AAAI 2026 Papers — Page 12
AAAI Conference on Artificial Intelligence · 4149 papers
EdgeMTSC: A Lightweight Large-Kernel ConvNet for Multivariate Time Series Classification
Xueyi Zhou (Hanyang University), Dong-Kyu Chae (Hanyang University)
ClassificationConvolutional Neural NetworkTime Series
🎯 What it does: Propose EdgeMTSC, a lightweight ConvNet with large convolutional kernels for multivariate time series classification (MTSC) on edge devices, achieving inter-channel information passing and temporal feature learning through the IMP-KB module.
Editing Is a Bargaining Game: Balanced Knowledge Editing in Large Language Models
Chenghao Xu (Xidian University), Cheng Deng (Xidian University)
OptimizationComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: Propose a balanced knowledge editing framework NaKE based on Nash equilibrium, addressing gradient conflicts and dominance issues during LLM knowledge updates, maintaining existing knowledge while achieving new knowledge insertion and modification.
EduGuardBench: A Holistic Benchmark for Evaluating the Pedagogical Fidelity and Adversarial Safety of LLMs as Simulated Teachers
Yilin Jiang (Zhejiang University of Technology), Xiangjie Kong (Zhejiang University of Technology)
Safty and PrivacyAdversarial AttackLarge Language ModelTextBenchmark
🎯 What it does: Propose the EduGuardBench dual-mode evaluation framework, focusing on the teaching authenticity and safety of teacher-like LLMs;
EEG Agent: A Unified Framework for Automated EEG Analysis Using Large Language Models
Sha Zhao (Zhejiang University), Shijian Li (Zhejiang University)
ClassificationAnomaly DetectionTransformerLarge Language ModelAgentic AIBiomedical DataRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Built EEGAgent, a unified framework based on large language models, for sensing, exploring, detecting EEG, and enabling user interaction with automatic report generation.
EEG-DLite: Dataset Distillation for Efficient Large EEG Model Training
Yuting Tang (Nanyang Technological University), Cuntai Guan
Computational EfficiencyKnowledge DistillationTransformerAuto EncoderContrastive LearningBiomedical Data
🎯 What it does: To address pretraining of large-scale EEG foundation models, this paper proposes a data distillation framework, EEG-DLite, which constructs a concise subset occupying only 5% of the original 2,500-hour EEG dataset through three steps: self-supervised autoencoder compression, anomaly sample removal, and diversity sampling, while maintaining or even exceeding the performance of the full dataset on multiple downstream tasks.
Effective Robotic Cloth Grasping Through Suppressing False Discoveries
Xingyu Zhu (Jilin University), Yixing Gao (Jilin University)
SegmentationRobotic IntelligenceConvolutional Neural NetworkImage
🎯 What it does: Proposed an unsupervised fabric segmentation network and a grasp position estimation method based on wrinkle analysis, achieving robotic grasping and storage in cluttered fabric piles.
Efficient and Adaptive Simultaneous Speech Translation with Fully Unidirectional Architecture
Biao Fu (Xiamen University), Xiaodong Shi (Xiamen University)
Computational EfficiencyRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningTextAudio
🎯 What it does: Proposed an efficient adaptive simultaneous interpretation system EASiST with a fully unidirectional architecture, achieving end-to-end streaming translation through multi-delay data mining, interleaved read/write tagging, lightweight decision heads, and three-stage training.
Efficient and Effective In-context Demonstration Selection with Coreset
Zihua Wang (Southeast University), Yu Zhang (Southeast University)
RetrievalComputational EfficiencyData-Centric LearningVision Language ModelImageTextMultimodality
🎯 What it does: This paper proposes the CoDR framework, which first constructs diverse coresets using cluster-pruning, and then completes global demonstration selection by combining precomputed C-scores with similarity weight between queries and coreset samples.
Efficient and Reliable Hitting-Set Computations for the Implicit Hitting Set Approach
Hannes Ihalainen (University of Helsinki), Matti Järvisalo
OptimizationBenchmark
🎯 What it does: Under the Implicit Hitting Set (IHS) framework, a more efficient, reliable, and provable hitting set solving method is provided for Pseudo-Boolean Optimization (PBO) problems.
Efficient Diffusion Planning with Temporal Diffusion
Jiaming Guo (Institute of Computing Technology, Chinese Academy of Sciences), Ling Li (Institute of Software, Chinese Academy of Sciences)
Robotic IntelligenceConvolutional Neural NetworkReinforcement LearningDiffusion modelTime SeriesSequentialBenchmark
🎯 What it does: Propose Temporal Diffusion Planner (TDP), which dynamically allocates diffusion denoising steps over time to generate an initial triangular plan and then refines the existing plan with a small number of denoising steps in subsequent stages, while introducing an automatic replanning mechanism to prevent plan deviation.
Efficient Few-Step Solution Generation via Discrete Flow Matching for Combinatorial Optimization
Yuanshu Li (Jilin University), You Zhou (Jilin University)
OptimizationGraph Neural NetworkFlow-based ModelGraph
🎯 What it does: Propose an efficient solving method called EFLOCO based on discrete flow matching, which is used to quickly generate high-quality solutions for combinatorial optimization problems (such as TSP, ATSP).
Efficient Hallucination Detection: Adaptive Bayesian Estimation of Semantic Entropy with Guided Semantic Exploration
Qiyao Sun (National University of Defense Technology), Qingyong Hu (Chinese University of Hong Kong)
Anomaly DetectionComputational EfficiencyText
🎯 What it does: Proposes an adaptive semantic entropy estimation method based on a hierarchical Bayesian framework, combined with guided semantic exploration via importance sampling to efficiently detect hallucinatory outputs in large language models.
Efficient LLM-Jailbreaking via Multimodal-LLM Jailbreak
Haoxuan Ji (Xi'an Jiaotong University), Gang Hua (Xidian University)
Adversarial AttackPrompt EngineeringVision Language ModelMultimodality
🎯 What it does: Propose a method that constructs a multimodal large language model (MLLM) and generates text suffixes through image attacks on the MLLM, thereby efficiently achieving jailbreaking of large language models (LLMs).
Efficient Modality Translation via Arbitrary Conditioning and Wasserstein Regularization
Tomas Tokar (Wondeur AI), Scott Sanner (University of Toronto)
GenerationAuto EncoderImageTextMultimodalityAudio
🎯 What it does: This paper proposes a multimodal generative model called MAC-WAE based on arbitrary conditions and Wasserstein regularization, which uses a single shared encoder to handle arbitrary missing modalities, addressing the insufficient expressiveness caused by posterior collapse in traditional multimodal VAEs.
Efficient Multiagent Planning via Shared Action Suggestions
Dylan M. Asmar (Stanford University), Mykel J. Kochenderfer (Stanford University)
Computational EfficiencyReinforcement LearningBenchmark
🎯 What it does: Proposed an MCAS algorithm for multi-agent planning through shared action proposals, utilizing action information to prune the belief space and estimate joint beliefs, approximating centralized decision-making.
Efficient Multimodal Large Language Model via Dynamic KV Cache Quantization
Jiahao Fan (University of Sydney), Chien-Ming Chen (Nanjing University of Information Science and Technology)
Computational EfficiencyTransformerLarge Language ModelMultimodalityBenchmark
🎯 What it does: This paper proposes a dynamic KV cache quantization method for multi-modal large language models, achieving efficient compression of KV caches through channel-level and token-level quantization, online channel/token tracking, automatic scaling, and fine-grained grouping.
Efficient Plug-and-Play Weight Refinement for Sparse Large Models
Jingcheng Xie (University of Science and Technology of China), Zhiwei Xiong (University of Science and Technology of China)
OptimizationComputational EfficiencyTransformerLarge Language ModelText
🎯 What it does: Propose EIWR, a training-free, plug-and-play weight refinement method aimed at efficiently recovering sparse weights generated after one-time pruning.
Efficient Post-Training Refinement of Latent Reasoning in Large Language Models
Xinyuan Wang (Arizona State University), Yanjie Fu (Arizona State University)
Computational EfficiencyReinforcement Learning from Human FeedbackTransformerLarge Language ModelContrastive LearningText
🎯 What it does: Proposed a lightweight post-training framework that dynamically improves LLM's reasoning trajectories in an implicit reasoning space by leveraging contrastive reasoning feedback and residual embedding refinement.
Efficient Protein Optimization via Structure-aware Hamiltonian Dynamics
Jiahao Wang (Shanghai Jiao Tong University), Shuangjia Zheng (Shanghai Jiao Tong University)
OptimizationDrug DiscoveryProtein Structure PredictionTransformerBiomedical Data
🎯 What it does: Propose a structure-aware Bayesian optimization framework called HADES, which samples in continuous sequence space using Hamiltonian dynamics and generates discrete protein sequences through position discretization, thereby efficiently searching for the optimal combination of protein structure and function.
Efficient Reasoning for Large Reasoning Language Models via Certainty-Guided Reflection Suppression
Jiameng Huang (Peking University), Lu Hou (Huawei Technologies Co., Ltd)
Computational EfficiencyLarge Language ModelText
🎯 What it does: Propose a training-agnostic Certainty-Guided Reflection Suppression (CGRS) method that dynamically detects confidence levels of large models during inference and suppresses reflection triggers, thereby reducing redundant reasoning steps and token usage.
Efficient Reinforcement Learning for Zero-Shot Coordination in Evolving Games
Bingyu Hui (Tsinghua University), Jian Wang (Tsinghua University)
Meta LearningReinforcement LearningBenchmark
🎯 What it does: Propose an expandable group training framework ScaPT to address the generalization issues caused by limited group size in zero-shot collaboration (ZSC).
Efficient Rule Induction by Ignoring Pointless Rules
Andrew Cropper (ELLIS Institute Finland), David M. Cerna (Dynatrace Research)
Computational EfficiencyImageTabular
🎯 What it does: This paper proposes an ILP system called REDUCER, which can identify and ignore 'useless rules' (including reducible rules and indifferent rules), thereby performing effective pruning of the hypothesis space during the search process.
Efficient Segmentation with Multimodal Large Language Model via Token Routing
Changsong Wen (Shanghai Jiao Tong University), Wei Shen (Shanghai Jiao Tong University)
SegmentationComputational EfficiencyTransformerLarge Language ModelImageTextMultimodality
🎯 What it does: Propose an efficient segmentation framework based on a multi-modal large language model (MLLM), which significantly reduces computational load while maintaining segmentation performance through dynamic scheduling of image and mask token participation.
Efficient Solution and Learning of Robust Factored MDPs
Yannik Schnitzer (University of Oxford), David Parker (University of Oxford)
OptimizationReinforcement LearningBenchmark
🎯 What it does: Studied solution and learning methods for robust Markov decision processes (r-MDP) with decomposable structures;
Efficient Tensorized Multi-View Anchor Graph Clustering with Affinity Propagation for Remote Sensing Data
Yongshan Zhang (China University of Geosciences), Zhihua Cai (China University of Geosciences)
OptimizationRepresentation LearningImageMultimodalityAgriculture Related
🎯 What it does: Propose an efficient tensor quantized multi-view anchor graph clustering method ETAP, which utilizes superpixel preprocessing, joint learning of anchor graph and compressed anchor graph, tensor Schatten p-norm regularization, and connectivity constraints, and infers pixel clustering from anchors directly via affinity propagation.
Efficient Thought Space Exploration Through Strategic Intervention
Ziheng Li (Peking University), Dawei Yin (Baidu Inc)
Computational EfficiencyTransformerPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: Proposes the Hint-Practice Reasoning (HPR) framework, which strategically explores the idea space during reasoning through a few prompts from a powerful model and efficient execution by a small model.
Efficient Transcoder Adaptation for Fine-Tuned Models: Revealing Medical Reasoning Mechanisms in Large Language Models
Zhouxing Tan (Peking University), Junfei Liu (Peking University)
Explainability and InterpretabilityComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningAuto EncoderBiomedical Data
🎯 What it does: Fine-tune large language models for medical tasks by freezing attention layers and only updating feed-forward network parameters to construct interpretable encoders.
Efficient Verification and Falsification of ReLU Neural Barrier Certificates
Dejin Ren (Chinese Academy of Sciences), Bai Xue (Chinese Academy of Sciences)
OptimizationSafty and PrivacyComputational Efficiency
🎯 What it does: This paper proposes a necessary and sufficient condition for verifying and refuting the forward invariance of ReLU neural network barrier certificates under continuous-time systems.
Efficient, Secure, Differentially Private Deep Learning in the Two-Server Model
Jun Feng (Huazhong University of Science and Technology), Shunli Zhang (Hainan University)
Safty and PrivacyConvolutional Neural NetworkImage
🎯 What it does: Proposes CRYPTDP, an encryption-assisted differential privacy deep learning scheme implemented under a two-server model, which can train differential privacy-enabled deep models without trusting the servers, combining the advantages of both local and centralized differential privacy.
EfficientFlow: Efficient Equivariant Flow Policy Learning for Embodied AI
Jianlei Chang, Xiangyu Xu (Xi'an Jiaotong University)
Computational EfficiencyRobotic IntelligenceConvolutional Neural NetworkReinforcement LearningFlow-based ModelBenchmark
🎯 What it does: Proposed an equivariant visual-motor policy learning framework based on flow matching called EfficientFlow;
EfficientFSL: Enhancing Few-Shot Classification via Query-Only Tuning In Vision Transformers
Wenwen Liao, Xiaofeng Yang (East China University Of Science And Technology)
ClassificationComputational EfficiencyMeta LearningTransformerSupervised Fine-TuningImage
🎯 What it does: Proposed EfficientFSL, a ViT framework that performs query-side fine-tuning with minimal trainable parameters to efficiently accomplish few-shot classification tasks.
Efficiently Computing Compact Formal Explanations
Min Wu (Stanford University), Clark Barrett (Stanford University)
Explainability and InterpretabilityComputational EfficiencyTransformerImageText
🎯 What it does: Propose the VERIX+ framework for efficiently computing optimal formal explanations under ϵ-perturbation constraints, significantly reducing explanation size and generation time.
Efficiently Enhancing Long-term Series Forecasting via Adaptive Lookback with Wavelets
Suxin Tong (Wuhan University of Technology), Jingling Yuan (Wuhan University of Technology)
Computational EfficiencyTransformerTime SeriesBenchmark
🎯 What it does: Proposed a multi-scale adaptive observation window framework (ALW) based on wavelet transform, which can dynamically learn the optimal historical observation length for each instance and each frequency band in time series prediction, and reconstruct new input features by weighting and restoring the selected features.
Efficiently Seeking Flat Minima for Better Generalization in Fine-Tuning Large Language Models and Beyond
Jiaxin Deng (Beijing University Of Technology), Baochang Zhang (Communication University Of China)
OptimizationComputational EfficiencyRepresentation LearningLarge Language ModelSupervised Fine-TuningImageTextMultimodality
🎯 What it does: Introduce the concept of 'flat minima' into the LoRA parameter-efficient fine-tuning framework, designing FMLoRA and its efficient version EFMLoRA to approximate perturbations in the full parameter space within a low-rank subspace, thereby enhancing generalization performance.
EFX Allocation in (Multi)Hypergraphs
Thanasis Lianeas (University of West Attica), Minas Marios Sotiriou (Athens University of Economics and Business)
OptimizationGraph
🎯 What it does: This paper constructively proves the existence of fair allocations satisfying EFX (no envy for any item) under hypergraphs (especially those with girth ≥ 4) and multi-hypergraphs (satisfying specific multiplicity constraints), and provides polynomial-time or pseudo-polynomial-time algorithms.
EFX and PO Allocation Exists for Two Types of Goods
Vladimir Davidiuk (St. Petersburg State University), Danil Sagunov (ITMO University)
OptimizationComputational Efficiency
🎯 What it does: Prove that when there are two types of items and all agents' utilities are positive, it is always possible to construct an allocation satisfying EFX and PO, and present an algorithm that can complete the task in O(n log n + log m) time.
Ego-PMOVE: Prompt-aware Mixture of View Experts Network for Egocentric Gaze Prediction
Heqian Qiu (University of Electronic Science and Technology of China), Hongliang Li (University of Electronic Science and Technology of China)
RecognitionTransformerPrompt EngineeringMixture of ExpertsVision Language ModelContrastive LearningImageVideo
🎯 What it does: This paper proposes a prompt-aware Mixture of Experts (Ego-PMOVE) network to predict gaze points in first-person perspective, fully leveraging cross-perspective (internal and external) information.
EgoCross: Benchmarking Multimodal Large Language Models for Cross-Domain Egocentric Video Question Answering
Yanjun Li (East China Normal University), Xiaoling Wang (Sofia University St Kliment Ohridski)
TransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningPrompt EngineeringVision Language ModelVideoTextMultimodalityBenchmark
🎯 What it does: This paper proposes the EgoCross benchmark, specifically designed to evaluate the generalization ability of multi-modal large language models in cross-domain first-person video question answering (EgocentricQA), covering four domains: surgery, industry, extreme sports, and animal perspectives, with a total of 798 videos and 957 QA pairs.
Eguard: Defending LLM Embeddings Against Inversion Attacks via Text Mutual Information Optimization
Tiantian Liu (Zhejiang University), Kui Ren (Zhejiang University)
OptimizationSafty and PrivacyRepresentation LearningAdversarial AttackTransformerLarge Language ModelAuto EncoderContrastive LearningText
🎯 What it does: Propose the Eguard framework, which projects text embeddings into a secure space by optimizing global and local mutual information to defend against reverse embedding attacks.
EHL*: Memory-Budgeted Indexing for Ultrafast Optimal Euclidean Pathfinding
Jinchun Du (Monash University), Muhammad Aamir Cheema (Monash University)
CompressionOptimizationComputational EfficiencyGraphBenchmark
🎯 What it does: A novel EHL* algorithm is proposed for the Euclidean shortest path problem, capable of building an index under a given memory budget and supporting workload-aware compression using known query distributions.
EHRStruct: A Comprehensive Benchmark Framework for Evaluating Large Language Models on Structured Electronic Health Record Tasks
Xiao Yang, Zhiqi Shen (Nanyang Technological University)
TransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTabularBiomedical DataElectronic Health RecordsBenchmark
🎯 What it does: Proposed the EHRStruct framework, constructed a standardized structured EHR benchmark containing 11 types of clinical tasks and 2,200 samples, systematically evaluated 20 large language models and 11 enhancement methods, and finally proposed the EHRMaster code enhancement approach to achieve optimal performance.
EigenShield: Inference-Time, Model-Agnostic Jailbreaking Defense via Causal Subspace Filtering
Nastaran Darabi (University of Illinois Chicago), Amit Ranjan Trivedi (University of Illinois Chicago)
Safty and PrivacyRepresentation LearningTransformerVision Language ModelTextMultimodality
🎯 What it does: Propose a model-free defense method called EigenShield during inference, which uses random matrix theory (RMT) to filter out causal subspaces in the high-dimensional embeddings of LLMs and VLMs, thereby suppressing adversarial perturbations while preserving semantic structure;
ElastoGen: 4D Generative Elastodynamics
Yutao Feng (Zhejiang University), Yin Yang (Zhejiang University)
GenerationData SynthesisConvolutional Neural NetworkRecurrent Neural NetworkDiffusion modelMeshTime SeriesPhysics Related
🎯 What it does: Propose the ElastoGen model, capable of generating physically accurate 4D elastic dynamics sequences.
ElementaryNet: A Non-Strategic Neural Network for Predicting Human Behavior in Normal-Form Games
Greg d'Eon (University of British Columbia), James R. Wright (University of Alberta)
Explainability and InterpretabilityConvolutional Neural NetworkTabular
🎯 What it does: Proposed a novel neural network structure called ElementaryNet to replace the level-0 model in GameNet that simulates strategic reasoning, and proved that it cannot express any strategic behavior;
Eliciting Chain-of-Thought in Base LLMs via Gradient-Based Representation Optimization
Zijian Wang (University of Sydney), Chang Xu (University of Sydney)
OptimizationRepresentation LearningTransformerTextChain-of-Thought
🎯 What it does: Propose a scheme based on gradient-optimized hidden states, activating chain-of-thought (CoT) reasoning capabilities in Base LLMs through a Bayesian posterior inference framework.
Elite Pattern Reinforcement for Vehicle Routing Problems
Ning Li (Jilin University), Ruichen Tian (Jilin University)
OptimizationTransformerReinforcement LearningTabularBenchmark
🎯 What it does: This paper proposes the Elite-Pattern Reinforcement (EPR) strategy to improve neural network-based construction methods for solving the vehicle routing problem (VRP).
Ellipsoid-Based Decision Boundaries for Open Intent Classification
Yuetian Zou (Tsinghua University), Long Xiao (Hebei University of Science and Technology)
ClassificationTransformerContrastive LearningTextFinance Related
🎯 What it does: To address open-intent classification in dialogue systems, this paper proposes EliDecide, a model based on an elliptical decision boundary, which simultaneously accurately identifies known intents and reliably rejects unknown intents.
ELSPR: Evaluator LLM Training Data Self-Purification on Non-Transitive Preferences via Tournament Graph Reconstruction
Yan Yu, Fuliang Li (Huawei)
Data-Centric LearningLarge Language ModelTextBenchmark
🎯 What it does: Propose the ELSPR framework, which uses a tournament graph to self-purify bidirectional comparison data from LLM evaluations, eliminating non-transitive preferences;
EM-KD: Distilling Efficient Multimodal Large Language Model with Unbalanced Vision Tokens
Ze Feng (Southeast University), Jingdong Wang (Baidu Inc)
Computational EfficiencyKnowledge DistillationRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodality
🎯 What it does: This paper proposes EM-KD, a knowledge distillation framework for efficient multimodal large language models, which enhances the visual reasoning capability of the compressed student model by leveraging the teacher model's rich visual understanding.
EMAformer: Enhancing Transformer Through Embedding Armor for Time Series Forecasting
Zhiwei Zhang (Beijing Jiaotong University), Wenjuan Han (Beijing Jiaotong University)
TransformerTime Series
🎯 What it does: Propose EMAformer, which enhances the performance of multivariate time series forecasting by introducing channel, phase, and joint channel-phase embeddings into a Transformer with variant tokenization.
Embracing Positional Bias in Multiple-Choice Question Answering via Permutation Equivariant Neural Networks
Chengyu Jiao (Southern University of Science and Technology), Yu Zhang (Southern University of Science and Technology)
ClassificationTransformerLarge Language ModelTextBenchmark
🎯 What it does: Proposed the EMBER network, which learns option permutations suitable for LLMs to embrace position bias and achieve permutation equivariance, thereby improving the accuracy of answering multiple-choice questions.
EmbryoDiff: A Conditional Diffusion Framework with Multi-Focal Feature Fusion for Fine-Grained Embryo Developmental Stage Recognition
Yong Sun, Qiang Nie (Hong Kong University of Science and Technology (Guangzhou))
RecognitionConvolutional Neural NetworkTransformerDiffusion modelVideoBiomedical Data
🎯 What it does: Proposed a two-stage conditional diffusion framework called EmbryoDiff, which utilizes multi-focal plane time-stretched videos to achieve fine-grained identification of embryonic development stages.
Emergent Fast-Slow Dynamics in Multi-Agent Q-Learning for Networked Stochastic Games
Yuxin Geng (Beijing University of Posts and Telecommunications), Xingru Chen (Beihang University)
Reinforcement LearningGraphTabularPhysics RelatedStochastic Differential EquationOrdinary Differential Equation
🎯 What it does: Established a unified theoretical framework for describing large-scale networked multi-agent Q-learning, derived the Fokker-Planck equation and master equation through symmetric pair approximation, revealed the time-scale separation between learning and state transitions, and further validated the emergence mechanism of cooperation via simulations.
EMOD: A Unified EEG Emotion Representation Framework Leveraging V-A Guided Contrastive Learning
Yuning Chen (Zhejiang University), Gang Pan (Zhejiang University)
RecognitionComputational EfficiencyRepresentation LearningTransformerContrastive LearningBiomedical Data
🎯 What it does: Propose the EMOD framework, which utilizes the emotional dimensions (Valence-Arousal) to unify labels and structures across different EEG emotion datasets, constructing a cross-dataset emotion representation model;
EMODIS: A Benchmark for Context-Dependent Emoji Disambiguation in Large Language Models
Jiacheng Huang (Hubei Normal University), Xiaoyin Yi (Chongqing University of Posts and Telecommunications)
TransformerLarge Language ModelTextBenchmark
🎯 What it does: Propose the EMODIS benchmark to evaluate the ambiguity resolution capability of large language models in contexts containing emojis.
Emotion and Intention Guided Multi-Modal Learning for Sticker Response Selection
Yuxuan Hu (City University of Hong Kong), Xiangyu Zhao (City University of Hong Kong)
Recommendation SystemTransformerContrastive LearningTextMultimodality
🎯 What it does: Propose a joint emotion and intent guided multimodal learning framework EIGML for the sticker response selection task.
Emotion-Coherent Reasoning for Multimodal LLMs via Emotional Rationale Verifier
Hyeongseop Rha (KAIST), Yong Man Ro (KAIST)
Explainability and InterpretabilityTransformerLarge Language ModelReinforcement LearningVision Language ModelMultimodality
🎯 What it does: Guide multimodal large language models to generate natural language explanations consistent with target emotions through an Emotion Reasonableness Validator (ERV) and reward mechanism.
Emotion-Conditioned Motion Sub-spaces with Flow Matching for Real-Time Audio-Driven Talking Heads
Haoyu Wang (Tsinghua University), Jia Jia (Alibaba Group)
Image TranslationGenerationTransformerDiffusion modelFlow-based ModelVideoMultimodalityAudio
🎯 What it does: This paper proposes a real-time audio-driven speaker facial animation framework that combines emotion-conditioned motion subspace with flow matching.
EmoVid: A Multimodal Emotion Video Dataset for Emotion-Centric Video Understanding and Generation
Zongyang Qiu (Hong Kong University of Science and Technology), Zeyu Wang (Hong Kong University of Science and Technology)
ClassificationGenerationTransformerLarge Language ModelSupervised Fine-TuningVision Language ModelVideoTextMultimodalityBenchmarkAudio
🎯 What it does: Proposed the EmoVid multimodal emotional video dataset for emotion-centered video understanding and generation.
Empowering DINO Representations for Underwater Instance Segmentation via Aligner and Prompter
Zhiyang Chen (Shandong University), Runmin Cong (Shandong University)
SegmentationDomain AdaptationRepresentation LearningTransformerPrompt EngineeringImage
🎯 What it does: Proposed the DiveSeg framework based on DINOv2, achieving high-precision prediction for underwater instance segmentation through AquaStyle Aligner and ObjectPrior Prompter.
Empowering Semantic-Sensitive Underwater Image Enhancement with VLM
Guodong Fan (Shandong Technology and Business University), Jinjiang Li (Shandong Technology and Business University)
RestorationTransformerVision Language ModelAuto EncoderMultimodality
🎯 What it does: Proposes a semantic-sensitive underwater image enhancement strategy based on a vision-language model, which generates text descriptions and maps them to spatial semantic guidance maps. It improves the UIE network by integrating cross-attention and semantic alignment loss to achieve content-aware enhancement focused on key objects.
Empowering Sparse-Input Neural Radiance Fields with Dual-Level Semantic Guidance from Dense Novel Views
Yingji Zhong (Hong Kong University of Science and Technology), Dan Xu (Hong Kong University of Science and Technology)
Knowledge DistillationRepresentation LearningNeural Radiance FieldGaussian SplattingImagePoint Cloud
🎯 What it does: A self-supervised framework utilizes dense semantic maps rendered by a NeRF teacher model from sparse inputs as auxiliary data to train a student NeRF model, enhancing the quality of sparse view synthesis.
Encode Geometric Diagram as Geo-Graph in Geometry Problem Solving
Wenjun Wu (Xi'an Jiaotong University), Yaqiang Wu (Lenovo Research)
Graph Neural NetworkTransformerLarge Language ModelTextGraph
🎯 What it does: Proposed the Geo-Graph Geometry Problem Solving (G³PS) model, which converts geometric diagrams into geo-graphs and encodes them using a multi-attention graph Transformer. It then fuses the graph features with text features to generate problem-solving steps and answers.
ENCORE: Entropy-guided Reward Composition for Multi-head Safety Reward Models
Xiaomin Li (Harvard University), Mingye Gao (Massachusetts Institute of Technology)
Reinforcement Learning from Human FeedbackReinforcement LearningTextBenchmark
🎯 What it does: Investigated the relationship between safety rule entropy and human preference accuracy in multi-head reward models, and proposed the ENCORE method that aggregates rules via entropy penalty weights to construct more reliable safety reward models.
End-to-End Contrastive Language-Speech Pretraining Model for Long-Form Spoken Question Answering
Jiliang Hu (Wuhan University), Ping Wang (Xiaomi)
RetrievalTransformerContrastive LearningTextMultimodalityAudio
🎯 What it does: Propose an end-to-end contrastive language-voice retrieval model, CLSR, for retrieving the most relevant segments in long audio to support long-text question answering.
End-to-End Knowledge Distillation for Unsupervised Domain Adaptation with Large Vision-language Models
Yangtao Wang (Guangzhou University), Meie Fang (Guangzhou University)
Domain AdaptationKnowledge DistillationTransformerPrompt EngineeringVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Leveraging a large vision-language model (CLIP) through lightweight prompt learning and teacher-student alternating training to achieve end-to-end knowledge distillation, addressing the domain gap in unsupervised domain adaptation (UDA).
End-to-End Multi-Person Pose Estimation with Pose-Aware Video Transformer
Yonghui Yu (Zhejiang Gongshang University), Wenwu Yang (Zhejiang Gongshang University)
Pose EstimationTransformerVideo
🎯 What it does: Propose a full end-to-end multi-person human pose estimation framework called PAVE-Net, eliminating heuristic steps such as detection, RoI cropping, and NMS; achieve precise estimation of multi-person poses in videos through a spatial encoder, pose-aware spatiotemporal decoder, and joint decoder.
EndoIR: Degradation-Agnostic All-in-One Endoscopic Image Restoration via Noise-Aware Routing Diffusion
Tong Chen (University of Sydney), Luping Zhou (University of Sydney)
RestorationConvolutional Neural NetworkDiffusion modelBiomedical Data
🎯 What it does: Developed a full-featured, degradation-agnostic endoscopic image restoration framework called EndoIR
Endowing Vision-Language Models with System 2 Thinking for Fine-grained Visual Recognition
Yutong Yang (Sichuan University), Mouxing Yang (Sichuan University)
RecognitionLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: Propose the SCAN framework, introducing System 2 reasoning into VLM under zero-shot settings to achieve fine-grained visual recognition.
Energy-based Autoregressive Generation for Neural Population Dynamics
Ningling Ge (Institute of Automation, Chinese Academy of Sciences), Shan Yu (Institute of Automation, Chinese Academy of Sciences)
GenerationTransformerDiffusion modelScore-based ModelAuto EncoderTime SeriesBiomedical Data
🎯 What it does: Proposed an energy-based autoregressive generation framework (EAG) for efficiently generating temporal samples of neural population dynamics in latent space, and achieving conditional generation to generalize to unseen behavioral contexts.
Energy-guided Dual Domain-invariant Prompting Framework with Fourier Regularization for Generalized Few-Shot Medical Segmentation
Shaolei Liu (Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences), Jiamao Li (University of Chinese Academy of Sciences)
SegmentationTransformerPrompt EngineeringVision Language ModelBiomedical DataMagnetic Resonance Imaging
🎯 What it does: Proposed an energy-guided dual-domain invariant prompt optimization framework that achieves cross-domain medical image segmentation with limited annotations.
Enhanced Federated Deep Multi-View Clustering Under Uncertainty Scenario
Bingjun Wei (Southwestern University of Finance and Economics), Xin Yang (Southwestern University of Finance and Economics)
Federated LearningRepresentation LearningAuto EncoderContrastive LearningImageTextMultimodalityTabular
🎯 What it does: Propose the EFDMVC framework in federated multi-view clustering to address issues of incomplete client perspectives and aggregation uncertainty.
Enhanced Privacy Leakage from Noise-Perturbed Gradients via Gradient-Guided Conditional Diffusion Models
Jiayang Meng (Renmin University of China), Guolong Zheng (Minjiang University)
Safty and PrivacyAdversarial AttackDiffusion modelImage
🎯 What it does: Proposes a gradient-guided conditional diffusion model (GG-CDM) to achieve gradient inversion attacks, capable of recovering high-resolution images from noisy perturbed gradients.
Enhancing All-to-X Backdoor Attacks with Optimized Target Class Mapping
Lei Wang (Nanjing University of Aeronautics and Astronautics), Fengyuan Xu (Nanjing University)
Adversarial AttackConvolutional Neural NetworkImage
🎯 What it does: Propose a new attack strategy for All-to-X backdoor attacks, significantly improving attack success rates by optimizing source class grouping and target class mapping, while maintaining robustness against existing defenses.
Enhancing Chemical Explainability Through Counterfactual Masking
Łukasz Janisiów (Jagiellonian University), Tomasz Danel (Jagiellonian University)
Explainability and InterpretabilityDrug DiscoveryGraph Neural NetworkDiffusion modelGraph
🎯 What it does: Propose a counterfactual masking method based on generative models, which enhances the interpretability and evaluation metrics of molecular property prediction models by replacing masked substructures with chemically reasonable fragments;
Enhancing Control Policy Smoothness by Aligning Actions with Predictions from Preceding States
Kyoleen Kwak (Kyung Hee University), Hyoseok Hwang (Kyung Hee University)
OptimizationReinforcement LearningTime SeriesSequentialBenchmark
🎯 What it does: Proposed the ASAP method, achieving action smoothing without structural loss by predicting actions using the transition distribution of the previous state and incorporating a second-order difference penalty.
Enhancing Conversational Recommender Systems with Tree-Structured Knowledge and Pretrained Language Models
Yongwen Ren (University of Science and Technology of China), Hui Xiong (Nanjing University of Aeronautics and Astronautics)
Recommendation SystemGraph Neural NetworkTransformerPrompt EngineeringContrastive LearningTextGraphRetrieval-Augmented Generation
🎯 What it does: Propose a dialogue recommendation framework PCRS-TKA that integrates pre-trained language models (PLMs) with knowledge graphs, utilizing retrieval-augmented generation to construct dialogue-specific tree-structured knowledge and serialize it, while combining user multi-round collaborative preferences and semantic alignment to achieve fine-grained reasoning and response generation.
Enhancing Diffusion Policies with Distribution-Matching Generator in Offline Reinforcement Learning
Xuemin Hu (Hubei University), Long Chen (Hubei University)
Reinforcement LearningDiffusion modelGenerative Adversarial NetworkTabularBenchmark
🎯 What it does: Proposed the DMGDP algorithm, a diffusion strategy offline reinforcement learning method based on a distribution-matching generator, which aims to simultaneously maximize the expected return and deceive the discriminator to address out-of-distribution (OOD) problems.
Enhancing DPSGD via Per-Sample Momentum and Low-Pass Filtering
Xincheng Xu (Australian National University), David Smith (CSIRO)
OptimizationConvolutional Neural NetworkTransformerImageText
🎯 What it does: Propose a new variant of differential privacy stochastic gradient descent (DP-SGD) called DP-PMLF, which combines per-sample momentum smoothing and low-pass filtering to simultaneously mitigate DP noise and clipping bias.
Enhancing Exploration and Exploitation in Hierarchical Reinforcement Learning with Subgoal Graph Learning
Yibo Zhang (Institute of Automation, Chinese Academy of Sciences), Dengpeng Xing (Institute of Automation, Chinese Academy of Sciences)
OptimizationGraph Neural NetworkReinforcement LearningContrastive LearningGraph
🎯 What it does: Propose the ASI framework, which enhances the exploration and exploitation efficiency of GCHRL by constructing a subgoal potential graph to achieve active boundary exploration and hierarchical self-imitation.
Enhancing Generalization of Depth Estimation Foundation Model via Weakly-Supervised Adaptation with Regularization
Yan Huang (South China University Of Technology), Xun Xu (Institute For Infocomm Research)
Depth EstimationDomain AdaptationConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: Propose a source-free adaptation framework named WeSTAR, which efficiently and robustly fine-tunes pre-trained monocular depth estimation base models to enhance generalization performance on unseen domains and damaged data by leveraging dense self-training, weakly supervised (point-to-rank constraint), and low-rank incremental regularization.
Enhancing Interpretability for Vision Models via Shapley Value Optimization
Kanglong Fan (City University of Hong Kong), Chen Ma (City University of Hong Kong)
ClassificationSegmentationExplainability and InterpretabilityConvolutional Neural NetworkTransformerImageMultimodality
🎯 What it does: Propose a self-explaining framework that integrates Shapley value estimation as an auxiliary task into visual model training.
Enhancing Kernel Power $K$-means: Scalable and Robust Clustering with Random Fourier Features and Possibilistic Method
Yixi Chen (National University of Defense Technology), Xinwang Liu (National University of Defense Technology)
OptimizationComputational EfficiencyImageMultimodality
🎯 What it does: Propose a kernel power k-means based on random Fourier features (RFF-KPKM) and its improved multi-kernel multi-view version (IP-RFF-MKPKM) to address large-scale data and noise robustness issues.
Enhancing Logical Expressiveness in Graph Neural Networks via Path-Neighbor Aggregation
Han Yu, Zhichao Peng (National University Of Defense Technology)
Computational EfficiencyRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: Proposed Path-Neighbor Enhanced GNN (PN-GNN), which enhances the logical expressiveness of knowledge graph reasoning by aggregating neighbor nodes along reasoning paths.
Enhancing Medical Large Vision-Language Models via Alignment Distillation
Aofei Chang (Pennsylvania State University), Fenglong Ma (Pennsylvania State University)
Explainability and InterpretabilityKnowledge DistillationSupervised Fine-TuningVision Language ModelBiomedical Data
🎯 What it does: Proposes the MEDALIGN framework, which enhances the accuracy and interpretability of image understanding and text generation by transferring visual representations and attention knowledge from a professional CLIP model to a medical large vision-language model (Med-LVLM) through alignment distillation.
Enhancing Meme Emotion Understanding with Multi-Level Modality Enhancement and Dual-Stage Modal Fusion
Yi Shi (Zhejiang University), Wenzhi Chen (Zhejiang University)
ClassificationRecognitionTransformerLarge Language ModelVision Language ModelMultimodalityChain-of-Thought
🎯 What it does: Propose the MemoDetector framework, which utilizes a multi-modal large language model (MLLM) for four-step text enhancement and designs a two-stage cross-modal fusion strategy to achieve more precise identification of meme sentiment.
Enhancing Multimodal Misinformation Detection by Replaying the Whole Story from Image Modality Perspective
Bing Wang (Jilin University), Shengsheng Wang (Jilin University)
ClassificationGenerationGraph Neural NetworkSupervised Fine-TuningDiffusion modelImageTextMultimodality
🎯 What it does: In the multimodal misinformation detection task, this paper proposes the RETSIMD method: first, the text is divided into several paragraphs, and a pre-trained text-image generator is used to generate a supplementary image for each paragraph; then, a graph structure incorporating the original and generated images is constructed, and graph neural networks are utilized to fuse features from multiple images; finally, the fused image features are combined with text features for authenticity judgment.
Enhancing Noise Resilience in Face Clustering via Sparse Differential Transformer
Dafeng Zhang (Samsung Research and Development Institute), Shizhuo Liu (Samsung Research and Development Institute)
RecognitionTransformerMixture of ExpertsImage
🎯 What it does: Propose a facial clustering framework based on a sparse differential transformer (SDT), enhancing noise robustness and neighbor discovery accuracy by predicting the optimal number of neighbors and improving the Top-K Jaccard similarity
Enhancing PIBT via Multi-Action Operations
Egor Yukhnevich (HSE University), Anton Andreychuk (Cognitive AI Systems Lab)
OptimizationGraphBenchmark
🎯 What it does: Propose an enhanced version of PIBT (EPIBT) to achieve fast online multi-agent path planning in LMAPF-T scenarios;
Enhancing Pre-training Data Detection in LLMs Through Discriminative and Symmetric Prefix Selection
Kai Sun (Xi'an Jiaotong University), Bin Shi (Xi'an Jiaotong-Liverpool University)
Data-Centric LearningTransformerLarge Language ModelText
🎯 What it does: This paper proposes an unlabeled prefix selection framework called DSC-Prefix to enhance the accuracy and robustness of LLM pre-training data detection methods RECALL and CON-RECALL.
Enhancing Retrieval-Augmented Large Vision Language Models via Knowledge Conflict Mitigation
Wenbin An (Xi'an Jiaotong University), Shijian Lu (Zhejiang University of Technology)
TransformerVision Language ModelMultimodalityRetrieval-Augmented Generation
🎯 What it does: This paper proposes a training-agnostic, plug-and-play framework called KCM, aiming to reduce knowledge conflicts in retrieval-augmented vision-language models and improve generation quality by modulating attention allocation, pruning parameterized neurons, and reinforcing context logits.
Enhancing Rotation-Invariant 3D Learning with Global Pose Awareness and Attention Mechanisms
Jiaxun Guo (Concordia University), Wentao Fan (Beijing Normal-Hong Kong Baptist University)
ClassificationSegmentationPose EstimationGraph Neural NetworkPoint Cloud
🎯 What it does: This paper proposes Shadow-informed Pose Feature (SiPF) and Rotation-Invariant Attention Convolution (RIAttnConv), enhancing rotation-invariant learning in 3D point clouds through global pose reference and attention mechanisms.
Enhancing Spatial Reasoning Through Visual and Textual Thinking
Xun Liang (Zhejiang University), Jieping Ye (Alibaba Cloud Computing)
TransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringVision Language ModelImageTextMultimodalityBenchmarkChain-of-Thought
🎯 What it does: Propose the SpatialVTS method, combining visual reasoning and textual reasoning stages to enhance the performance of Vision-Language Models (VLMs) in 2D/3D spatial reasoning tasks.
Enhancing Stability and Fidelity for Zero-Shot TTS with a Multi-Level Evaluator
Hualei Wang, Dong Yu (Tencent AI Lab)
GenerationData-Centric LearningRecurrent Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextAudio
🎯 What it does: Proposed the Vox-Evaluator multi-layer evaluator for automatically locating and correcting pronunciation errors and sound quality issues in zero-shot TTS synthesized speech, achieving refined preference alignment and significantly enhancing speech stability and naturalness.
Enhancing Strategy Logic with Procedural Rationality
Ruiqi Jin (Sun Yat-sen University), Yongmei Liu (Sun Yat-sen University)
🎯 What it does: This paper proposes an extension of Strategy Logic (SL) called SL IEDS, which adds two operators—Elimination of Dominated Strategies (EDS) and Iterated Elimination of Dominated Strategies (IEDS)—to the original SL, along with their semantic definitions.
Enhancing the Knowledge Tracing via a Plug-In Guided Diffusion Model
Shuaishuai Zu (Renmin University of China), Biao Qin (Renmin University of China)
TransformerDiffusion modelSequential
🎯 What it does: By introducing a plug-and-play guided diffusion module GOOD, the prediction performance of existing knowledge tracing models is improved and enhanced.
Enhancing Uncertainty Estimation in LLMs with Expectation of Aggregated Internal Belief
Zeguan Xiao (Shanghai University of Finance and Economics), Guanhua Chen (Southern University of Science and Technology)
Explainability and InterpretabilityReinforcement Learning from Human FeedbackTransformerLarge Language ModelText
🎯 What it does: Proposes EAGLE, a no-training calibration method that leverages LLM hidden layer information for self-assessment to more accurately estimate the confidence of model responses.
ENHash: Error Notebook-Guided Fine-Grained Learning for Unsupervised Cross-Modal Hashing
Hao Fu (Yanshan University), Guanghua Gu (Yanshan University)
RetrievalTransformerContrastive LearningMultimodality
🎯 What it does: Proposed an unsupervised fine-grained cross-modal hashing framework called ENHash based on human error notes, achieving unsupervised fine-grained retrieval through incremental clustering, error logging, and multi-cluster boundary loss.
Enumerating Minimal Unsatisfiable Cores of LTLf Formulae
Antonio Ielo (University of Calabria), Francesco Ricca (University of Calabria)
Benchmark
🎯 What it does: Studied a method that utilizes Answer Set Programming (ASP) to enumerate minimal unsatisfiable cores (MUC) of finite trace linear temporal logic (LTLf) formulas.
EnViT: Enhancing the Performance of Early-Exit Vision Transformers via Exit-Aware Structured Dropout-Enabled Self-Distillation
Yonghao Dong (National Engineering Research Center for Big Data Technology and System), Yun Yang (Swinburne University of Technology)
Computational EfficiencyKnowledge DistillationTransformerImage
🎯 What it does: Proposed a training framework called EnViT, which combines exit-aware structured dropout with self-distillation to improve the early exit accuracy of Vision Transformers while maintaining the performance of later exits.