AAAI 2026 Papers — Page 13
AAAI Conference on Artificial Intelligence · 4149 papers
EoH-S: Evolution of Heuristic Set Using LLMs for Automated Heuristic Design
Fei Liu (City University of Hong Kong), Mingxuan Yuan (Huawei Noah's Ark Lab)
OptimizationTransformerLarge Language ModelBenchmark
🎯 What it does: This paper proposes the Automated Heuristic Set Design (AHSD) framework and the EoH-S evolutionary algorithm, which utilize large language models to generate and evolve a set of small-scale complementary heuristics, addressing the generalization issues of traditional single heuristic designs.
EPIC: Explanation of Pretrained Image Classification Networks via Prototypes
Piotr Borycki (Jagiellonian University), Łukasz Struski (Jagiellonian University)
ClassificationExplainability and InterpretabilityConvolutional Neural NetworkImage
🎯 What it does: Propose EPIC, a post-hoc prototype explanation framework that provides interpretable prototypes for image classification models without modifying the pre-trained network architecture or requiring retraining
EPO: Diverse and Realistic Protein Ensemble Generation via Energy Preference Optimization
Yuancheng Sun (Chinese Academy of Sciences), Qiwei Ye (Beijing Academy of Artificial Intelligence)
Protein Structure PredictionFlow-based ModelBiomedical DataStochastic Differential Equation
🎯 What it does: Proposed an online refinement framework EPO based on energy preference optimization, enabling pre-trained protein conformation generators to produce diverse energy states that conform to the Boltzmann distribution.
EPSegFZ: Efficient Point Cloud Semantic Segmentation for Few- and Zero-Shot Scenarios with Language Guidance
Jiahui Wang (National University of Singapore), Tong Heng Lee (National University of Singapore)
SegmentationGraph Neural NetworkTransformerVision Language ModelContrastive LearningPoint Cloud
🎯 What it does: This study proposes an efficient point cloud semantic segmentation framework called EPSegFZ, which does not require pre-training and is applicable to few-shot and zero-shot scenarios.
EquaCode: A Multi-Strategy Jailbreak Approach for Large Language Models via Equation Solving and Code Completion
Zhen Liang (Zhejiang Sci-Tech University), Zhengkui Chen (Zhejiang Sci-Tech University)
Adversarial AttackPrompt EngineeringText
🎯 What it does: Proposes a multi-strategy jailbreak method called EquaCode, which first transforms malicious requests into mathematical equation-solving tasks and then embeds the solution into a Python template via code completion, inducing LLMs to generate harmful content with a single query.
Equilibrium-Driven Vertical Federated Learning with Selective Privacy Protection
Yuanzhe Peng (University of Florida), Jie Xu (University of Florida)
Federated LearningSafty and PrivacyImageVideoBiomedical DataElectronic Health RecordsAudio
🎯 What it does: Proposed a balance-based vertical federated learning framework using NashCoder and Shapley values for adaptive decomposition, selectively protecting sensitive attributes of each client while maintaining model accuracy.
Equivariant Atomic and Lattice Modeling Using Geometric Deep Learning for Crystal Structure Optimization
Ziduo Yang (Jinan University), Lei Shen (Jinan University)
OptimizationGraph Neural NetworkGraphPhysics Related
🎯 What it does: This study proposes an end-to-end E3Relax network that directly maps unrelaxed crystal structures to their DFT-relaxed lowest energy structures;
ERank: Fusing Supervised Fine-Tuning and Reinforcement Learning for Effective and Efficient Text Reranking
Yuzheng Cai (Fudan University), Weiguo Zheng (Fudan University)
RetrievalTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningText
🎯 What it does: ERANK is a pointwise text re-ranker based on large language models, employing a two-stage training approach: first using supervised fine-tuning with fine-grained integer scoring, followed by optimization through reinforcement learning with list-wise rewards.
Error Slice Discovery via Manifold Compactness
Han Yu (Tsinghua University), Peng Cui (Tsinghua University)
ClassificationObject DetectionComputational EfficiencyImageTextBiomedical DataBenchmark
🎯 What it does: Propose a method for discovering erroneous slices based on Manifold Compactness, directly using risk and slice coherence as optimization objectives;
ESCA: An Emotional Support Conversation Agent for Enhancing Reasonable Strategy Planning and Effective Expression
Jing Li (Northeastern University), Gang Wu (Northeastern University)
TransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningPrompt EngineeringTextRetrieval-Augmented Generation
🎯 What it does: This paper proposes ESCA, an emotional support chat agent based on a large language model that simultaneously enhances strategy planning and response generation.
Escaping Optimization Stagnation: Taking Steps Beyond Task Arithmetic via Difference Vectors
Jinping Wang (Wenzhou-Kean University), Zhiwu Xie (Wenzhou-Kean University)
ClassificationOptimizationTransformerImage
🎯 What it does: A multi-step optimization framework DV-BASI is studied, which uses differential vectors to continuously explore task arithmetic for improving model editing effectiveness.
Escaping the CAM Shadow: Uncertainty-Guided Reliable Learning for Weakly Supervised Semantic Segmentation
Luyao Chang (University of Electronic Science and Technology of China), Chuan Zhou (University of Electronic Science and Technology of China)
SegmentationTransformerContrastive LearningImage
🎯 What it does: Proposes the Uncertainty-Guided Reliable Learning (UGRL) framework for weakly supervised semantic segmentation, integrating three modules: Prototype-driven Uncertainty Modeling, Loss Modulation, and Reliable Semantic Enhancement, to improve CAM pseudo-labels and decoder learning.
ESMC: MLLM-Based Embedding Selection for Explainable Multiple Clustering
Xinyue Wang, Junhui Hou (Saint Francis University)
Explainability and InterpretabilityLarge Language ModelPrompt EngineeringVision Language ModelMultimodality
🎯 What it does: Propose an interpretable multi-clustering framework (ESMC) based on the hidden states of multimodal large language models, achieving user-specified semantic clustering by extracting feature embeddings from text prompts and combining them with a lightweight MLP clustering head.
Estimating Online Influence Needs Causal Modeling! Counterfactual Analysis of Misinformation Engagement on Social Media
Lin Tian (University of Technology Sydney), Marian-Andrei Rizoiu (University of Technology Sydney)
TransformerTime Series
🎯 What it does: Propose the CITRUS framework, jointly modeling external signals (search trends, news reports, influencer activities) and their interactions with social media content, utilizing causal inference to analyze the true impact of misinformation dissemination.
Estimating the True Distribution of Data Collected with Randomized Response
Carlos Antonio Pinzón (INRIA Saclay), Catuscia Palamidessi (École Normale Supérieure de Lyon)
Safty and PrivacyTabular
🎯 What it does: The paper proposes an analytical formula for maximum likelihood estimation (MLE) based on the randomized response (RR) mechanism and its fast implementation, comparing it with existing iterative estimation methods.
Ev-iCRF: Self-supervised Event-guided iCRF Estimation for HDR Image Reconstruction
Xucheng Guo (Shandong University), Yiran Shen (Shandong University)
RestorationConvolutional Neural NetworkImageMultimodality
🎯 What it does: Propose the Ev-iCRF framework that estimates the inverse camera response function (iCRF) and generates high dynamic range (HDR) images from single-exposure LDR images, using asynchronous event streams generated by an event camera as self-supervised signals.
EvalMuse-40K: A Fine-Grained Benchmark with Comprehensive Human Annotations for Text-to-Image Generation Model Alignment Evaluation
Shuhao Han (Nankai University), Chongyi Li (Nankai University)
TransformerPrompt EngineeringVision Language ModelDiffusion modelMultimodalityBenchmark
🎯 What it does: Constructed the EvalMuse-40K large-scale text-image alignment evaluation dataset with fine-grained human annotations, and proposed an efficient FGA-BLIP2 evaluation method.
Evaluating, Synthesizing, and Enhancing for Customer Support Conversation
Jie Zhu (Soochow University), Fang Kong (Alibaba Cloud Computing)
GenerationData SynthesisLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose the Customer Service Conversation (CSC) task and construct a high-quality evaluation dataset CSConv and a role-play-based synthetic dataset RoleCS; significantly improve the model's strategy consistency and problem-solving ability on CSConv through fine-tuning LLMs on RoleCS.
EvDiff3D: Event-Aware Diffusion Repair for High-Fidelity Event-Based 3D Reconstruction
Kanghao Chen (Hong Kong University of Science and Technology), Zeyu Wang (Hong Kong University of Science and Technology)
GenerationData SynthesisSupervised Fine-TuningDiffusion modelGaussian SplattingImageTime Series
🎯 What it does: Propose EvDiff3D, a two-stage 3D reconstruction framework that uses only event camera data; the first stage generates a rough 3D Gaussian Splatting (3DGS) structure through event loss and event depth constraints, while the second stage employs an event-aware diffusion model to repair rough textures and gradually enhance details and visual quality;
Event-Guided Scene Text Image Super-Resolution
Zihan Qi, Wei Jia (Hefei University Of Technology)
RecognitionSuper ResolutionConvolutional Neural NetworkMultimodality
🎯 What it does: Achieve super-resolution restoration of text images using event camera data, enhancing text readability in low-light and motion-blurred scenarios.
Event-Guided Super-Resolving Blurry Image via Asymmetric Integral Driven Consistency
Chi Zhang (Peng Cheng Laboratory), Wenhan Yang (Peng Cheng Laboratory)
Super ResolutionConvolutional Neural NetworkImageMultimodalityTime Series
🎯 What it does: This paper proposes a self-supervised event-driven blurry image super-resolution framework SE-SRB, achieving high-resolution sharpening by leveraging the spatiotemporal consistency of event streams and RGB blurry images.
Every Little Bit Helps: Exploring Better Utilization of Unlabeled Data for Semi-supervised Singing Melody Extraction Using Multi-bands Diffusion Model
Shuai Yu (Dalian University of Technology), Yi Yu (Donghua University)
RecognitionDiffusion modelAudio
🎯 What it does: Propose the ELH-SME framework for semi-supervised singing melody extraction using unlabeled data, incorporating three modules: Multi-Band Augmentation (DMA), Global-Class Confidence (GCC), and Channel Cross-Attention (CCA);
Evidence-aware Integration and Domain Identification of Spatial Transcriptomics Data
Wei Zhang (Sichuan University), Wei Ju (Nankai University)
Domain AdaptationRepresentation LearningGraph Neural NetworkBiomedical Data
🎯 What it does: Proposes the PREST framework for downstream analyses of spatial transcriptomics data, including domain identification, DEG detection, and expression interpolation.
EvoEmpirBench: Dynamic Spatial Reasoning with Agent-ExpVer
Pukun Zhao (Guangdong University of Finance and Economics), Haojian Huang (Hong Kong University of Science and Technology)
Reinforcement Learning from Human FeedbackLarge Language ModelAgentic AIBenchmark
🎯 What it does: Designed a dynamic interactive spatial reasoning benchmark EEB (local visibility maze navigation and Match-2 elimination) and a three-agent online learning framework Agent-ExpVer.
EvoFMVC: Trusted Federated Multi-View Clustering with Evolutionary Fusion
Li Zhang (Shanxi University), Xinyan Liang (Shanxi University)
Federated LearningSafty and PrivacyRepresentation LearningNeural Architecture SearchContrastive Learning
🎯 What it does: Designed and implemented a trustworthy federated multi-view clustering framework named EvoFMVC, leveraging lightweight trustworthy evidence exchange and evolutionary fusion to achieve privacy protection, low communication cost, and adaptive aggregation.
EvoGrad: Evolutionary-Weighted Gradient and Hessian Learning for Black-Box Optimization
Yedidya Kfir (Bar-Ilan University), Sarit Kraus (Nvidia)
OptimizationAI Code AssistantImageTextBenchmark
🎯 What it does: Proposed a black-box optimization algorithm called EvoGrad2, which combines evolutionary weighted gradient learning with second-order Taylor expansion correction to efficiently search for optimal solutions in high-dimensional non-convex spaces.
EVOKE: Efficient and High-Fidelity EEG-to-Video Reconstruction via Decoupling Implicit Neural Representation
Haodong Jing (Xi'an Jiao Tong University), Yongqiang Ma (Xi'an Jiao Tong University)
GenerationDiffusion modelContrastive LearningBiomedical Data
🎯 What it does: Proposed the EVOKE framework, achieving zero-shot high-fidelity video reconstruction based on EEG.
Evolving Generalist Virtual Agents with Generative and Associative Memory
Zhenkui Zhang (Zhejiang University), Siliang Tang (Zhejiang University)
Graph Neural NetworkLarge Language ModelAgentic AITextBenchmark
🎯 What it does: Proposes the CA3Mem framework, leveraging the memory reorganization and associative retrieval mechanisms of the human hippocampal CA3 region, enabling general virtual agents to achieve continuous learning and self-evolution through generative memory reorganization and propagation activation retrieval.
Evolving Semantic Propagation for Aerial Semantic 3D Gaussian Splatting
Zihan Gao (Xidian University), Shuyuan Yang (Xidian University)
SegmentationVision Language ModelGaussian SplattingImage
🎯 What it does: Propose EvoPropGS, a method that achieves aerial 3D Gaussian Splatting semantic segmentation with only about 2% pixel annotation.
EvoMoE: Expert Evolution in Mixture of Experts for Multimodal Large Language Models
Linglin Jing (Shanghai AI Lab), Qingpei Guo (Ant Group)
Computational EfficiencyRepresentation LearningLarge Language ModelMixture of ExpertsVision Language ModelMultimodality
🎯 What it does: This paper proposes the EvoMoE framework, which performs sparse MoE fine-tuning on multimodal large language models by leveraging Expert Evolution and a Dynamic Token-aware Router.
Exact Algorithms for Distance to Unique Vertex Cover
Foivos Fioravantes (Czech Technical University in Prague), Manolis Vasilakis (Universitè Paris-Dauphine)
Graph
🎯 What it does: This paper studies the problem of finding a 'modulator to unique minimum vertex cover (MU-VC)' by deleting at most k vertices from a given graph such that the remaining graph has a unique minimum vertex cover, and presents its complexity, linear-time tree algorithm, and parameterized algorithms based on treewidth, maximum degree, and cliquewidth.
Exact and Approximate Maximin Share Allocations in Multi-Graphs
George Christodoulou (Aristotle University of Thessaloniki), Symeon Mastrakoulis (Aristotle University of Thessaloniki)
OptimizationGraph
🎯 What it does: Under the multi-graph model, the paper systematically studies maximum minimum share (MMS), symmetric maximum minimum share (PMMS), and 1-out-of-d fair allocation, providing positive existence proofs and negative upper bounds for three types of agents: additive, XOS, and subadditive.
Exact Combinatorial Multi-Class Graph Cuts for Semi-Supervised Learning
Mohammad Mahdi Omati (Sharif University of Technology), Arash Amini (Sharif University of Technology)
ClassificationOptimizationImageGraph
🎯 What it does: Propose a fully discrete multi-class graph semi-supervised learning framework CombCut, achieving precise one-to-one label assignment through stepwise linear allocation;
Exact Optimization for Minimum Dominating Sets
Enqiang Zhu (Guangzhou University), Pu Wu (Hithink RoyalFlush Information Network Co., Ltd.)
OptimizationGraph
🎯 What it does: Developed an exact branch-and-bound algorithm named ParDS to solve the Minimum Dominating Set (MDS) problem, significantly improving solving efficiency through an enhanced linear programming lower bound and multiple instance simplification rules.
Exact Shapley Attributions in Quadratic-time for FANOVA Gaussian Processes
Majid Mohammadi (Vrije Universiteit Amsterdam), Siu Lun Chau (CISPA Helmholtz Center for Information Security)
Explainability and InterpretabilityComputational EfficiencyTabular
🎯 What it does: Proposed a method to compute exact Shapley values for FANOVA Gaussian processes in quadratic time, for local and global explanations.
Exclusion Zones of Instant Runoff Voting
Kiran Tomlinson (Microsoft Research), Jon Kleinberg (Cornell University)
Graph
🎯 What it does: Investigated the exclusion zone properties of Instant-Runoff Voting (IRV) in high-dimensional spaces and graph structures, proving the absence of non-trivial exclusion zones in high-dimensional hyperrectangles while constructing exclusion zones under irregular shapes.
Existence of 2-EFX Allocations of Chores
Jugal Garg (University of Illinois at Urbana-Champaign), Aniket Murhekar (Northwestern University)
Optimization
🎯 What it does: This paper studies the allocation of indivisible chores to agents with additive disutility values and proves that a 2-EFX allocation (i.e., no envy after removing any chore) exists for any instance.
Exo2Ego: Exocentric Knowledge Guided MLLM for Egocentric Video Understanding
Haoyu Zhang (Harbin Institute of Technology (Shenzhen)), Liqiang Nie (Harbin Institute of Technology (Shenzhen))
RecognitionRetrievalDomain AdaptationConvolutional Neural NetworkLarge Language ModelSupervised Fine-TuningVision Language ModelVideoTextMultimodality
🎯 What it does: Constructed a 1.1M aligned front-view and back-view video-text dataset called Ego-ExoClip, and designed a three-stage transfer training process (self-preparation, demonstrator-learner guidance, self-practice) to achieve mapping from third-person (exocentric) to first-person (egocentric) perspectives, thereby enhancing the performance of multi-modal large language models in self-perspective video understanding tasks.
ExoTimer: Leveraging Large Language Models for Time Series Forecasting with Exogenous Variables
Lan Wu (Information Engineering University), LinYu Wang (Information Engineering University)
TransformerLarge Language ModelPrompt EngineeringTime Series
🎯 What it does: Propose the ExoTimer framework, integrating large language models with exogenous variables for time series forecasting.
ExPairT-LLM: Exact Learning for LLM Code Selection by Pairwise Queries
Tom Yuviler (Technion), Dana Drachsler-Cohen (Technion)
Computational EfficiencyAI Code AssistantTransformerLarge Language ModelContrastive LearningTextBenchmark
🎯 What it does: Propose ExPairT-LLM, a precise learning code selection algorithm based on LLM, which identifies the correct implementation among candidate programs by utilizing pairwise membership and equivalence queries.
Expand Your SCOPE: Semantic Cognition over Potential-Based Exploration for Embodied Visual Navigation
Ningnan Wang (Xi'an Jiaotong University), Hongbin Sun (Xi'an Jiaotong University)
Robotic IntelligenceVision Language ModelImageTextBenchmark
🎯 What it does: Proposed the SCOPE framework, achieving zero-shot embedded visual navigation through frontier boundary-driven potential estimation and potential maps;
Expandable and Differentiable Dual Memories with Orthogonal Regularization for Exemplar-free Continual Learning
Hyung-Jun Moon (Yonsei University), Sung-Bae Cho (Yonsei University)
Representation LearningConvolutional Neural NetworkImage
🎯 What it does: Propose a scalable, differentiable dual-memory framework (EDD) for sample-free continual learning, utilizing shared memory and task-specific memory, and maintaining efficient knowledge utilization and stability through orthogonal regularization and memory alignment.
Experiential Fairness: Bridging the Gap Between User Experience and Resource-Centric Fairness in Online LLM Services
Jiahua Huang (South China University of Technology), Weiwei Lin (Pengcheng Laboratory)
OptimizationTransformerLarge Language ModelText
🎯 What it does: Propose an experience fairness concept oriented towards user experience, design and implement the ExFairS scheduling framework, which balances resource utilization and SLO compliance rate, achieving dual optimal scheduling for multi-tenant LLM inference services.
Experimentation on Endogenous Graphs
Wenshuo Wang (Meta), Dominic Coey (Meta)
Graph
🎯 What it does: In this paper, the authors propose a method for unbiased estimation of the total treatment effect in a bipartite graph environment where the intrinsic edges (i.e., edges affected by treatment) change with the intervention.
Expert-Guided Prompting and Retrieval-Augmented Generation for Emergency Medical Service Question Answering
Xueren Ge (University of Virginia), Homa Alemzadeh (University of Virginia)
TransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextBiomedical DataElectronic Health RecordsRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Constructed a multiple-choice question dataset EMSQA containing 24.3K questions, 10 clinical subjects, and 4 certification levels, combined with a 40K-document knowledge base and a 4M-token case library, and proposed two expert-guided LLM frameworks: Expert-CoT and ExpertRAG;
Expert-Inspired Multi-Agent Coordination for Multi-Objective Molecular Optimization
Daojian Zeng, Xiangxiang Zeng (Hunan University)
OptimizationDrug DiscoveryTransformerLarge Language ModelReinforcement LearningAgentic AIMixture of ExpertsBiomedical Data
🎯 What it does: Proposed a multi-objective molecular optimization framework called MAMO based on multi-agent large language models, which simulates expert collaboration cycles for molecular structure modification and property evaluation.
ExpertAD: Enhancing Autonomous Driving Systems with Mixture of Experts
Haowen Jiang (Fudan University), Xin Peng (Fudan University)
Autonomous DrivingMixture of ExpertsImage
🎯 What it does: Propose the ExpertAD framework, improving the perception and planning performance of end-to-end autonomous driving systems through Mixture of Experts (MoE);
Explain with Visual Keypoints Like a Real Mentor! A Benchmark for Multimodal Solution Explanation
Jaewoo Park (Yonsei University), Youngjae Yu (Mathpresso)
Explainability and InterpretabilityTransformerLarge Language ModelPrompt EngineeringVision Language ModelMultimodalityBenchmark
🎯 What it does: Proposed a multimodal problem-solving explanation task and the ME2 benchmark to evaluate models' ability to identify visual key points and generate corresponding explanatory text.
Explainable Depression Assessment from Face Videos by Weakly Supervised Learning
Rongfan Liao (University of Leicester), Siyang Song (Hohai University)
ClassificationExplainability and InterpretabilityTransformerVideoBenchmark
🎯 What it does: Propose a weakly supervised learning method that extracts and selects depression-related facial behaviors using multi-scale time intervals for video-level automatic depression assessment
Explainable Synthetic Image Detection Through Diffusion Timestep Ensembling
Yixin Wu (Fudan University), Xuanjing Huang (Fudan University)
Anomaly DetectionExplainability and InterpretabilityTransformerLarge Language ModelVision Language ModelDiffusion modelImageMultimodality
🎯 What it does: This paper proposes a framework called ESIDE, which detects synthetic images by utilizing intermediate noise images from the DDIM inverse process, and generates and optimizes interpretable defect descriptions through a multimodal large model.
Explaining Decentralized Multi-Agent Reinforcement Learning Policies
Kayla Boggess (University of Virginia), Lu Feng (University of Virginia)
Explainability and InterpretabilityReinforcement LearningBenchmark
🎯 What it does: This paper studies the explanation and summarization of distributed multi-agent reinforcement learning (MARL) strategies, proposing a policy summarization method based on Hasse diagrams and an explanation framework supporting When/Why Not/What queries.
Explaining Temporal Graph Neural Network via Quantum-Inspired Evolutionary Algorithm
Masahiro Mitani (University of Osaka), Yuya Sasaki (University of Osaka)
Explainability and InterpretabilityGraph Neural NetworkGraph
🎯 What it does: Propose a unified explanation framework called QIEA-TGX based on a quantum-inspired evolutionary algorithm (QIEA) to generate factual and counterfactual explanations for Temporal Graph Neural Networks (TGNN), capable of handling both existing and non-existing events.
Explaining Tournament Solutions with Minimal Supports
Clément Contet (University of Toulouse), Jérôme Mengin (University of Toulouse)
OptimizationExplainability and InterpretabilityComputational EfficiencyGraph
🎯 What it does: Proposes the concept of generating a minimal support sub-tournament for the winner in a tournament as evidence for the explainability of various tournament solutions.
Explanation-Preserving Augmentation for Semi-Supervised Graph Representation Learning
Zhuomin Chen (Florida International University), Dongsheng Luo (Florida International University)
Explainability and InterpretabilityRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Propose a semi-supervised data augmentation method called EPA-GRL based on graph explanation, which uses a small number of labeled samples to train a graph interpreter to extract semantic subgraphs, preserving core semantics while perturbing the remaining structure, thereby improving the quality of graph contrastive learning representations.
Explicit Intent-Enhanced Knowledge Distillation for Trip Recommendation
Shuliang Wang (Beijing Institute of Technology), Hanning Yuan (Beijing Institute of Technology)
Recommendation SystemKnowledge DistillationTabularTime SeriesSequential
🎯 What it does: Propose the EKD-Trip framework in travel recommendations, enhancing journey recommendation effectiveness through explicit intent-aware knowledge distillation.
Explicit Modeling of Causal Factors and Confounders for Image Classification
Wei Wu (Shandong University), Xiangxu Meng (Shandong University)
ClassificationVision Language ModelGenerative Adversarial NetworkImageMultimodality
🎯 What it does: This paper proposes the Explicit Modeling Causal Model (EMCM) framework, which explicitly distinguishes and extracts causal factors and confounding factors in image classification to enhance generalization ability.
Explicit Temporal-Semantic Modeling for Dense Video Captioning via Context-Aware Cross-Modal Interaction
Mingda Jia (Chinese Academy of Sciences), Xiaopeng Zhang (Chinese Academy of Sciences)
GenerationTransformerVision Language ModelVideoTextMultimodalityBenchmarkRetrieval-Augmented Generation
🎯 What it does: Propose the CACMI framework, which achieves dense video captioning through explicit temporal-semantic modeling, simultaneously addressing event localization and description.
Exploiting All Mamba Fusion for Efficient RGB-D Tracking
Ge Ying (Zhejiang Normal University), Zhonglong Zheng (Zhejiang Normal University)
Object TrackingComputational EfficiencyMultimodality
🎯 What it does: This paper proposes and implements a fully Mamba-based single-stream RGB-D tracking framework called AMTrack, which enhances the efficiency and robustness of multi-modal tracking through two-stage feature extraction and a low-parameter cross-modal fusion module.
Exploiting Blurry Representations for Event-guided Video Super-Resolution
Zeyu Xiao (National University of Singapore), Xinchao Wang (National University of Singapore)
Super ResolutionRecurrent Neural NetworkMixture of ExpertsVideoMultimodality
🎯 What it does: This paper proposes a blur-aware video super-resolution framework called BluR-EVSR based on event cameras, which can achieve high-quality reconstruction from inputs simultaneously suffering from blur and low resolution.
Exploiting Geometric Structures for Modeling Multi-Agent Behaviors: A New Thinking
Bohao Qu, Qing Guo (Agency for Science Technology and Research A STAR)
Representation LearningReinforcement LearningContrastive LearningSequentialBenchmark
🎯 What it does: Propose HMAR, a hyperbolic multi-agent representation method based on Poincaré balls, to model multi-agent decision-making behaviors.
Exploiting Inter-Session Information with Frequency-enhanced Dual-Path Networks for Sequential Recommendation
Peng He (University of Electronic Science and Technology of China), Qiao Liu (University of Electronic Science and Technology of China)
Recommendation SystemSequential
🎯 What it does: Proposed a frequency-enhanced dual-path network called FreqRec for sequential recommendation, which parallelly integrates a global spectral aggregator and a local spectral refiner based on time-domain self-attention, and jointly trains with frequency-domain consistency loss and cross-entropy.
Exploiting Missing Data Remediation Strategies Using Adversarial Missingness Attacks
Deniz Koyuncu (Rensselaer Polytechnic Institute), Moti Yung (Google LLC)
Adversarial AttackTabularBiomedical Data
🎯 What it does: Proposed a generic two-layer optimization framework to implement adversarial missing attacks under various missing data compensation strategies (complete case analysis, mean imputation, regression imputation, etc.); by learning differentiable missing mechanism parameters, attackers can shift learning models toward their expected targets while introducing only a small amount of missing data;
Exploiting Pre-trained Language Model for Cross-city Urban Flow Prediction Guided by Information-theoretic Analysis
Qiang Zhou (Nanjing University of Aeronautics and Astronautics), Jingjing Gu (Nanjing University of Aeronautics and Astronautics)
Domain AdaptationGraph Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningContrastive LearningTime Series
🎯 What it does: Propose a cross-city traffic flow prediction framework based on pre-trained language models (PLM-CUP), achieving distribution alignment between source domain (city + language) and target domain through constructing a semantic bridge encoder and task-specific adapters.
Exploiting Space Folding by Neural Networks
Michal Lewandowski (Software Competence Center Hagenberg), Bernhard A. Moser (Software Competence Center Hagenberg)
Explainability and InterpretabilityAdversarial AttackImage
🎯 What it does: This paper partitions the input space of neural networks with ReLU and any monotonic activation functions, proposing a generic spatial folding metric χ to quantitatively measure the folding behavior of networks during training, and further investigates its mathematical properties and computational methods.
Exploiting Synergistic Cognitive Biases to Bypass Safety in LLMs
Xikang Yang (Institute of Information Engineering, Chinese Academy of Sciences), Songlin Hu (Institute of Information Engineering, Chinese Academy of Sciences)
Safty and PrivacyAdversarial AttackTransformerSupervised Fine-TuningReinforcement LearningPrompt EngineeringText
🎯 What it does: Proposed and implemented a red team framework called CognitiveAttack based on collaborative cognitive biases, designed to generate adversarial prompts capable of bypassing LLM security mechanisms.
Explore and Establish Synergistic Effects Between Weight Pruning and Coreset Selection in Neural Network Training
Weilin Wan (Fudan University), Cheng Jin (Fudan University)
ClassificationComputational EfficiencyImage
🎯 What it does: Investigate the interaction between weight pruning and core set selection, proposing the SWaST mechanism to achieve synchronous iteration of both, addressing the dual loss problem.
Explore How to Inject Beneficial Noise in MLLMs
Ruishu Zhu (Northwestern Polytechnical University), Hongyuan Zhang (HuaWei Technologies Co., Ltd.)
TransformerMultimodality
🎯 What it does: Proposes a parameter-efficient fine-tuning method called MuNG (Multimodal Noise Generator), which generates and injects 'beneficial noise' while keeping the multimodal large language model (MLLM) frozen;
Explore to Learn: Latent Exploration Through Disentangled Synergy Patterns for Reinforcement Learning in Overactuated Control
Yiming Wang (University of Macau), Leong Hou U (University of Macau)
Robotic IntelligenceSupervised Fine-TuningReinforcement LearningAuto Encoder
🎯 What it does: Developed a two-stage framework ETL: first, latent space-guided exploration discovers muscle synergy patterns, then maps action representations to this low-dimensional synergy space, directly learning policies in this space and fine-tuning the decoder for downstream tasks;
Exploring Category-level Articulated Object Pose Tracking on SE(3) Manifolds
Xianhui Meng (University of Science and Technology of China), Jun Liu (University of Science and Technology of China)
Object TrackingPose EstimationOptimizationRobotic IntelligencePoint Cloud
🎯 What it does: Propose a pose tracking framework called PPF-Tracker based on pairwise features, focusing on 6D pose incremental learning and tracking for category-level articulated objects on the SE(3) manifold.
Exploring Depth Generalization in Large Language Models for Solving Recursive Logic Tasks
Zhiyuan He (University College London)
TransformerLarge Language ModelText
🎯 What it does: Investigates the depth generalization capability of large language models in recursive logic tasks and proposes a cyclic localization-replacement method to alleviate depth degradation issues.
Exploring Domain Generalization and Subpopulation Shift for Generalizable Graph-Level Anomaly Detection
Xiaoxiang Li (Tsinghua University), Xibin Zhao (Tsinghua University)
Anomaly DetectionGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Propose a framework named FGS-GLAD for graph-level anomaly detection under out-of-distribution (OOD) scenarios, which can simultaneously address domain generalization and subpopulation shift problems.
Exploring Efficient Open-Vocabulary Segmentation in the Remote Sensing
Bingyu Li (University of Science and Technology of China), Junyu Gao (Institute of Artificial Intelligence Teleai)
SegmentationDomain AdaptationTransformerVision Language ModelContrastive LearningImageBenchmark
🎯 What it does: Proposed a unified open-vocabulary remote sensing image segmentation benchmark named OVRSISBench, and designed an efficient open-vocabulary segmentation framework called RSKT-Seg, specifically addressing rotation invariance and domain adaptation issues in remote sensing scenes.
Exploring Generalizable Remote Sensing Change Detection via Low-Rank Exchange Adaptation of Vision Foundation Model
Mingwei Zhang (Northwestern Polytechnical University), Qi Wang (Northwestern Polytechnical University)
SegmentationDomain AdaptationTransformerSupervised Fine-TuningContrastive LearningImage
🎯 What it does: Proposed a general remote sensing change detection framework GenCD based on vision foundation models (VFM), focusing on the generalizability of cross-domain change detection.
Exploring High-order-aware Prompt Learning for Zero-shot Anomaly Detection
Shun Wei (Nanjing University of Information Science and Technology), Xiaolong Xu (Nanjing University of Information Science and Technology)
Anomaly DetectionGraph Neural NetworkPrompt EngineeringMixture of ExpertsVision Language ModelImageTextMultimodalityBiomedical Data
🎯 What it does: Proposed a high-order perceptual prompting framework called HiPL, which leverages hypergraphs to dynamically model high-order relationships between image patches and generates diverse text and visual prompts through a Mixture-of-Experts Prompt Learner to enhance zero-shot anomaly detection performance.
Exploring Modality-Aware Fusion and Decoupled Temporal Propagation for Multi-Modal Object Tracking
Shilei Wang (Northwestern Polytechnical University), Gong Cheng (Northwestern Polytechnical University)
Object TrackingTransformerMixture of ExpertsMultimodality
🎯 What it does: Propose the MDTrack framework, addressing the issues of neglected modal differences and mixed temporal information in multi-modal tracking through modal-aware fusion and decoupled temporal propagation, achieving more robust target tracking.
Exploring Non-Convex Discrete Energy Landscapes: An Efficient Langevin-Like Sampler with Replica Exchange
Haoyang Zheng (Purdue University), Guang Lin (Purdue University)
OptimizationComputational EfficiencyImageStochastic Differential Equation
🎯 What it does: In high-dimensional discrete energy scenarios, two samplers, DREXEL and DREAM, combining discrete Langevin sampling with replica exchange, are proposed to better explore non-convex energy landscapes.
Exploring Position Encoding Mechanism in Diffusion U-Net for Training-free High-resolution Image Generation
Feng Zhou (Beijing University of Posts and Telecommunications), Jianqin Yin (Beijing University of Posts and Telecommunications)
GenerationConvolutional Neural NetworkDiffusion modelImageText
🎯 What it does: This paper proposes a training-free Progressive Boundary Complement (PBC) method, which completes the positional information of the diffusion U-Net using virtual boundaries to address the repetition and disorder phenomena in high-resolution image generation.
Exploring Reliable Spatiotemporal Dependencies for Efficient Visual Tracking
Junze Shi (Chinese Academy of Sciences), Haibo Luo (Chinese Academy of Sciences)
Object TrackingComputational EfficiencyConvolutional Neural NetworkTransformerVideo
🎯 What it does: Proposed a lightweight visual tracking framework called STDTrack, which achieves reliable utilization of historical information through dense video sampling during training, introducing spatiotemporal tokens and employing a multi-frame information fusion module (MFIFM) and spatiotemporal token maintainer (STM), while adopting a multi-scale prediction head to enhance adaptability to different target scales.
Exploring Selective Avoidance for Online User Behavior Analysis: A Forest of Thought Explanation
Xiaohua Wu (Wuhan University of Technology), Yuefeng Li (University of Southern Queensland)
Explainability and InterpretabilityLarge Language ModelTextChain-of-Thought
🎯 What it does: This paper proposes the Forest of Thought Explanation (FoTE) prompting framework, which utilizes adaptive Chain of Thought to generate multi-perspective reasoning paths and evaluates and selects high-contribution thoughts through cooperative game theory, ultimately constructing a multi-tree reasoning forest to enhance the interpretability and reliability of LLMs in social questionnaires regarding selective avoidance.
Exploring Surround-View Fisheye Camera 3D Object Detection
Changcai Li (Sun Yat-sen University), Weishi Zheng (Pengcheng Laboratory)
Object DetectionAutonomous DrivingTransformerImage
🎯 What it does: Investigated the use of wide-field fisheye camera systems for end-to-end 3D object detection and proposed a corresponding detection framework and dataset.
Exploring the Potentials of Spiking Neural Networks for Image Deraining
Shuang Chen (Durham University), Amir Atapour-Abarghouei (Durham University)
RestorationSpiking Neural NetworkImage
🎯 What it does: This work proposes a single-image de-raining method based on spiking neural networks (SNN), introducing visual LIF neurons, a spiking decomposition and enhancement module (SDEM), and a multi-scale unit (SMU) to achieve efficient rain removal.
Exploring Transferability of Self-Supervised Learning by Task Conflict Calibration
Huijie Guo (Institute of Software Chinese Academy of Sciences), Wenwen Qiang (Institute of Software Chinese Academy of Sciences)
Meta LearningContrastive LearningImageVideo
🎯 What it does: This paper proposes the Task Conflict Correction (TC2) method, which leverages a meta-learning framework to enhance the transfer performance of self-supervised learning models.
Exposing the Cracks: Vulnerabilities of Retrieval-Augmented LLM-based Machine Translation
Yanming Sun (University of Macau), Derek F. Wong (University of Macau)
GenerationRetrievalTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: Investigate the robustness and reliability of retrieval-augmented large language model (REAL-MT) machine translation under noisy retrieval conditions, proposing a systematic noise synthesis framework and two novel evaluation metrics;
Expressive Power of Graph Transformers via Logic
Veeti Ahvonen (Tampere University), Carsten Lutz (Leipzig University)
Representation LearningGraph Neural NetworkTransformerGraph
🎯 What it does: This paper investigates the expressive power of Graph Transformers (GT) and GPS networks (a variant of Transformers incorporating GNN message passing layers) in processing graph structures through logical methods, analyzing them under both real-number and floating-point numerical implementations.
Expressive Recursive Answers for Ontological Knowledge Bases
Luca Andolfi (Sapienza University of Rome), Maurizio Lenzerini (Sapienza University of Rome)
🎯 What it does: Proposes a new answer language called RRAs for more informative query answering on UCQ over DL-Lite R knowledge bases.
Expressive Temporal Specifications for Reward Monitoring
Omar Adalat (Imperial College London), Francesco Belardinelli (Imperial College London)
Reinforcement LearningTime SeriesSequential
🎯 What it does: This study proposes a method that utilizes quantified linear temporal logic (LTL_f[F]) to construct a reward monitor, which can generate dense numerical rewards for each state trajectory during runtime, thereby alleviating the sparsity problem caused by traditional Boolean rewards.
Extendable Planning via Multiscale Diffusion
Chang Chen (Rutgers University), Sungjin Ahn (KAIST)
Robotic IntelligenceReinforcement LearningDiffusion modelTime SeriesSequentialBenchmark
🎯 What it does: Proposed an scalable long-term planning framework, first extending short trajectories into long trajectories through Progressive Trajectory Extension (PTE), then learning and executing long-term planning at multiple scales via Hierarchical Multiscale Diffuser (HM-Diffuser).
ExtendAttack: Attacking Servers of LRMs via Extending Reasoning
Zhenhao Zhu (Tsinghua University), Jiaheng Zhang (National University of Singapore)
Adversarial AttackPrompt EngineeringText
🎯 What it does: For large language models (LLMs), a stealthy slowdown attack method called ExtendAttack is proposed, which embeds multi-base ASCII encoded characters into input prompts to force the model to perform extensive decoding computations before understanding the prompt, significantly increasing computational overhead and latency while maintaining answer accuracy.
Extending Description Logics with Generic Concepts – the Case of Terminologies
Joshua Hirschbrunn (University of Ulm), Yevgeny Kazakov (University of Ulm)
🎯 What it does: Proposes an extension introducing generic concepts and conditional axioms in description logic (DL), and proves a theoretical framework ensuring decidability in SROIQ and its subclasses.
Extracting Events Like Code: A Multi-Agent Programming Framework for Zero-Shot Event Extraction
Quanjiang Guo (University of Electronic Science and Technology of China), Ke Yan (University of Electronic Science and Technology of China)
Large Language ModelAgentic AIText
🎯 What it does: Propose the Agent-Event-Coder (AEC) framework, transforming zero-shot event extraction into a multi-agent code generation task;
Extracting Interaction-Aware Monosemantic Concepts in Recommender Systems
Dor Arviv, Noam Koenigstein (Tel Aviv University)
Recommendation SystemExplainability and InterpretabilityAuto EncoderTabular
🎯 What it does: Extract interpretable monosemantic neurons from user and item embeddings in recommendation systems, and implement post-hoc control
Extracting Multimodal Learngene in CLIP: Unveiling the Multimodal Generalizable Knowledge
Ruiming Chen (Southeast University), Xin Geng (Southeast University)
ClassificationGenerationRetrievalKnowledge DistillationTransformerContrastive LearningMultimodality
🎯 What it does: Extract and retain multi-modal generalizable knowledge (learngene) from the CLIP model, use this learngene to initialize sub-models of different scales and modalities (visual, language, or combined) via a weighted approach, and perform fine-tuning on tasks such as image classification, cross-modal retrieval, and image captioning.
Extreme Value Monte Carlo Tree Search for Classical Planning
Masataro Asai (MIT-IBM Watson AI Lab), Stephen Wissow (University of New Hampshire)
OptimizationReinforcement LearningTabularBenchmark
🎯 What it does: This paper proposes a new multi-armed bandit algorithm based on extreme value theory, named UCB1-Uniform, and integrates it into MCTS to form GUCT-Uniform, which is used for classical planning problems without domain-specific knowledge.
F.A.C.U.L.: Language-Based Interaction with AI Companions in Gaming
Wenya Wei (Tencent Games), Elvis S. Liu (Tencent Games)
TransformerLarge Language ModelSupervised Fine-TuningVision Language ModelTextMultimodalityRetrieval-Augmented GenerationAudio
🎯 What it does: Designed and implemented the F.A.C.U.L. system to enable real-time AI companion interaction in FPS games based on natural language.
F2RVLM: Boosting Fine-grained Fragment Retrieval for Multi-Modal Long-form Dialogue with Vision Language Model
Hanbo Bi (Pattern Recognition Center, WeChat AI, Tencent Inc), Jinchao Zhang (Pattern Recognition Center, WeChat AI, Tencent Inc)
RetrievalSupervised Fine-TuningReinforcement LearningVision Language ModelMultimodalityBenchmark
🎯 What it does: This paper proposes the Fine-grained Fragment Retrieval (FFR) task, aiming to retrieve short fragments (including sentences and images) semantically consistent with queries in multi-modal long-text dialogues, and designs a specialized retrieval generation model called F2RVLM.
F2SST: Frequency-to-Spatial Semantic Transfer for Few-Shot Image Classification
Xueyi Chen (Soochow University), Fanzhang Li (Soochow University)
ClassificationMeta LearningImage
🎯 What it does: Propose a framework named F2SST that leverages frequency domain information as implicit semantics to enhance few-shot image classification.
FaceShield: Explainable Face Anti-Spoofing with Multimodal Large Language Models
Hongyang Wang (Shijiazhuang Tiedao University), Xiaochun Cao (Sun Yat-sen University)
ClassificationAnomaly DetectionExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Proposed a multi-modal large language model called FaceShield specifically for Face Anti-Spoofing (FAS), capable of performing four tasks: coarse classification, fine classification, reasoning, and attack localization; simultaneously constructed two FAS instruction datasets, FaceShield-pre10K and FaceShield-sft45K.
Facial Dynamics in Video: Instruction Tuning for Improved Facial Expression Perception and Contextual Awareness
Jiaxing Zhao (Tongyi Lab, Alibaba Group), Xihan Wei (Tongyi Lab, Alibaba Group)
RecognitionObject TrackingTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelVideo
🎯 What it does: Propose a dynamic facial expression description task for videos, construct a high-quality annotated dataset with 5,033 entries and over 700,000 words, and perform instruction fine-tuning based on this dataset to develop the FaceTrack-MM model for better tracking and focus on the main character's facial expressions.
Facial-R1: Aligning Reasoning and Recognition for Facial Emotion Analysis
Jiulong Wu (Soochow University), Min Cao (Baidu Inc.)
ClassificationRecognitionData SynthesisLarge Language ModelSupervised Fine-TuningReinforcement LearningVision Language ModelImageMultimodalityBenchmark
🎯 What it does: In the facial emotion analysis task, we propose a three-stage alignment framework called Facial-R1. It first performs few-shot instruction fine-tuning, then uses emotional factors (AU and emotion labels) as verifiable rewards for reinforcement learning, and finally iteratively expands the training set through self-supervised data synthesis, constructing a large-scale FEA-20K dataset.
Facilitating Early Maladaptive Schema–Guided Polite and Empathetic Psychotherapeutic Support: An LLM-Driven MoE-RL-Based Dialogue System
Priyanshu Priya (Indian Institute of Technology Patna), Asif Ekbal (Indian Institute of Technology Patna)
Large Language ModelReinforcement LearningMixture of ExpertsText
🎯 What it does: This paper proposes MATE, a psychotherapy dialogue system based on large language models and Mixture-of-Experts (MoE) reinforcement learning, capable of generating polite and empathetic therapeutic support according to patients' early maladaptive schemas (EMS).