AAAI 2026 Papers — Page 20
AAAI Conference on Artificial Intelligence · 4149 papers
Information Theoretic Optimal Surveillance for Epidemic Prevalence in Networks
Ritwick Mishra (University of Virginia), Anil Vullikanti (University of Virginia)
OptimizationGraphBiomedical Data
🎯 What it does: This paper proposes an information-theoretic disease monitoring framework called TESTPREV, aiming to select a set of nodes to maximize the mutual information between their observations and the pandemic scale distribution; it also provides a greedy solution method called GREEDYMI for different network structures (trees, paths, one-hop propagation) and general networks. Experimental results demonstrate that this method outperforms traditional degree- or vulnerability-based baselines on various synthetic networks and real-world ICU contact networks, significantly reducing the variance of pandemic scale.
Information-Theoretic Minimal Sufficient Representation for Multi-Domain Knowledge Graph Completion
Jiawei Sheng (Chinese Academy of Sciences), Tingwen Liu (Chinese Academy of Sciences)
Representation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposes the IMKGC framework, which leverages the information bottleneck principle to learn the minimal sufficient representation for multi-domain knowledge graph completion, significantly enhancing the effectiveness of cross-domain entity and relation reasoning.
Informative Subgraph Extraction with Deep Reinforcement Learning for Drug-Drug Interaction Prediction
Jiancong Xie (Sun Yat-sen University), Yuedong Yang (Sun Yat-sen University)
Drug DiscoveryGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: This study proposes RISE-DDI, an interpretable subgraph extraction framework based on deep reinforcement learning, for predicting drug-drug interactions (DDI) and explaining their mechanisms.
Infrared-Privileged UAV Detection via Cross-Modal Vector-Quantization
Zhibo Lou (Jiangxi University of Finance and Economics), Yuming Fang (Jiangxi University of Finance and Economics)
Object DetectionConvolutional Neural NetworkAuto EncoderImageMultimodality
🎯 What it does: Propose a drone target detection framework that utilizes infrared images as privileged information during training. It employs cross-modal vector quantization (CMVQ) to predict and reconstruct infrared codebook indices from RGB features, enabling the acquisition of infrared complementary features without infrared input during inference.
Injection Without Distortion: Geometrically Constrained Knowledge Enhancement for Vision-Language Models
Zhongze Wu, Yueyi Luo
Representation LearningImageTextMultimodalityBenchmark
🎯 What it does: Propose the GeCoin framework, which securely injects external knowledge into VLMs via zero-space projection without training, enhancing hierarchical semantic understanding and avoiding performance collapse.
Injection, Attack and Erasure: Revocable Backdoor Attacks via Machine Unlearning
Baogang Song (Wuhan University of Technology), Zizhuo Yu (Wuhan University of Technology)
Adversarial AttackConvolutional Neural NetworkImage
🎯 What it does: Proposed a backdoor attack method that can be actively revoked through machine unlearning, achieving the first realization of recoverable backdoors.
Inpaint-Anywhere: Zero-Shot Multi-Identity Inpainting with Efficient Diffusion Transformer
Junsheng Luan (Zhejiang University), Wei Xing (Zhejiang University)
RestorationGenerationTransformerDiffusion modelImage
🎯 What it does: Proposed a multi-identity zero-shot image inpainting method called Inpaint-Anywhere.
Insert Anything: Image Insertion via In-Context Editing in DiT
Wensong Song (Zhejiang University), Yi Yang (Zhejiang University)
GenerationTransformerVision Language ModelDiffusion modelImageText
🎯 What it does: Proposed a unified reference image insertion framework called Insert Anything, which can insert any reference image elements into the target image via mask or text instructions.
Instance Dependent Testing of Samplers Using Interval Conditioning
Rishiraj Bhattacharyya (University of Birmingham), Sayantan Sen (National University of Singapore)
Anomaly DetectionComputational Efficiency
🎯 What it does: Proposes an instance-dependent sampler verification method based on the interval conditioning (interval conditioning) model, which can efficiently verify the distribution correctness of samplers on infinite domains, and implements a practical tester Lachesis for inverse transform samplers (e.g., geometric, binomial, Poisson distributions).
Instance Generation for Meta-Black-Box Optimization Through Latent Space Reverse Engineering
Chen Wang (South China University of Technology), Zeyuan Ma (South China University of Technology)
OptimizationMeta LearningAuto EncoderTabularBenchmark
🎯 What it does: Proposed a method generating diverse training instances through latent space reverse engineering (LSRE), constructing a new Diverse-BBO benchmark to enhance the generalization capability of Meta-Black-Box Optimization (MetaBBO).
Instance-Guided Scene Adaptation for Unsupervised Person Search
Linfeng Qi (Dalian Maritime University), Jiqing Zhang (Dalian Maritime University)
RetrievalDomain AdaptationContrastive LearningImage
🎯 What it does: Proposed an instance-guided scene adaptation framework (IGSA) to achieve unsupervised cross-domain person search.
InstructDubber: Instruction-based Alignment for Zero-shot Movie Dubbing
Zhedong Zhang (Hangzhou Dianzi University), Yuankai Qi (Tsinghua University)
GenerationRecurrent Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningVideoMultimodalityAudio
🎯 What it does: Proposes InstructDubber, an instruction-based movie dubbing alignment method that generates detailed speaking rate and emotion instructions via multimodal large language models, and achieves lip-sync and emotion alignment without visual preprocessing through instruction-driven duration extraction and emotion calibration modules.
Instruction-Guided Cross-Modal Clustering for Training-Free Visual Token Pruning in Vision-Language Models
Yunqian Yu (University of Electronic Science and Technology of China), Lin Zuo (University of Electronic Science and Technology of China)
Computational EfficiencyRepresentation LearningTransformerVision Language ModelTextMultimodality
🎯 What it does: Proposed an untrained, instruction-based cross-modal clustering visual token pruning method called ICCTP, which can significantly reduce the number of visual tokens in large vision-language models and lower inference costs.
InteChar: A Unified Oracle Bone Character List for Ancient Chinese Language Modeling
Xiaolei Diao (Jilin University), Hao Xu (Jilin University)
Representation LearningData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: Construct a unified and scalable Oracle Bone character set named InteChar, and build the OracleCS corpus based on this character set along with expert annotations and LLM-assisted generation, thereby enhancing the training and evaluation of ancient Chinese language models.
Integral-based Knockoffs Inference for Partially Linear Models
Hao Wang (Huazhong Agricultural University), Hong Chen (Huazhong Agricultural University)
Tabular
🎯 What it does: Proposed an integral knockoff inference method (IKO) for partial linear models (PLM), achieving FDR control for variable selection in both linear and nonlinear components;
Integrating Diverse Assignment Strategies into DETRs
Yiwei Zhang (Chinese Academy of Sciences), Zhipeng Zhang (Shanghai Jiao Tong University)
Object DetectionTransformerImage
🎯 What it does: Insert a lightweight LoRA branch into the DETR decoder to parallelly integrate various one-to-many label assignment strategies, enhancing the diversity of training supervision.
Integrating Reweighted Least Squares with Plug-and-Play Diffusion Priors for Noisy Image Restoration
Ji Li (Capital Normal University), Chao Wang (China Agricultural University)
RestorationSuper ResolutionDiffusion modelImage
🎯 What it does: Proposed a plug-and-play image restoration framework based on diffusion models, leveraging the gGSM model for noise modeling and employing the ℓq‖ data fidelity term to achieve robust denoising against arbitrary noise (including impulse noise);
INTENT: Invariance and Discrimination-aware Noise Mitigation for Robust Composed Image Retrieval
Zhiwei Chen (Shandong University), Yinwei Wei (Shandong University)
RetrievalTransformerGenerative Adversarial NetworkContrastive LearningMultimodality
🎯 What it does: This paper proposes a framework named INTENT for robust compositional image retrieval (CIR) under the presence of noisy triplet correspondences (NTC).
Intention Chain-of-Thought Prompting with Dynamic Routing for Code Generation
Shen Li (Chongqing University), Meng Yan (Chongqing University)
AI Code AssistantTransformerPrompt EngineeringTextChain-of-Thought
🎯 What it does: Designed and implemented a code generation framework called RoutingGen based on difficulty-aware dynamic routing, using few-shot prompts for simple tasks and Intention Chain-of-Thought (ICoT) two-phase reasoning for complex tasks.
Intention-Aware Diffusion Model for Pedestrian Trajectory Prediction
Yu Liu (Southern University of Science and Technology), He Kong (Southern University of Science and Technology)
Object TrackingGenerationAutonomous DrivingTransformerDiffusion modelVideoTime SeriesSequential
🎯 What it does: This paper proposes a pedestrian trajectory prediction framework based on diffusion models, which can simultaneously model short-term fine-tuning intentions and long-term goal intentions, and achieve more accurate trajectory generation through soft mask guidance and residual noise correction.
Intention-Guided Cognitive Reasoning for Egocentric Long-Term Action Anticipation
Qiaohui Chu (Harbin Institute of Technology), Liqiang Nie (Harbin Institute of Technology)
TransformerLarge Language ModelReinforcement LearningPrompt EngineeringVision Language ModelVision-Language-Action ModelVideoChain-of-Thought
🎯 What it does: Propose a two-stage long-term action anticipation framework named INSIGHT, which first extracts semantically rich visual features from hand-object interaction (HOI) regions and uses a verb-noun co-occurrence matrix for semantic correction, then employs a reinforcement learning-based explicit cognitive reasoning module (think → reason → answer) on a large vision-language model to complete action prediction.
IntentMotion: Learning Intent-Aware Human Motion from Language in 3D Scenes
Wenfeng Song (Beijing Information Science and Technology University), Shuai Li (Beihang University)
GenerationData SynthesisVision-Language-Action ModelDiffusion modelTextMultimodality
🎯 What it does: Propose the IntentMotion framework to generate 3D human motion that aligns with natural language instructions, significantly enhancing the realism of physical contact.
Inter-Client Dependency Recovery with Hidden Global Components for Federated Traffic Prediction
Hang Zhou (Nanjing University of Science and Technology), Chen Gong (Shanghai Jiao Tong University)
Federated LearningGraph Neural NetworkTime Series
🎯 What it does: Proposed a traffic flow prediction method FedHINT under the federated learning framework, which generates proxy nodes from hidden global components within each client's internal data to restore cross-client dependencies and improve prediction accuracy.
Interactive Evaluation of Large Language Models for Multi-Requirement Software Engineering Tasks
Dimitrios Rontogiannis (Max Planck Institute for Software Systems), Dimitrios Gunopulos (National and Kapodistrian University of Athens)
AI Code AssistantTransformerLarge Language ModelPrompt EngineeringTextGraphBenchmark
🎯 What it does: Propose an interactive evaluation framework based on a demand dependency graph, using minimized, targeted prompts to progressively guide LLMs in error correction and completing multi-demand software engineering tasks.
InterCoser: Interactive 3D Character Creation with Disentangled Fine-Grained Features
Yi Wang (Tianjin University), Kun Li (ByteDance China)
GenerationDiffusion modelGaussian SplattingMesh
🎯 What it does: Propose the InterCoser framework to achieve decoupled 3D character generation and editing based on user interaction, supporting layered drawing and 3D proxy editing;
Interest-driven Deep Multi-modal Clustering
Guoliang Zou (Zhengzhou University), Shizhe Hu (Zhengzhou University)
Representation LearningAuto EncoderContrastive LearningMultimodality
🎯 What it does: Proposed an interest-driven deep multi-modal clustering method (IDMC), which dynamically focuses on important samples through sample interest values to enhance clustering performance.
Interest-Shift-Aware Logical Reasoning for Efficient Long-Sequence Recommendation
Fei Li (Northeastern University), Guibing Guo (Northeastern University)
Recommendation SystemContrastive LearningSequential
🎯 What it does: Propose the ELECTOR framework to address the interest drift and computational complexity issues in logical reasoning methods for long-sequence recommendations.
Intermediate N-Gramming: Deterministic and Fast N-Grams for Large N and Large Datasets
Ryan R. Curtin (Booz Allen Hamilton), Priyanka Ranade (Laboratory for Physical Sciences)
Computational EfficiencyTextBiomedical Data
🎯 What it does: This paper proposes a multi-stage algorithm called Intergrams for rapidly and deterministically retrieving the top-k most frequent n-grams from large-scale datasets.
InterMoE: Individual-Specific 3D Human Interaction Generation via Dynamic Temporal-Selective MoE
Lipeng Wang (Beihang University), Lu Sheng (Beihang University)
GenerationTransformerMixture of ExpertsDiffusion modelText
🎯 What it does: Propose a text-driven 3D human interactive generation framework called InterMoE based on dynamic time selection Mixture-of-Experts (MoE), which can achieve high semantic fidelity while preserving individual identity features.
Interpolated Stochastic Interventions Based on Propensity Scores, Target Policies and Treatment-Specific Costs
Johan de Aguas (University of Oslo)
TabularElectronic Health Records
🎯 What it does: This paper proposes two families of interpolation-based random interventions grounded in cost and goal strategies, leveraging information projection to obtain Boltzmann-Gibbs couplings, and deriving their efficient influence functions and first-order corrected estimators;
Interpretable and Robust Behavior Abstraction via Environment-Disentangled Heterogeneous Graph
Zhibin Ni (Tsinghua University), Xibin Zhao (Tsinghua University)
Explainability and InterpretabilityRepresentation LearningGraph Neural NetworkAuto EncoderGenerative Adversarial NetworkGraph
🎯 What it does: Propose the EDHGNN model to achieve interpretable and robust behavior abstraction tasks;
Interpretable Reward Model via Sparse Autoencoder
Shuyi Zhang (University of Science and Technology of China), Xiang Wang (University of Science and Technology of China)
Explainability and InterpretabilityRepresentation LearningReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningAuto EncoderText
🎯 What it does: This study proposes the Sparse Autoencoder-Enhanced Reward Model (SARM), which maps LLM hidden states to a sparse, disentangled feature space by embedding a pre-trained sparse autoencoder into the reward model, and uses a linear head to aggregate these features to obtain interpretable reward scores.
Interpreting Fedspeak with Confidence: A LLM-Based Uncertainty-Aware Framework Guided by Monetary Policy Transmission Paths
Rui Yao (Hong Kong University of Science and Technology Guangzhou), Hao Wang (Hong Kong University of Science and Technology Guangzhou)
Explainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningTextFinance Related
🎯 What it does: This paper proposes a 'Fedspeak' parsing framework based on large language models (LLMs), integrating monetary policy transmission paths for domain reasoning and incorporating a dynamic uncertainty decoding module.
Intervention-Aware Time Series Modeling: Capturing and Evaluating Feature Dependencies
Ibrahim Delibasoglu (Linkoping University), Mattias Tiger (Linkoping University)
Explainability and InterpretabilityAuto EncoderTime Series
🎯 What it does: Designed a conditional variational autoencoder, CFOR-VAE, for interpretable reconstruction when performing feature-level interventions in multivariate time series.
Intra-Class Unbiased Prototype Aggregation and Classifier Collaboration for Personalized Federated Learning
Hao Zheng (Central South University), Boyu Wang (Central South University)
ClassificationFederated LearningRepresentation LearningImage
🎯 What it does: Propose the FedIC framework, which first clusters client prototypes for each class, then aggregates prototypes only within similar prototype clusters, and performs classifier collaborative optimization within the same cluster to address prototype aggregation bias in non-IID environments.
Intra-Image Mining and Symmetric Maximum Concept Matching for Few Shot Out-of-Distribution Detection
Kaixiang Chen (Southeast University), Hui Xue (Southeast University)
Anomaly DetectionTransformerPrompt EngineeringContrastive LearningImageTextMultimodalityBenchmark
🎯 What it does: This paper proposes a lightweight few-shot OOD detection framework named IIM (Intra-Image Mining), which selects the top k local patches most relevant to the class prototype within a single image as positive samples to optimize local feature prototypes. It also selects the top k patches from the remaining patches that have high similarity to the prototype but mismatch the category as ID-like OOD samples to construct OOD prototypes. Meanwhile, it introduces a bidirectional reasoning strategy called Symmetric Maximum Concept Matching (S-MCM), which considers both image-text and text-image similarity.
Introducing Decomposed Causality with Spatiotemporal Object-Centric Representation for Video Classification
Yachong Zhang (Shandong University), Xiangxu Meng (Shandong University)
ClassificationObject DetectionObject TrackingTransformerVideoBenchmark
🎯 What it does: Proposes a SurdCRL model based on SURD causality theory, which decomposes object-background interactions in videos into synergistic, exclusive, and redundant causal components, achieving fine-grained event recognition through object-centric spatiotemporal modeling and causal interventions.
Introducing Visual Scenes and Reasoning: A More Realistic Benchmark for Spoken Language Understanding
Di Wu (Xinjiang University), Xuelong Li (Institute of Artificial Intelligence of China Telecom (TeleAI))
GenerationExplainability and InterpretabilityData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringVision Language ModelDiffusion modelImageTextMultimodalityBenchmarkAudio
🎯 What it does: Constructed the first voice-language understanding dataset VRSLU that simultaneously includes user environment images and reasoning processes, and proposed the LR-Instruct instruction template to let the model first perform label prediction and then generate reasoning.
Intuitive Thinking: Expanding Large Language Models’ Thinking for Rapid Decision-Making on Candidate Corrections in Chinese Grammar Error Correction
Lintao Long (Guizhou University), Qihang Fu (Guizhou University)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose a lightweight Expanding Intuitive Thinking Model (ExIT), which injects confidence information from large language models (LLMs) into the embedding space of error correction candidates and performs joint evaluation of source sentences and correction candidates at a global level, mimicking human intuitive rapid decision-making processes;
Invariant Conditional Molecular Generation Under Distribution Shift
Chunyu Hu (Nankai University), Ziwei Zhang (Beihang University)
Domain AdaptationDrug DiscoveryGraph Neural NetworkTransformerDiffusion modelGraph
🎯 What it does: Propose a framework called IC-MOL for conditional molecular generation under distribution drift conditions, combining invariant learning with graph diffusion models;
Invariant Feature Learning for Counterfactual Watch-time Prediction in Video Recommendation
Chenghou Jin (Fudan University), Shuigeng Zhou (Fudan University)
Recommendation SystemRepresentation LearningVideo
🎯 What it does: Proposes a solution to the duration bias problem in video watch time prediction, constructing the Duration-Invariant Feature Learning (DIFL) framework
Invariant Representation Learning for Memory Behavior Modeling via Adaptive Environment Separation
Xiaoxuan Shen (Central China Normal University), Jianwen Sun (Central China Normal University)
Representation LearningTransformerContrastive LearningSequential
🎯 What it does: Proposed the I-Mem framework, which learns transferable memory behavior models without requiring environmental labels by leveraging self-supervised contrastive learning and decorrelation constraints.
Inverse Optimal Transport for Efficient Adaptation of Vision-Language Models
Shupeng Qiu (Sun Yat-Sen University), Chuan-Xian Ren (Sun Yat-Sen University)
Domain AdaptationComputational EfficiencyRepresentation LearningVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: This paper proposes a lightweight adaptation method called IOTA, which utilizes an inverse optimal transport framework for zero/few-shot reasoning and adaptation of Vision-Language models;
Investigating Data Pruning for Pretraining Biological Foundation Models at Scale
Yifan Wu (Chinese University of Hong Kong), Yu Li (Chinese University of Hong Kong)
Data-Centric LearningBiomedical Data
🎯 What it does: This paper proposes a post-hoc influence-guided data reduction framework to select a small yet effective subset during the pre-training of biological foundation models (BioFMs), maintaining or even improving model performance under extreme data pruning (99%).
Investigating Prosocial Behavior Theory in LLM Agents Under Policy-Induced Inequities
Yujia Zhou (Tsinghua University), Yiqun Liu (Tsinghua University)
Explainability and InterpretabilityTransformerLarge Language ModelAgentic AITextTabular
🎯 What it does: Simulate prosocial behaviors in multiple scenarios using LLM agents through the PROSIM framework, and systematically evaluate their sensitivity to unfair policies.
Invisible Triggers, Visible Threats! Road-Style Adversarial Creation Attack for Visual 3D Detection in Autonomous Driving
Jian Wang (Xi'an Jiaotong University), Fan Li (Xi'an Jiaotong University)
Autonomous DrivingAdversarial AttackGenerative Adversarial NetworkImagePoint Cloud
🎯 What it does: Designed and implemented a visualization-based 3D detection adversarial creation attack (AdvRoad) that places naturally appearing adversarial stickers on roads to induce false targets.
IO-RAE: Information-Obfuscation Reversible Adversarial Example for Audio Privacy Protection
Jiajie Zhu, Chi-Man Pun (Xiamen University of Technology)
Safty and PrivacyAdversarial AttackTransformerLarge Language ModelAudio
🎯 What it does: Propose the IO-RAE framework, leveraging reversible adversarial examples to achieve audio privacy protection
IPDA:Intelligent Perception Delay Alignment Method Based on Spatio-Temporal Co-Sensing Calibration
Jianhang Liu (China University of Petroleum (East China)), Runda Fan (Shandong Key Laboratory of Intelligent Oil & Gas Industrial Software)
Autonomous DrivingConvolutional Neural NetworkGraph Neural NetworkTransformerPoint CloudBenchmark
🎯 What it does: Propose a spatiotemporal co-sensory calibration method named IPDA for addressing data inconsistency caused by communication delays in multi-agent collaborative perception. First, historical alignment attention is utilized to achieve temporal delay compensation, and then a difference quantization co-sensory calibration network optimizes spatial pose errors.
IPFormer: Instance Prompt-guided Transformer for Multi-modal Multi-shot Video Understanding
Yujia Liang (JD Explore Academy Deepeleph Intelligent Technology), Zhicheng Wang (Huazhong University Of Science And Technology)
RecognitionComputational EfficiencyTransformerPrompt EngineeringVision Language ModelVideoTextMultimodalityBenchmark
🎯 What it does: Propose the MultiClip-Bench dataset and design IPFormer (Instance Prompt-guided Transformer) to achieve instance-aware understanding of multimodal multi-camera videos.
IQGS: Instance Query-based Gaussian Segmentation
Yichao Gao (Institute Of Computing Technology Chinese Academy Of Science), Feng Dai (Institute Of Computing Technology Chinese Academy Of Science)
SegmentationTransformerGaussian SplattingImage
🎯 What it does: For scenes reconstructed by 2D Gaussian Splatting, the IQGS (Instance Query-based Gaussian Segmentation) framework is proposed to achieve panoramic instance segmentation.
IROTE: Human-like Traits Elicitation of Large Language Model via In-Context Self-Reflective Optimization
Yuzhuo Bai (Tsinghua University), Xing Xie (Microsoft Research Asia)
OptimizationTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: This paper proposes the IROTE method, which enables large language models (LLMs) to stably and transferably exhibit specified human traits through self-reflective text without fine-tuning.
Is Symbolic Music a Specific Language? Exploring Inspiration-to-Structure Machine Composition via LLMs
Zhejing Hu (Hong Kong Polytechnic University), Gong Chen (Hong Kong Polytechnic University)
GenerationData SynthesisTransformerLarge Language ModelSupervised Fine-TuningContrastive LearningSequentialAudio
🎯 What it does: Proposed the Inspiration-to-Structure (IoS) framework, leveraging a three-stage cognitive process and conversational data to enable LLMs to generate structured musical passages from melodic inspiration.
Is the Information Bottleneck Robust Enough? Towards Label-Noise Resistant Information Bottleneck Learning
Yi Huang, Jianxin Li (Hong Kong University Of Science And Technology)
ClassificationRepresentation LearningImageGraph
🎯 What it does: Proposed the LaT-IB method, which splits representations into clean label space and noisy label space, and designs a three-stage training framework to learn Minimal-Sufficient-Clean representations under label noise environments.
Is Your (Reasoning) Multimodal Language Model Vulnerable Toward Distractions?
Ming Liu (Iowa State University), Wensheng Zhang (Iowa State University)
Adversarial AttackPrompt EngineeringVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Constructed the I-ScienceQA benchmark to systematically evaluate the robustness of Vision-Language Models under visual and textual perturbation conditions.
IS-Bench: Evaluating Interactive Safety of VLM-Driven Embodied Agents in Daily Household Tasks
Xiaoya Lu (Shanghai Jiao Tong University), Jing Shao (Shanghai Artificial Intelligence Laboratory)
Safty and PrivacyRobotic IntelligenceTransformerPrompt EngineeringVision Language ModelTextMultimodalityBenchmarkChain-of-Thought
🎯 What it does: This paper proposes the IS-Bench benchmark, specifically designed to evaluate the interaction safety of vision-language model-driven embodied agents in daily household tasks, constructing 161 interactive scenarios, 388 safety risks, and introducing a process-oriented evaluation method;
iSeal: Encrypted Fingerprinting for Reliable LLM Ownership Verification
Zixun Xiong (Stevens Institute of Technology), Hao Wang (Stevens Institute of Technology)
Safty and PrivacyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Developed iSeal, a reliable verification method for the ownership of large language models through external encryption fingerprints.
Iterative Multi-Granular RAG with Contextual Hierarchical Graph
Yanli Hu (National University of Defense Technology), Xiang Zhao (National University of Defense Technology)
GenerationRetrievalGraph Neural NetworkTransformerTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Propose a multi-granularity retrieval-augmented generation framework (MGranRAG) that combines local iterative retrieval with a lightweight context hierarchy graph, leveraging LLM to extract fine-grained evidence and propagate global semantics on the graph.
ITPP: Learning Disentangled Event Dynamics in Marked Temporal Point Processes
Wang-Tao Zhou (University of Electronic Science and Technology of China), Ling Tian (Shenzhen Institute for Advanced Study)
TransformerTime SeriesSequentialElectronic Health RecordsOrdinary Differential Equation
🎯 What it does: Proposed the ITPP model, which utilizes channel-independent encoding and type-aware inverted self-attention to model marked point processes.
JELV: A Judge of Edit-Level Validity for Evaluation and Automated Reference Expansion in Grammatical Error Correction
Yuhao Zhan (Zhejiang University), Fei Wu (Zhejiang University)
ClassificationComputational EfficiencyKnowledge DistillationTransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: Proposed the Judge of Edit-Level Validity (JELV) framework for automated evaluation of the validity of revisions in grammar error correction (GEC), and implemented two judgment implementations based on LLM and DeBERTa;
JoDiffusion: Jointly Diffusing Image with Pixel-Level Annotations for Semantic Segmentation Promotion
Haoyu Wang (Northwestern Polytechnical University), Chen Ding (Northwestern Polytechnical University)
SegmentationGenerationVision Language ModelDiffusion modelAuto EncoderImageText
🎯 What it does: Proposes JoDiffusion, a joint diffusion model capable of simultaneously generating images and corresponding pixel-level segmentation annotations.
Joint Evaluation of Answer and Reasoning Consistency for Hallucination Detection in Large Reasoning Models
Changyue Wang (Tsinghua University), Yiqun Liu (Tsinghua University)
Anomaly DetectionExplainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelTextChain-of-Thought
🎯 What it does: Proposed a black-box hallucination detection framework named RACE, specifically designed for large reasoning models (LRM), which detects factual hallucinations by jointly evaluating the consistency between the model's reasoning trajectory and the final answer.
Joint Implicit and Explicit Language Learning for Pedestrian Attribute Recognition
Yukang Zhang (Xiamen University), Hanzi Wang (Xiamen University)
RecognitionTransformerPrompt EngineeringVision Language ModelContrastive LearningImageText
🎯 What it does: Propose a Joint Implicit and Explicit Language-Guided Reinforcement Learning (JGEL) method that enhances pedestrian attribute recognition performance by converting pedestrian images into personalized language descriptions.
Joint-GCG: Unified Gradient-Based Poisoning Attacks on Retrieval-Augmented Generation Systems
Haowei Wang (State Key Laboratory of Complex System Modeling and Simulation Technology), Qing Wang (State Key Laboratory of Complex System Modeling and Simulation Technology)
RetrievalAdversarial AttackTransformerTextRetrieval-Augmented Generation
🎯 What it does: Propose the Joint-GCG framework to perform poisoning attacks on RAG systems by jointly manipulating gradients of the retriever and generator.
JRDB-Reasoning: A Difficulty-Graded Benchmark for Visual Reasoning in Robotics
Simindokht Jahangard (Monash University), Hamid Rezatofighi (Sharif University of Technology)
Robotic IntelligenceGraph Neural NetworkVision Language ModelVision-Language-Action ModelImageVideoBenchmarkChain-of-Thought
🎯 What it does: Propose the JRDB-Reasoning benchmark and adaptive query engine to enable vision reasoning tasks with adjustable difficulty levels.
Judge Q: Trainable Queries for Optimized Information Retention in KV Cache Eviction
Yijun Liu (Harbin Institute of Technology), Wanxiang Che (Harbin Institute of Technology)
OptimizationComputational EfficiencyLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose Judge Q, which inserts learnable soft tokens during the pre-filling phase and only fine-tunes the embedding layer, allowing these soft tokens to fit the attention maps of real decoding tokens, thereby better capturing global information during KV cache pruning;
JudgeBoard: Benchmarking and Enhancing Small Language Models for Reasoning Evaluation
Zhenyu Bi (Virginia Tech), Xuan Wang (Virginia Tech)
TransformerAgentic AIPrompt EngineeringTextBenchmark
🎯 What it does: Built an evaluation pipeline named JudgeBoard, directly using language models (especially small models) as judges to assess the correctness of candidate answers, and further established task-specific leaderboards through the Elo rating system and multi-metric evaluations (overall accuracy, accuracy of error/correct judgments). Meanwhile, proposed the MAJ (Multi-Agent Judging) framework, enabling multiple configurations of SMLs to collaborate, debate, and vote to enhance judgment quality.
Jump-teaching: Combating Sample Selection Bias via Temporal Disagreement
Kangye Ji (Xidian University), Bohu Huang (Xidian University)
ClassificationData-Centric LearningImage
🎯 What it does: Propose an efficient sample selection framework named Jump-teaching, which eliminates sample selection bias by leveraging temporal differences of a single network across different training iterations, and achieves finer-grained sample screening through single-sample loss splitting.
Jupiter: Enhancing LLM Data Analysis Capabilities via Notebook and Inference-Time Value-Guided Search
Shuocheng Li (Peking University), Dongmei Zhang (Microsoft)
AI Code AssistantTransformerLarge Language ModelSupervised Fine-TuningTextTabular
🎯 What it does: This paper proposes the NbQA dataset and the JUPITER framework, aiming to enhance the performance of large language models (LLMs) in multi-step data analysis tasks.
Just Few States Are Enough: Randomized Sparse Feedback for Stability of Dynamical Systems
Zaid Hadach (Mohammed VI Polytechnic University), Adnane Saoud (Mohammed VI Polytechnic University)
OptimizationPhysics Related
🎯 What it does: Propose using a randomly selected subset of states for feedback in control systems to achieve the system's mean-square asymptotic stability
K-12EduBench: A Benchmark for Evaluating Large Language Models’ Knowledge, Problem-Solving, and Educational Goal Cognition in K-12 Education
Yuqing Ye (Northeast Normal University), Dongdai Zhou (Cornell University)
Large Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: Propose the K-12EduBench benchmark to evaluate three core capabilities of LLMs in K-12 education: knowledge mastery, interdisciplinary problem-solving, and educational goal cognition;
K-ProtoDiff: Key Prototypes-Guided Diffusion for Time Series Generation
Yuhang Duan (Dalian University of Technology), Xiaoshuai Wu (Dalian University of Technology)
GenerationTransformerDiffusion modelTime Series
🎯 What it does: Propose a diffusion model based on key prototypes (K-ProtoDiff) for time series generation, which can maintain the global distribution while significantly preserving local key patterns.
K-STaR: Knowledge-Aware Self-Taught Reasoner
Guozheng Li (Southeast University), Xinyu Zhang (Southeast University)
TransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningPrompt EngineeringTextChain-of-Thought
🎯 What it does: Propose a self-training framework called K-STaR based on knowledge units, which evaluates and filters reasoning paths generated by LLMs using knowledge retrieval and frequency consistency, and further employs high-quality paths for supervised fine-tuning.
KALL-E: Autoregressive Speech Synthesis with Next-Distribution Prediction
Kangxiang Xia (Northwestern Polytechnical University), Lei Xie (Northwestern Polytechnical University)
GenerationData SynthesisTransformerFlow-based ModelAuto EncoderAudio
🎯 What it does: Propose KALL-E, an autoregressive text-to-speech model based on Flow-VAE continuous latent representations, directly predicting the next-frame speech distribution.
KCLNet: Electrically Equivalence-Oriented Graph Representation Learning for Analog Circuits
Peng Xu (Chinese University of Hong Kong), Bei Yu (Chinese University of Hong Kong)
ClassificationObject DetectionRepresentation LearningGraph Neural NetworkContrastive LearningGraphPhysics Related
🎯 What it does: Proposes the KCLNet framework, which utilizes the current equivalence principle based on KCL and asynchronous current-oriented GNN to perform graph representation learning on analog circuits;
KeenKT: Knowledge Mastery-State Disambiguation for Knowledge Tracing
Zhifei Li (Hubei University), Bing Yang (Hubei University)
Representation LearningTransformerDiffusion modelContrastive LearningSequential
🎯 What it does: Propose the KeenKT model, which represents students' knowledge states as Normal-Inverse-Gaussian (NIG) distributions, enabling the capture of fluctuations in learning behaviors during each interaction and achieving distributed representation of states.
Keep On Going: Learning Robust Humanoid Motion Skills via Selective Adversarial Training
Yang Zhang, Yue Gao (Lenovo Research)
Adversarial AttackRobotic IntelligenceReinforcement Learning
🎯 What it does: Proposed a selective adversarial training framework (SA2RT) that enhances the robustness of whole-body motion skills by learning to identify and selectively perturb vulnerabilities in robotic motion policies, validated on the Unitree G1 robot.
KeepKV: Achieving Periodic Lossless KV Cache Compression for Efficient LLM Inference
Yuxuan Tian (Peking University), Tong Yang (ByteDance)
CompressionComputational EfficiencyTransformerLarge Language Model
🎯 What it does: Proposes the KeepKV method, which periodically compresses the KV cache of LLMs without sacrificing generation quality, addressing the attention inconsistency problem caused by traditional merging.
Kernelized Edge Attention: Addressing Semantic Attention Blurring in Temporal Graph Neural Networks
Govind Waghmare (Mastercard), Srikanta Bedathur (Mastercard)
Graph Neural NetworkTransformerGraphBenchmark
🎯 What it does: Proposes KEAT—a novel attention mechanism that modulates edge features through continuous-time kernels to address the semantic ambiguity between node and edge temporal features in Temporal Graph Neural Networks (TGNNs).
Key Decision-Makers in Multi-Agent Debates: Who Holds the Power?
Qian Zhang (Tianjin University), Lanjun Wang (Tianjin University)
Large Language ModelAgentic AITextBenchmark
🎯 What it does: This paper systematically investigates the impact of role allocation strategies in multi-agent debate (MAD) on reasoning task performance, and proposes the MADC consistency ranking strategy based on path consistency without requiring prior knowledge of the truth.
Khan-GCL: Kolmogorov–Arnold Network Based Graph Contrastive Learning with Hard Negatives
Zihu Wang (University of California Santa Barbara), Peng Li (University of California Santa Barbara)
Representation LearningDrug DiscoveryGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Designed a graph contrastive learning framework called Khan-GCL based on the Kolmogorov-Arnold network (KAN), and proposed a method for critical feature identification and hard negative sample generation using KAN coefficients
KineST: A Kinematics-guided Spatiotemporal State Space Model for Human Motion Tracking from Sparse Signals
Shuting Zhao (Fudan University), Xinrong Chen (Fudan University)
Pose EstimationTime Series
🎯 What it does: Proposed a lightweight, kinematic-guided spatiotemporal state space model called KineST for full-body motion tracking from sparse signals of head-mounted displays (HMDs).
KNNDA: A New Perspective of Alignment Recovery for Partially View-Aligned Clustering
Liang Zhao (Dalian University of Technology), Bo Xu (Dalian University of Technology)
Representation LearningAuto EncoderContrastive LearningMultimodality
🎯 What it does: Propose a direct alignment recovery method based on k-nearest neighbors (KNNDA), which simultaneously accomplishes alignment recovery and consistent representation learning in partially viewed multi-view clustering (PVC) without requiring pre-training.
Know Your Neighbors: Subgraph Importance Sampling for Heterophilic Graph Active Learning
Wenjie Yang, Zengfeng Huang (Fudan University)
ClassificationComputational EfficiencyGraph Neural NetworkGraph
🎯 What it does: This paper proposes KyN, an active learning framework specifically designed for heterophilic graphs, aiming to improve the training efficiency of GNNs by selecting nodes and their neighbors for joint labeling.
Knowledge Boundary Discovery for Large Language Models
Ziquan Wang (China University of Petroleum-Beijing), Zhongqi Lu (China University of Petroleum-Beijing)
Explainability and InterpretabilityLarge Language ModelReinforcement LearningText
🎯 What it does: Propose Knowledge Boundary Discovery (KBD), a reinforcement learning-based framework for dynamically exploring the knowledge boundaries of large language models (LLMs) and automatically generating sets of answerable and unanswerable questions.
Knowledge Completes the Vision: A Multimodal Entity-aware Retrieval-Augmented Generation Framework for News Image Captioning
Xiaoxing You (Hangzhou Dianzi University), Jun Yu (Harbin Institute of Technology (Shenzhen))
GenerationRetrievalGraph Neural NetworkTransformerPrompt EngineeringMultimodalityRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Proposed and implemented a multi-modal retrieval-enhanced generation framework named MERGE for news image captioning.
Knowledge Graph Guided Heterogeneity-Informed Diffusion Model for Spatio-Temporal Generation
Zi'ang Wang, Yu Zhao (Beihang University)
GenerationData SynthesisMixture of ExpertsDiffusion modelGraphTime Series
🎯 What it does: This paper proposes a knowledge graph-based heterogeneous-aware diffusion model (KGDiff) for generating urban spatiotemporal data with spatial structure and temporal heterogeneity.
Knowledge-Enhanced Explainable Hypergraph Convolution Network for Medication Recommendation
Zihan Zhang (Peking University), Zhonghai Wu (Peking University)
Recommendation SystemGraph Neural NetworkTabularBiomedical DataElectronic Health Records
🎯 What it does: Construct a hierarchical hypergraph combined with a knowledge graph for drug recommendation, providing explainable and safe predictions
Knowledge-Enhanced Explainable Prompting for Vision-Language Models
Yequan Bie (Hong Kong University of Science and Technology), Hao Chen (Hong Kong University of Science and Technology)
Explainability and InterpretabilityLarge Language ModelPrompt EngineeringVision Language ModelContrastive LearningMultimodalityBiomedical DataRetrieval-Augmented Generation
🎯 What it does: Designed and implemented a framework called KEEP to enhance prompt learning of visual-language models (e.g., CLIP) with fine-grained domain knowledge, providing both visual and textual interpretability.
Knowledge-Enhanced Image Captioning with Adaptive Graph-based Multimodal Alignment and LLM
Guoyi Li (Chinese Academy of Sciences), Honglei Lyu (University of Chinese Academy of Sciences)
GenerationRepresentation LearningGraph Neural NetworkTransformerLarge Language ModelReinforcement LearningVision Language ModelMultimodalityGraphRetrieval-Augmented Generation
🎯 What it does: Enhance the accuracy and semantic consistency of image descriptions through adaptive knowledge graph structures and scene graph knowledge fusion.
Koopman Invariants as Drivers of Emergent Time-Series Clustering in Joint-Embedding Predictive Architectures
Pablo Ruiz-Morales (KU Leuven), Mathias Verbeke (KU Leuven)
Explainability and InterpretabilityRepresentation LearningAuto EncoderTime Series
🎯 What it does: Investigated why Joint-Embedding Predictive Architecture (JEPA) can cluster latent representations of different dynamical modes in time series data through the Koopman operator theory.
KPDM: Key Phrase Dynamic Masking for Robust Text-to-Image Person Retrieval
Shaofeng You (China University of Geosciences), Dapeng Luo (China University of Geosciences)
RetrievalTransformerVision Language ModelContrastive LearningMultimodality
🎯 What it does: This paper proposes a text-to-image person retrieval model based on Key Phrase Dynamic Masking (KPDM), addressing issues caused by traditional random masking, such as semantic relationship destruction, image redundancy interference, and inaccurate noise alignment.
KPLM-STA: Physically-Accurate Shadow Synthesis for Human Relighting via Keypoint-Based Light Modeling
Xinhui Yin (Jilin University), Xiaoli Zhang (Jilin University)
GenerationPose EstimationDiffusion modelGenerative Adversarial NetworkImagePhysics Related
🎯 What it does: Propose a diffraction model based on the Key Point Linear Model (KPLM) and Shadow Triangle Algorithm (STA) to generate physically accurate and geometrically precise shadows after human relighting.
Kronos: A Foundation Model for the Language of Financial Markets
Yu Shi (Tsinghua University), Jian Li (Tsinghua University)
TransformerLarge Language ModelTime SeriesSequentialFinance Related
🎯 What it does: A two-stage pre-training framework named Kronos is constructed specifically for financial K-line sequences.
KSS-MoE: Knowledge Space Synergy Framework in Mixture of Experts for Continual Visual Instruction Tuning
Lingyun Song (Northwestern Polytechnical University), Xuequn Shang (Northwestern Polytechnical University)
Mixture of ExpertsMultimodalityBenchmark
🎯 What it does: Studied a novel knowledge space collaborative framework KSS-MoE, using Mixture of Experts in continuous visual instruction tuning to alleviate catastrophic forgetting.
KTV: Keyframes and Key Tokens Selection for Efficient Training-Free Video LLMs
Baiyang Song (Xiamen University), Jianyuan Guo (Xiamen University)
Computational EfficiencyRepresentation LearningTransformerVision Language ModelContrastive LearningVideoTextMultimodality
🎯 What it does: Designed and implemented the KTV framework, which leverages an untrained vision-language model for video understanding. It first selects keyframes through clustering, then extracts key visual tokens from each frame to reduce spatiotemporal redundancy and improve inference efficiency.
KVmix: Gradient-Based Layer Importance-Aware Mixed-Precision Quantization for KV Cache
Fei Li (Xi'an Jiaotong University), Jinyu Wang (Xi'an Jiaotong University)
Computational EfficiencyTransformerLarge Language ModelTextBenchmark
🎯 What it does: To address the excessive memory consumption of KV cache in large language models, the KVmix method is proposed, which leverages gradient importance analysis to achieve hierarchical mixed-precision quantization, combined with dynamic critical context selection and efficient CUDA kernel fusion, significantly compressing the KV cache and enhancing inference throughput.
L2-LoRA: Improving Low-Rank Adaptation with Layer-Specific Regularization
Xiang Zhang (Peking University), Shikun Zhang (Peking University)
Computational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose L2-LoRA: Introduce layer-wise differentiated L2 regularization into LoRA training, leveraging knowledge localization results to maintain pre-trained knowledge in lower layers while making higher layers more adaptable to downstream tasks, significantly reducing catastrophic forgetting.
L2V-CoT: Cross-Modal Transfer of Chain-of-Thought Reasoning via Latent Intervention
Yu-Liang Zhan (Renmin University of China), Hao Sun (Renmin University of China)
Representation LearningTransformerLarge Language ModelVision Language ModelMultimodalityBenchmarkChain-of-Thought
🎯 What it does: This paper proposes a training-free cross-modal chain-of-thought (CoT) knowledge transfer method called L2V-CoT, which transfers the low-frequency CoT representations of large language models (LLMs) to vision-language models (VLMs) by performing latent interventions on the hidden states of VLMs during inference, thereby enhancing multi-step reasoning capabilities.
La La LiDAR: Large-Scale Layout Generation from LiDAR Data
Youquan Liu, Tongliang Liu (University Of Sydney)
GenerationData SynthesisAutonomous DrivingGraph Neural NetworkDiffusion modelPoint CloudGraph
🎯 What it does: Propose a controllable LiDAR generation framework named La La LiDAR based on scene graphs, achieving adjustable positions and relationships of foreground objects through layout guidance;
LaF-GRPO: In-Situ Navigation Instruction Generation for the Visually Impaired via GRPO with LLM-as-Follower Reward
Yi Zhao (Hong Kong Polytechnic University), Jing Li (Hong Kong Polytechnic University)
Autonomous DrivingReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningVision-Language-Action ModelImageTextMultimodalityBenchmark
🎯 What it does: Proposed the LaF-GRPO framework, which utilizes LLM to simulate responses of visually impaired users as a reward for post-training Vision-Language Models, generating precise, context-aware step-by-step navigation instructions, and constructing the NIG4VI dataset with 27k samples.