurrent Neural NetworkSpiking Neural NetworkTransformerImage
🎯 What it does: Constructed a deep spiking neural network (TDSNN) capable of spontaneously forming topological organization across multi-layer visual pathways, and achieved spatial clustering of functionally similar neurons through a novel spatiotemporal constraint (STC) loss.
TDSS: Task Dynamic-Synergistic Skill Adaptation for Boosting Efficient and Scalable Multi-Task Learning in Dense Visual Prediction
Haiming Yao (Tsinghua University), Wei You (Huawei Technologies Co. Ltd)
SegmentationDepth EstimationTransformerSupervised Fine-TuningMixture of ExpertsImage
🎯 What it does: Proposed a parameter-efficient fine-tuning based multi-task dense visual prediction framework TDSS, mainly achieving efficient multi-task adaptation of pre-trained vision Transformers through two modules: Task-Dynamic Skill Adapters (TDSA) and Task-Synergistic Adaptation Interaction (TSAI);
Teaching Large Language Models to Maintain Contextual Faithfulness via Synthetic Tasks and Reinforcement Learning
Shuzheng Si (Tsinghua University), Maosong Sun (University Of Illinois Urbana Champaign)
GenerationData SynthesisReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningText
🎯 What it does: Proposes the CANOE framework, leveraging synthetic short-text QA data without manual annotations and rule-based reinforcement learning methods to enhance the contextual faithfulness of large language models in both short and long-text generation tasks
TechCoach: Towards Technical-Point-Aware Descriptive Action Coaching
Yuan-Ming Li (Sun Yat-sen University), Weishi Zheng
Reinforcement Learning from Human FeedbackTransformerLarge Language ModelVision Language ModelVideoText
🎯 What it does: Proposes the Descriptive Action Coaching task, constructs the EE4D-DescCoach dataset and the TechCoach framework, which can generate detailed 'well done/needs improvement' comments and overall scores based on action videos.
Tell as You Want: Customizing Image Narrative with Knowledge and Thoughts
Ziwei Yao (Chinese Academy of Sciences), Xilin Chen (Chinese Academy of Sciences)
GenerationRetrievalExplainability and InterpretabilityTransformerLarge Language ModelVision Language ModelImageTextMultimodalityRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Proposed a novel image narrative generation framework called T4, capable of generating knowledge-rich and interpretable image narratives based on user interests;
Template-Theorems Graph Construction to Enhance Mathematical Reasoning Capabilities of LLM
Yarong Lan (Zhejiang University), Huajun Chen (Zhejiang University)
Graph Neural NetworkTransformerLarge Language ModelTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Built a human memory-inspired template-theorem graph knowledge base and designed an offline problem distillation and retrieval-augmented generation process to enhance the performance of large language models in mathematical reasoning.
TEMPLE: Incentivizing Temporal Understanding of Video Large Language Models via Progressive Pre-SFT Alignment
Shicheng Li (Peking University), Xu Sun (Kuaishou Technology)
OptimizationTransformerSupervised Fine-TuningReinforcement LearningVision Language ModelVideoTextMultimodality
🎯 What it does: Propose the TEMPLE framework, leveraging direct preference optimization (DPO) and Pre-SFT alignment to significantly enhance the temporal understanding capabilities of video LLMs.
Temporal and Spatial Representation Learning for Multimodal Low-Beam 3D Object Detection
Lin Wang (East China Normal University), Jing Zhao (East China Normal University)
Object DetectionAutonomous DrivingRepresentation LearningConvolutional Neural NetworkTransformerImageMultimodalityPoint Cloud
🎯 What it does: Propose a multi-modal 3D detection framework for low-beam LiDAR, enhancing perception accuracy of sparse LiDAR point clouds through spatiotemporal representation learning.
Temporal Calibrating and Distilling for Scene-Text Aware Text-Video Retrieval
Zhiqian Zhao (Hangzhou Dianzi University), Chenggang Yan (University of Oulu)
RetrievalKnowledge DistillationRepresentation LearningTransformerVision Language ModelContrastive LearningVideoTextMultimodality
🎯 What it does: Propose a text-video retrieval framework TCD based on scene text, using a window OCR captioner to compress dense text into window-level captions, aligning video frames with OCR captions via an HSC module, and enhancing retrieval performance through a CTCD module that distills temporal cues.
Temporal Dynamics Enhancer for Directly Trained Spiking Object Detectors
Fan Luo (Institute of Automation, Chinese Academy of Sciences), Yanfeng Lu (Institute of Automation, Chinese Academy of Sciences)
Object DetectionSpiking Neural NetworkImageTime Series
🎯 What it does: Proposed the Temporal Dynamics Enhancer (TDE) module to enhance the temporal information modeling capability of directly trained spiking neural networks (SNNs) in object detection.
Temporal Inconsistency Guidance for Super-resolution Video Quality Assessment
Yixiao Li (Beihang University), Wei Zhou (Cardiff University)
RestorationSuper ResolutionConvolutional Neural NetworkRecurrent Neural NetworkTransformerOptical FlowVideo
🎯 What it does: Propose a super-resolution video quality assessment framework based on time inconsistency guidance, quantifying and utilizing inter-frame inconsistency information to enhance SR video quality prediction.
Temporal Object-Aware Vision Transformer for Few-Shot Video Object Detection
Yogesh Kumar (Indian Institute of Technology Jodhpur), Anand Mishra (Indian Institute of Technology Jodhpur)
Object DetectionTransformerVision Language ModelVideo
🎯 What it does: This paper proposes a few-shot video object detection framework based on language-aligned vision transformers and object-aware temporal fusion mechanisms, which can achieve high-precision detection of novel targets in videos with only a few support images provided.
Temporal Properties of Conditional Independence in Dynamic Bayesian Networks
Rajab Aghamov (Technische Universitat Dresden), Isa Vialard (Max Planck Institute for Software Systems, Saarland Informatics Campus)
🎯 What it does: This paper studies the formal verification of the time-evolving conditional independence (CI) in dynamic Bayesian networks (DBN), proposing the use of two specification frameworks: linear temporal logic (LTL) and non-deterministic Büchi automata (NBA), and conducting complexity analysis for two categories of problems: structured CI and stochastic CI.
Temporal-Consistent Video Restoration with Pre-trained Diffusion Models
Hengkang Wang (University of Minnesota), Ju Sun (Amazon.com, Inc.)
RestorationDiffusion modelOptical FlowVideo
🎯 What it does: Designed and implemented a Maximum A Posteriori (MAP) framework based on pre-trained diffusion models for unsupervised video restoration, addressing issues such as approximation errors, temporal consistency, and computational efficiency.
Tensor Decomposition and Language Description for Open-Vocabulary Object Detection
Qiuyu Liang (Inner Mongolia University), Yongqiang Zhang (Inner Mongolia University)
Object DetectionKnowledge DistillationConvolutional Neural NetworkLarge Language ModelVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Propose a novel open-domain object detection framework named TLDet, comprising three modules: Low-Rank Proposal Filtering (LRPF), Core Tensor Distillation (CTKD), and Language Description Enhancement (LDE).
Tensorized Label Learning via Balanced Tensor Regression
Guangyu Yang (Xidian University), Rui Wang (Harbin Engineering University)
Representation LearningMultimodality
🎯 What it does: Propose a collaborative label learning framework TLL-BTR based on balanced tensor regression, utilizing the probabilistic properties of the anchor graph to simultaneously learn anchors and sample labels, achieving one-click multi-view clustering;
TermGPT: Multi-Level Contrastive Fine-Tuning for Terminology Adaptation in Legal and Financial Domains
Yidan Sun (Zhejiang University), Shenglin Ben (Zhejiang University)
Domain AdaptationRepresentation LearningTransformerSupervised Fine-TuningContrastive LearningTextFinance Related
🎯 What it does: Propose the TermGPT framework, achieving term adaptation in legal and financial domains through multi-layer contrastive learning, addressing term ambiguity caused by isotropy in LLM embedding spaces.
Test-driven Reinforcement Learning in Continuous Control
Zhao Yu (Xi'an Jiaotong University), Liangjun Ke (Xi'an Jiaotong University)
OptimizationRobotic IntelligenceReinforcement Learning
🎯 What it does: Proposes a test-driven reinforcement learning (TdRL) framework, using a set of pass-fail and indicative test functions evaluated through trajectories to replace traditional scalar reward functions, defining task objectives while providing learning guidance; implements a complete algorithm under maximum entropy RL (SAC/PPO);
Test-time Diverse Reasoning by Riemannian Activation Steering
Ly Tran Ho Khanh (Chinese University of Hong Kong), Viet Anh Nguyen (University of Hong Kong)
OptimizationComputational EfficiencyRepresentation LearningTransformerLarge Language ModelTextBenchmark
🎯 What it does: Proposed an unsupervised activation regulation method for language models during inference (SPREAD) to enhance the diversity and accuracy of best-N sampling.
Test-Time Preference Optimization for Image Restoration
Bingchen Li (University Of Science And Technology Of China), Zhibo Chen (Huawei Noah's Ark Lab)
RestorationDiffusion modelImage
🎯 What it does: Propose a Test-Time Preference Optimization (TTPO) framework that generates candidate preference images online during testing using diffusion inverse editing, automatically selects superior and inferior samples via no-reference image quality assessment, and guides the restoration process during diffusion denoising to enhance human preference quality;
Test-time Prompt Intervention
Chenxu Yang (Chinese Academy of Sciences), Weiping Wang (Chinese Academy of Sciences)
Computational EfficiencyLarge Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: Proposed the test-time prompt intervention framework PI, dynamically controlling the LLM's chain-of-thought reasoning path to reduce redundant steps and improve inference efficiency.
Test-Time Reinforcement Learning for GUI Grounding via Region Consistency
Yong Du (Zhejiang University), Yongliang Shen (SF Technology)
Object DetectionReinforcement LearningVision Language ModelImage
🎯 What it does: Proposes two unsupervised test-time improvement methods—GUI-RC (spatial voting based on multiple sampling) and GUI-RCPO (test-time reinforcement learning with region consistency rewards)—which can enhance GUI localization accuracy without increasing labeled data or retraining.
Testing Under Strategic Manipulation: Mechanism Design for Human and AI Institutions
Xiaoyun Qiu (Dartmouth College), Liren Shan (Toyota Technological Institute at Chicago)
OptimizationFinance Related
🎯 What it does: This paper studies how to design testing institutions (the tests themselves and testing procedures) to achieve optimal selection outcomes in a multi-criteria and strategic agent environment, focusing on analyzing two institutional structures—-independent bureaucrats and joint bureaucrats—and their impacts on test order, randomization, and personalization.
Text-based Aerial-Ground Person Retrieval
Xinyu Zhou (Soochow University), Mang Ye (Wuhan University)
RetrievalTransformerLarge Language ModelMixture of ExpertsVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: This paper proposes a text-driven aerial and ground person retrieval task (TAG-PR), and designs the corresponding dataset TAG-PEDES as well as a novel retrieval framework TAG-CLIP.
Text-Guided Channel Perturbation and Pre-Trained Knowledge Integration for Unified Multi-Modality Image Fusion
Xilai Li (Foshan University), Weijun Jiang (Foshan University)
TransformerVision Language ModelTextMultimodalityMagnetic Resonance ImagingComputed TomographyPositron Emission Tomography
🎯 What it does: Proposed a unified multi-modal image fusion framework named UP-Fusion, combining channel perturbation with pre-trained knowledge to suppress redundant cross-modal features and enhance fusion quality.
Text-guided Controllable Diffusion for Realistic Camouflage Images Generation
Yuhang Qian (Nanjing University of Aeronautics and Astronautics), Jie Qin (Nanjing University of Aeronautics and Astronautics)
GenerationPrompt EngineeringVision Language ModelDiffusion modelImageText
🎯 What it does: Proposed a controllable text-guided camouflage image generation method called CT-CIG, which can generate visually realistic camouflage images that are logically consistent with the background based on text prompts and masks.
Text-Guided Gradient Refinement: Resolving Multimodal Gradient Conflicts to Boost Adversarial Attacks on Vision-Language Models
Yuyang Huang (ByteDance Inc.), Yuren Zhang (ByteDance Inc.)
Adversarial AttackVision Language ModelDiffusion modelMultimodality
🎯 What it does: Studied transfer attacks against vision-language models in black-box scenarios, and proposed a Text-Guided Gradient Refinement (TGGR) framework based on conflict-aware gradient projection to enhance attack effectiveness.
Text-to-Scene with Large Reasoning Models
Frédéric Berdoz (ETH Zurich), Roger Wattenhofer (ETH Zurich)
GenerationData SynthesisRetrievalTransformerLarge Language ModelVision Language ModelTextPoint CloudMeshRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Developed a text-to-scene synthesis framework called Reason-3D based on a large inference model, which can retrieve and place 3D objects from natural language descriptions in an unsupervised manner, generating complete indoor and outdoor environments.
TextGround4M: A Prompt-Aligned Dataset for Layout-Aware Text Rendering
Dongxing Mao (Central South University), Alex Jinpeng Wang (Central South University)
Image TranslationGenerationData SynthesisTransformerSupervised Fine-TuningPrompt EngineeringVision Language ModelMultimodalityBenchmark
🎯 What it does: Constructed a prompt-aligned text rendering dataset named TextGround4M with over 4 million entries, and fine-tuned a lightweight autoregressive text-to-image (T2I) model on this dataset;
TextShield-R1: Reinforced Reasoning for Tampered Text Detection
Chenfan Qu (South China University of Technology), Lianwen Jin (Ant Group)
Anomaly DetectionExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: This paper proposes TextShield-R1, a reinforcement learning-based multimodal large language model (MLLM) framework for detecting and explaining tampered text, enhancing detection, localization, and explanation capabilities through Forensic Continual Pre-training, Group Relative Policy Optimization (GRPO), and OCR Rectification, while constructing the TFR benchmark containing 45k multilingual, diverse tampering methods images;
Textual Self-Attention Network: Test-Time Preference Optimization Through Textual Gradient-Based Attention
Shibing Mo (Xidian University), Jing Liu (Xidian University)
OptimizationReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: Analyze, weight, and synthesize multiple candidate answers generated by LLMs using a parameter-free update text self-attention network (TSAN) during inference to produce answers more aligned with human preferences.
Textured Geometry Evaluation: Perceptual 3D Textured Shape Metric via 3D Latent-Geometry Network
Tianyu Luan (State University of New York at Buffalo), Junsong Yuan (State University of New York at Buffalo)
Computational EfficiencyRepresentation LearningConvolutional Neural NetworkTransformerMeshBenchmark
🎯 What it does: Designed a metric called TGE for directly assessing the quality of textured 3D meshes.
TFD-Net: Towards Intelligent Time-Frequency Mode Decomposition with Practical Applications
Pingping Pan (Guangdong Laboratory of Artificial Intelligence and Digital Economy), Jinyi Liu (Shenzhen Kaihong Digital Industry Development Co., Ltd)
SegmentationConvolutional Neural NetworkTransformerTime SeriesPhysics Related
🎯 What it does: This paper proposes a deep learning-based time-frequency analysis and mode decomposition framework called TFD-Net, which can automatically generate multi-resolution time-frequency representations and achieve signal component separation and reconstruction.
TFRank: Think-Free Reasoning Enables Practical Pointwise LLM Ranking
Yongqi Fan (East China University of Science and Technology), Tong Ruan (University of Chinese Academy of Sciences)
RetrievalTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextChain-of-Thought
🎯 What it does: This paper proposes TFRank, a pointwise reasoning ranking framework based on small-scale LLMs, which does not generate reasoning chains during the inference phase;
TG-Field: Geometry-Aware Radiative Gaussian Fields for Tomographic Reconstruction
Yuxiang Zhong (Shenzhen University), Hui Huang (Shenzhen University)
Neural Radiance FieldGaussian SplattingBiomedical DataComputed Tomography
🎯 What it does: Proposed a geometry-aware Gaussian deformation framework called TG-Field, which can achieve static and dynamic CT reconstruction under extremely sparse projections;
TGCA-LLM: Time-Aware Graph-Text Contrastive Alignment for Enhancing LLMs in Temporal Knowledge Graph Completion
Zexuan Wan (Beijing University of Posts and Telecommunications), Bin Wu (Beijing University of Posts and Telecommunications)
Representation LearningGraph Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningContrastive LearningTextGraph
🎯 What it does: Proposed the TGCA-LLM framework, combining time-aware graph-text contrastive alignment with a two-stage instruction tuning approach to enhance the reasoning capability of large language models in temporal knowledge graph completion (TKGC) tasks.
TGCD: A Framework for Generalized Category Discovery in Time-Series Data
Chandan Gautam (Institute for Infocomm Research, Agency for Science, Technology and Research), Savitha Ramasamy (National University of Singapore)
ClassificationTransformerTime Series
🎯 What it does: This paper proposes the TGCD framework, achieving the goal of generic category discovery in time series data.
TGDD: Trajectory Guided Dataset Distillation with Balanced Distribution
Fengli Ran (Chongqing University of Posts and Telecommunications), Bin Xiao (Chongqing University of Posts and Telecommunications)
Knowledge DistillationData-Centric LearningImage
🎯 What it does: Proposes a training trajectory-guided dataset distillation method called TGDD to compress large-scale datasets while maintaining efficient performance.
The Avengers: A Routing Recipe for Collective Intelligence in Language Models
Yiqun Zhang (Northeastern University), Shuyue Hu (Singapore Management University)
Computational EfficiencyKnowledge DistillationTransformerLarge Language ModelMixture of ExpertsTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Propose the Avengers framework, which employs an lightweight routing mechanism combining embedding-clustering-scoring-voting to aggregate the collective intelligence of 10 7B open-source models into a single answer.