AAAI 2025 Papers — Page 2
AAAI Conference on Artificial Intelligence · 3028 papers
Adaptive Calibration: A Unified Conversion Framework of Spiking Neural Networks
Ziqing Wang (Northwestern University), Renjing Xu (Hong Kong University of Science and Technology)
Object DetectionSegmentationComputational EfficiencyNeural Architecture SearchSpiking Neural NetworkImage
🎯 What it does: A training-free ANN-to-SNN conversion framework is proposed, combining Adaptive Firing Neurons (AdaFire), Layer Sensitivity Threshold Compression (SSC), and Input-Aware Adaptive Time Steps (IAT), significantly improving the accuracy and energy efficiency of the converted SNN.
Adaptive Computation Modules: Granular Conditional Computation for Efficient Inference
Bartosz Wójcik (Jagiellonian University), Simone Scardapane (Sapienza University of Rome)
Computational EfficiencyKnowledge DistillationTransformerSupervised Fine-TuningMixture of ExpertsImageAudio
🎯 What it does: This paper proposes Adaptive Computation Modules (ACM) for the Transformer model, enabling dynamic adjustment of computation width for each token as needed, thereby reducing inference costs.
Adaptive Dataset Quantization
Muquan Li (University of Electronic Science and Technology of China), Ke Qin (University of Electronic Science and Technology of China)
CompressionData-Centric LearningContrastive LearningImage
🎯 What it does: Proposes Adaptive Dataset Quantization (ADQ), which compresses datasets using an adaptive sampling method by evaluating the representativeness, richness, and importance of each bin based on Dataset Quantization.
Adaptive Decision Boundary for Few-Shot Class-Incremental Learning
Linhao Li (Hebei University of Technology), Liang Yang (Hebei University of Technology)
ClassificationConvolutional Neural NetworkImage
🎯 What it does: A pluggable adaptive decision boundary strategy (ADBS) is proposed for few-shot incremental learning (FSCIL) to dynamically allocate decision boundaries for each category and further enhance inter-class separability through inter-class constraint (IC).
Adaptive Draft-Verification for Efficient Large Language Model Decoding
Xukun Liu (Northwestern University), Dongkuan (DK) Xu (North Carolina State University)
GenerationComputational EfficiencyTransformerLarge Language ModelReinforcement LearningText
🎯 What it does: This paper presents Adaptix, an adaptive draft-validation scheme that does not require fine-tuning. It utilizes a triplet matrix to approximate the output distribution of LLMs and generates drafts that are highly consistent with the LLM distribution through MCTS, significantly accelerating LLM decoding.
Adaptive Dual Guidance Knowledge Distillation
Tong Li (Xi'an University of Technology), Kai Lu (Xi'an University of Technology)
ClassificationObject DetectionKnowledge DistillationImage
🎯 What it does: The Adaptive Dual Guidance Knowledge Distillation (ADG-KD) method is proposed, which retains the guidance of the pre-trained teacher during the knowledge distillation process and constructs a Bidirectional Optimization Route (BOR) for the teacher, allowing the student to gradually accept knowledge that matches their representation capability, from easy to difficult.
Adaptive Dual-domain Learning for Underwater Image Enhancement
Lintao Peng (Beijing Institute of Technology), Liheng Bian (Beijing Institute of Technology)
RestorationImage
🎯 What it does: This paper proposes an underwater image enhancement network SS-UIE based on spatial-spectral dual-domain adaptive learning.
Adaptive Few-shot Prompting for Machine Translation with Pre-trained Language Models
Lei Tang (Guangdong University of Technology), Zhijing Yang (The Chinese University of Hong Kong)
TransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Proposes the Adaptive Few-shot Prompting (AFSP) framework, which automatically selects the most suitable translation examples for large language models (LLMs) and uses a self-supervised re-ranker to choose the best translation after generating multiple candidates, in order to improve machine translation quality.
Adaptive Guidance: Training-free Acceleration of Conditional Diffusion Models
Angela Castillo (Universidad de los Andes), Ali Thabet (King Abdullah University of Science and Technology)
OptimizationComputational EfficiencyKnowledge DistillationNeural Architecture SearchDiffusion modelImage
🎯 What it does: Proposes Adaptive Guidance technology to reduce the number of network evaluations of Classifier-Free Guidance in text-conditioned diffusion models, thereby improving inference efficiency.
Adaptive Manipulation for Coalitions in Knockout Tournaments
Juhi Chaudhary (Tata Institute of Fundamental Research), Meirav Zehavi (Ben-Gurion University of the Negev)
🎯 What it does: This paper studies the probability-based adaptive coalition manipulation problem in single-elimination tournaments, providing definitions and solution methods for the problem.
Adaptive Market Making with Inventory Constraints via Online Learning
Shan Xue (Leshan Normal University), Liang Xu (Southwestern University of Finance and Economics)
OptimizationReinforcement Learning from Human FeedbackReinforcement LearningTime SeriesSequentialFinance Related
🎯 What it does: Designed and implemented two types of market-making reference strategies based on inventory constraints (hard constraints and soft constraints), and on this basis, constructed an adaptive no-drawdown online learning market-making strategy using polynomial weights (MW) and the FPL algorithm, proving that it satisfies no-drawdown properties and effectively controls inventory risk.
Adaptive Multi-Scale Decomposition Framework for Time Series Forecasting
Yifan Hu (Tsinghua University), Tao Dai (Shenzhen University)
OptimizationMixture of ExpertsTime Series
🎯 What it does: An adaptive multi-scale decomposition framework (AMD) based on MLP is proposed, which achieves finer time series prediction through multi-scale splitting, mixing, dual dependency interaction, and adaptive expert fusion of time series.
Adaptive Multimodal Fusion: Dynamic Attention Allocation for Intent Recognition
Bo Hu (RWTH Aachen University), Yuyang Ye (University of Science and Technology of China)
RecognitionRecurrent Neural NetworkTransformerContrastive LearningTextMultimodalityAudio
🎯 What it does: Dynamic Attention Fusion (DAF) and Multi-View Contrastive Learning (MVCL-DAF) are proposed for multimodal intent recognition;
Adaptive Prompt-Based Semantic Embedding with Inspire Potential of Implicit Knowledge for Cross-Modal Retrieval
Xin Huang (Nanyang Normal University), Jingjing Li (Peking University)
RetrievalTransformerPrompt EngineeringVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: A semantic embedding method based on adaptive prompts, APSE‑IPIK, is proposed to enhance the accuracy of cross-modal retrieval.
Adaptive Prompting for Continual Relation Extraction: A Within-Task Variance Perspective
Minh Le (VinAI Research), Thien Huu Nguyen (University of Oregon)
GenerationTransformerPrompt EngineeringMixture of ExpertsText
🎯 What it does: A no-replay continuous relation extraction framework WAVE-CRE is proposed, utilizing a task-specific prompt pool and a generative model for Prompt-Tuning on BERT to alleviate catastrophic forgetting.
Adaptive Prototype Replay for Class Incremental Semantic Segmentation
Guilin Zhu (Huazhong University of Science and Technology), Nong Sang (Hunan Normal University)
SegmentationKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: Proposes the Adapter method to address the catastrophic forgetting problem caused by model representation shift during prototype replay in class-incremental semantic segmentation, and enhances classification performance through uncertainty constraints and prototype distinction.
Adaptive Sampling to Reduce Epistemic Uncertainty Using Prediction Interval-Generation Neural Networks
Giorgio Morales (Montana State University), John W. Sheppard (Montana State University)
TabularTime SeriesAgriculture Related
🎯 What it does: This paper proposes an adaptive sampling method based on a predictive interval generating neural network (ASPINN), aimed at reducing the model's epistemic uncertainty through active sampling.
Adaptive Siamese Masked Autoencoder with Global Optimization for Unsupervised Point Cloud Shape Correspondence
Jiacheng Deng (University of Science and Technology of China), Jiahao Lu (University of Science and Technology of China)
OptimizationRepresentation LearningTransformerAuto EncoderPoint Cloud
🎯 What it does: An adaptive twin masking autoencoder and global optimization framework (AMIGO) is proposed, achieving efficient and accurate correspondence in unsupervised point cloud shape matching through an adaptive mask generator, twin MAE, and optimal transport.
Adaptive Wavelet-Positional Encoding for High-Frequency Information Learning in Implicit Neural Representation
Hongxu Zhao (Zhejiang University), Yu Zhang (Zhejiang University)
RestorationRepresentation LearningNeural Radiance FieldImageTime Series
🎯 What it does: Proposes an Adaptive Wavelet-Positional Encoding (WPE) and High-Frequency Perception (HFP) module to address the spectral bias of Implicit Neural Representation and achieve complete high-frequency detail reconstruction.
Adaptive-Grained Label Distribution Learning
Yunan Lu (Nanjing University of Science and Technology), Xiuyi Jia (Nanjing University of Science and Technology)
ClassificationData-Centric LearningSupervised Fine-TuningImage
🎯 What it does: Proposes the Adaptive-Grained Label Distribution Learning (AGLDL) framework, which first coarsens unreliable label distributions into discrete coarse-grained labels, and then refines them back to continuous label distributions using the LDL algorithm.
AdaSkip: Adaptive Sublayer Skipping for Accelerating Long-Context LLM Inference
Zhuomin He (Shanghai Jiao Tong University), Fan Wu (Shanghai Jiao Tong University)
OptimizationComputational EfficiencyTransformerLarge Language ModelText
🎯 What it does: AdaSkip is proposed, an adaptive sub-layer skipping strategy for accelerating long-context LLM inference.
ADBA: Approximation Decision Boundary Approach for Black-Box Adversarial Attacks
Feiyang Wang (Beijing University of Posts and Telecommunications), Gang Chen (Victoria University of Wellington)
OptimizationAdversarial AttackConvolutional Neural NetworkTransformerImage
🎯 What it does: This paper proposes an Approximation Decision Boundary Approach (ADBA) and its variant ADBA-md, which quickly compares and optimizes perturbation directions in black-box decision attacks, significantly reducing the number of queries.
Addressing Cold-Start Problem in Click-Through Rate Prediction via Supervised Diffusion Modeling
Wenqiao Zhu (HiThink Research), Jun Wu (HiThink Research)
Recommendation SystemDiffusion modelTabular
🎯 What it does: A supervised diffusion model (CSDM) was designed and implemented to generate warm-up embeddings for cold-start projects, thereby improving the accuracy of CTR prediction.
Addressing Multi-Label Learning with Partial Labels: From Sample Selection to Label Selection
Gengyu Lyu (Beijing University of Technology), Songhe Feng (Beijing Jiaotong University)
ClassificationObject DetectionConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: In the multi-label partially annotated scenario, a Co-Label Selection (CLS) dual-network co-training framework is proposed, which utilizes label-level selection to replace sample-level selection, automatically eliminating false negative labels while retaining training samples, thereby reducing confirmation bias and enhancing learning effectiveness.
ADELA: Accelerating Evolutionary Design of Machine Learning Pipelines with the Accompanying Surrogate Model
Yang Gu (Shanghai Jiao Tong University), Shiyou Qian (Shanghai Jiao Tong University)
ClassificationOptimizationMeta LearningRecurrent Neural NetworkAuto EncoderTabular
🎯 What it does: The ADELA method is proposed, which accelerates the design of machine learning pipelines in evolutionary AutoML by constructing an accompanying surrogate model.
Advancing Audio-Based Text Generation with Imbalance Preference Optimization
Zhenghao Zhou (National Supercomputing Center in Wuxi), Chen Cao (University of Sheffield)
GenerationOptimizationKnowledge DistillationReinforcement Learning from Human FeedbackTransformerReinforcement LearningMultimodalityAudio
🎯 What it does: A framework of 'model adversarial sampling + imbalanced preference optimization' (IPO) is proposed, which integrates limited human feedback into the alignment and performance enhancement of audio-to-text generation models.
Advancing Comprehensive Aesthetic Insight with Multi-Scale Text-Guided Self-Supervised Learning
Yuti Liu (Vivo Mobile Communication Co. Ltd), Bo Li (Vivo Mobile Communication Co. Ltd)
Recommendation SystemTransformerLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: Trained and deployed CALM - a multimodal large language model capable of providing scores, comments, personalized ratings in image aesthetic assessment, and achieving zero-shot aesthetic suggestions.
Advancing Loss Functions in Recommender Systems: A Comparative Study with a Rényi Divergence-Based Solution
Shengjia Zhang (Zhejiang University), Can Wang (Zhejiang University)
Recommendation SystemGraph Neural NetworkContrastive LearningTabular
🎯 What it does: A distributionally robust recommendation loss DrRL based on R' enyi divergence is proposed to unify the advantages of Softmax Loss and Cosine Contrastive Loss while overcoming their limitations.
Advancing Retrosynthesis with Retrieval-Augmented Graph Generation
Anjie Qiao (Sun Yat-sen University), Zhewei Wei (Renmin University of China)
GenerationRetrievalDrug DiscoveryGraph Neural NetworkTransformerDiffusion modelGraphRetrieval-Augmented Generation
🎯 What it does: A retrieval-augmented molecular graph generation framework RARB is proposed for template-free single-step backward synthesis prediction, implemented on the RetroBridge base model.
Advancing Spiking Neural Networks Towards Multiscale Spatiotemporal Interaction Learning
Yimeng Shan (Liaoning Technical University), Haicheng Qu (Liaoning Technical University)
ClassificationSpiking Neural NetworkImageTime Series
🎯 What it does: A pluggable multi-scale attention module (SMA) and attention-based zone-out regularization (AZO) are proposed, enabling spiking neural networks to simultaneously utilize multi-scale features and spatiotemporal correlations, thereby enhancing the learning effectiveness on event stream data.
AdvDisplay: Adversarial Display Assembled by Thermoelectric Cooler for Fooling Thermal Infrared Detectors
Hao Li (Xidian University), Maoguo Gong (Xidian University)
Object DetectionAdversarial AttackImage
🎯 What it does: This paper designs and implements a reusable thermoelectric cooler array (AdvDisplay) that can precisely, continuously, and bidirectionally adjust infrared radiation, used for manufacturing physically counteractive thermal infrared patches, successfully allowing pedestrians to evade detection by thermal infrared detectors.
Adversarial Attacks on Event-Based Pedestrian Detectors: A Physical Approach
Guixu Lin (University of Tokyo), Yinqiang Zheng (Singapore Management University)
Object DetectionAdversarial AttackVideo
🎯 What it does: Designed and validated a physically adversarial garment that prevents event-based camera pedestrian detectors from recognizing the wearer.
Adversarial Contrastive Graph Augmentation with Counterfactual Regularization
Tao Long (Shenzhen University), Laizhong Cui (Shenzhen University)
Representation LearningAdversarial AttackGraph Neural NetworkAuto EncoderContrastive LearningGraph
🎯 What it does: This paper proposes an Adversarial Contrastive Graph Augmentation framework (ACGA), which automatically generates positive samples containing minimal sufficient information and difficult negative samples through a conditional variational graph autoencoder, thereby enhancing model robustness in unsupervised graph representation learning.
Adversarial Contrastive Graph Masked AutoEncoder Against Graph Structure and Feature Dual Attacks
Weixuan Shen (Nanjing University of Science and Technology), Shirui Pan (Griffith University)
Representation LearningAdversarial AttackGraph Neural NetworkAuto EncoderContrastive LearningGraph
🎯 What it does: An unsupervised graph representation learning framework ACGMAE is proposed, aimed at simultaneously resisting adversarial attacks on graph structure and node features.
Adversarial Learning Under Hybrid Perturbations for Robust Acute Lymphoblastic Leukemia Classification
Jie Chen (Shenzhen University), Jianqiang Li (Shenzhen University)
ClassificationAdversarial AttackConvolutional Neural NetworkImageBiomedical Data
🎯 What it does: This paper proposes a Hybrid Perturbation Adversarial Training (HPAT) framework to enhance the robustness of acute lymphoblastic leukemia (ALL) cell classification models. It first generates pixel-level and spatial-level adversarial samples separately using Projected Gradient Descent (PGD) and Bayesian Optimization-based Spatial Transformation (STBO), and then mixes the two for training. During the training process, Mixed Batch Normalization (MixBN) is introduced to mitigate the drop in accuracy on clean samples.
Adversarial-Inspired Backdoor Defense via Bridging Backdoor and Adversarial Attacks
Jia-Li Yin (Fuzhou University), Ximeng Liu (Fuzhou University)
ClassificationAdversarial AttackConvolutional Neural NetworkImage
🎯 What it does: A backdoor defense method based on adversarial attack features, AIBD, is proposed, which identifies and removes backdoors by analyzing the differences in the number of iterations and adversarial labels of backdoor samples under adversarial attacks.
Adversity-aware Few-shot Named Entity Recognition via Augmentation Learning
Li Huang (Southwestern University of Finance and Economics), Xueqin Chen (Kash Institute of Electronics and Information Industry)
RecognitionDomain AdaptationAdversarial AttackTransformerPrompt EngineeringText
🎯 What it does: This paper proposes an adversarial adaptive augmentation learning framework (AAL) for few-shot named entity recognition, enhancing cross-domain robustness through prompt-driven data augmentation and domain-aware prototype learning.
AE-NeRF: Augmenting Event-Based Neural Radiance Fields for Non-ideal Conditions and Larger Scenes
Chaoran Feng (Peking University), Yonghong Tian (Peking University)
RestorationGenerationPose EstimationNeural Radiance FieldImageVideo
🎯 What it does: This paper studies a neural radiance field (NeRF) reconstruction method based on event cameras, called AE-NeRF, which can achieve high-quality 3D reconstruction and novel view synthesis under non-ideal conditions (such as pose noise, sparse events, and large scenes).
Aerodynamic Coefficients Prediction via Cross-Attention Fusion and Physical-Informed Training
Yueqing Wang (China Aerodynamics Research and Development Center), Yi Chen (Sichuan Tianfu Fluid Big Data Research Center)
Autonomous DrivingOptimizationTransformerPoint Cloud
🎯 What it does: A deep learning framework is proposed that integrates cross-attention fusion and physical information constraints to predict the aerodynamic coefficients of complex geometries under different flow conditions.
AeroGTO: An Efficient Graph-Transformer Operator for Learning Large-Scale Aerodynamics of 3D Vehicle Geometries
Pengwei Liu (Zhejiang University), Dong Ni (Zhejiang University)
Autonomous DrivingComputational EfficiencyGraph Neural NetworkTransformerMesh
🎯 What it does: AeroGTO is designed and implemented, an operator that combines graph neural networks and transformers to quickly and accurately predict surface pressure and drag coefficients on large-scale meshes of three-dimensional vehicle geometries.
AFFAKT: A Hierarchical Optimal Transport Based Method for Affective Facial Knowledge Transfer in Video Deception Detection
Zihan Ji (South China University of Technology), Ye Liu (Beijing Normal University)
ClassificationDomain AdaptationAnomaly DetectionExplainability and InterpretabilityRecurrent Neural NetworkSupervised Fine-TuningVideoMultimodalityAudio
🎯 What it does: A hierarchical optimal transport-based emotional facial knowledge transfer method (AFFAKT) is proposed to enhance video deception detection performance.
Affirm: Interactive Mamba with Adaptive Fourier Filters for Long-term Time Series Forecasting
Yuhan Wu (Zhejiang University), Dongming Lu (Zhejiang University)
Time Series
🎯 What it does: A lightweight time series forecasting model called Affirm is proposed, which combines adaptive Fourier filters and a dual-interaction Mamba module to achieve efficient modeling of long sequences.
Affordances-Oriented Planning Using Foundation Models for Continuous Vision-Language Navigation
Jiaqi Chen (University of Hong Kong), Kwan-Yee K. Wong (University of Hong Kong)
OptimizationKnowledge DistillationRobotic IntelligenceTransformerLarge Language ModelVision Language ModelMultimodality
🎯 What it does: This paper proposes AOPlanner, which utilizes multimodal foundational models to achieve low-level motion planning and high-level decision-making in continuous visual language navigation, completing navigation tasks under zero-shot conditions.
AFiRe: Anatomy-Driven Self-Supervised Learning for Fine-Grained Representation in Radiographic Images
Yihang Liu (Tongji University), Hongzhou Chen (Tongji University)
ClassificationSegmentationAnomaly DetectionRepresentation LearningTransformerContrastive LearningImageBiomedical Data
🎯 What it does: The AFiRe framework is proposed, combining anatomy-driven self-supervised learning with ViT-based token-level contrastive learning and pixel-level anomaly removal reconstruction to achieve fine-grained representation of chest X-rays.
Against All Odds: Overcoming Typology, Script, and Language Confusion in Multilingual Embedding Inversion Attacks
Yiyi Chen (Aalborg University), Johannes Bjerva (Aalborg University)
Adversarial AttackTransformerLarge Language ModelText
🎯 What it does: Evaluate embedding inversion attacks on multilingual large language models, studying vulnerabilities across languages and scripts, and exploring language confusion phenomena.
Agent-Aware Training for Agent-Agnostic Action Advising in Deep Reinforcement Learning
Yaoquan Wei (Zhejiang University), Mingli Song (Zhejiang University)
Reinforcement LearningContrastive LearningSequential
🎯 What it does: A framework called Agent-Aware trAining yet Agent-Agnostic Action Advising (A7) is proposed, which utilizes an agent model to extract state features and determines when to request expert action advice based on feature similarity, while also introducing behavior cloning reuse models and intrinsic rewards to accelerate DRL learning.
Agent4Edu: Generating Learner Response Data by Generative Agents for Intelligent Education Systems
Weibo Gao (University of Science and Technology of China), Zhenya Huang (University of Science and Technology of China)
GenerationData SynthesisRecommendation SystemTransformerLarge Language ModelReinforcement LearningAgentic AITextSequentialChain-of-Thought
🎯 What it does: Using large language model-driven generative agents to simulate personalized learner practice response data on online education platforms and record their learning processes.
AgentMixer: Multi-Agent Correlated Policy Factorization
Zhiyuan Li (Aalto University), Joni Pajarinen (Aalto University)
Reinforcement LearningSequential
🎯 What it does: The AgentMixer framework is proposed, which achieves relevant policy decomposition based on global information in multi-agent reinforcement learning through the Policy Modifier and Individual-Global-Consistency modules, and supports decentralized execution.
AGFSync: Leveraging AI-Generated Feedback for Preference Optimization in Text-to-Image Generation
Jingkun An (Beihang University), Chengwei Pan (Beihang University)
GenerationOptimizationVision Language ModelDiffusion modelImageTextMultimodality
🎯 What it does: Aiming to align text-to-image diffusion models using a fully AI-driven feedback loop (VLM evaluation, VQA, CLIP, aesthetic scoring), the AGFSync framework is proposed;
AGLLDiff: Guiding Diffusion Models Towards Unsupervised Training-free Real-world Low-light Image Enhancement
Yunlong Lin (Xiamen University), Xinghao Ding (Xiamen University)
RestorationGenerationDiffusion modelImage
🎯 What it does: A training-free and unsupervised Attribute Guidance Diffusion framework (AGLLDiff) is proposed, which guides the diffusion model to generate low-light images with high-quality attributes such as exposure, structure, and color.
AGMixup: Adaptive Graph Mixup for Semi-supervised Node Classification
Weigang Lu (Xidian University), Dapeng Tao (Xidian University)
ClassificationGraph Neural NetworkGraph
🎯 What it does: An Adaptive Graph Mixup (AGMixup) framework is proposed for semi-supervised node classification.
AI-generated Image Quality Assessment in Visual Communication
Yu Tian (City University of Hong Kong), Sam Kwong (Lingnan University)
GenerationData SynthesisTransformerLarge Language ModelDiffusion modelImageMultimodality
🎯 What it does: This study constructed the AIGI-VC dataset, which includes 2,500 AI-generated images covering 14 advertising themes and 8 types of emotions, and provides preference annotations from coarse to fine, used to evaluate the clarity of information and emotional interaction in visual communication.
AI-Powered Algorithm-Centric Quantum Processor Topology Design
Tian Li (Henan Key Laboratory of Quantum Information and Cryptography), He-Liang Huang (Henan Key Laboratory of Quantum Information and Cryptography)
OptimizationReinforcement LearningTabularBenchmarkPhysics Related
🎯 What it does: Dynamically designing the topology of quantum processors and mapping quantum circuits through reinforcement learning significantly reduces circuit depth.
AIA: Autoregression-Based Injection Attacks Against Text2SQL Models
Deyin Li (Zhejiang University), Chunming Wu (Stony Brook University)
Safty and PrivacyAdversarial AttackTransformerLarge Language ModelText
🎯 What it does: This study investigates the security of the Text2SQL model and proposes and evaluates an Autoregressive Injection Attack (AIA).
AIF-SFDA: Autonomous Information Filter Driven Source-Free Domain Adaptation for Medical Image Segmentation
Haojin Li (Southern University of Science and Technology), Jiang Liu (Beijing Information Science and Technology University)
SegmentationDomain AdaptationImageBiomedical DataUltrasound
🎯 What it does: Proposes AIF-SFDA, which implements domain adaptation for passive domain medical image segmentation using frequency domain adaptive information filters.
AIM: Additional Image Guided Generation of Transferable Adversarial Attacks
Teng Li (Fudan University), Yu-Gang Jiang (Fudan University)
GenerationAdversarial AttackConvolutional Neural NetworkTransformerGenerative Adversarial NetworkContrastive LearningImage
🎯 What it does: A lightweight Semantic Injection Module (SIM) is proposed to enhance transferability when generating adversarial examples by utilizing the semantic information of additional guiding images.
AIM: Let Any Multimodal Large Language Models Embrace Efficient In-Context Learning
Jun Gao (Soochow University), Wenjie Li (Hong Kong Polytechnic University)
TransformerLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: The AIM framework is proposed, enabling any multimodal large language model to achieve efficient context learning without updating model parameters.
AIQViT: Architecture-Informed Post-Training Quantization for Vision Transformers
Runqing Jiang (Sun Yat-sen University), Yulan Guo (Sun Yat-sen University)
ClassificationObject DetectionSegmentationNeural Architecture SearchTransformerImagePoint Cloud
🎯 What it does: A post-training quantization method for Vision Transformers (ViT) called AIQViT is proposed, which combines architecture-aware low-rank compensation and dynamic focus quantization to achieve low bit-width (3/4/6 bits) quantization.
AIR: Unifying Individual and Collective Exploration in Cooperative Multi-Agent Reinforcement Learning
Guangchong Zhou (Chinese Academy of Sciences), Guoliang Fan (Chinese Academy of Sciences)
Reinforcement Learning
🎯 What it does: The AIR (Adaptive exploration via Identity Recognition) framework is proposed, which implements individual and collective exploration in value-based multi-agent reinforcement learning through a unified identity recognizer, and dynamically adjusts the exploration method via adaptive temperature.
ALADE-SNN: Adaptive Logit Alignment in Dynamically Expandable Spiking Neural Networks for Class Incremental Learning
Wenyao Ni (Zhejiang University), Huajin Tang (Dalian University of Technology)
ClassificationKnowledge DistillationSpiking Neural NetworkImage
🎯 What it does: A dynamically expandable spiking neural network framework ALADE-SNN suitable for class-incremental learning is proposed.
Aligning and Prompting Anything for Zero-Shot Generalized Anomaly Detection
Jitao Ma (Xidian University), Leyuan Fang (Hunan University)
Anomaly DetectionTransformerPrompt EngineeringVision Language ModelImageBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: This paper proposes a zero-shot generalized anomaly detection (ZGAD) method based on Text Prompt Shunt (TPS), which can simultaneously achieve image-level anomaly classification and pixel-level anomaly segmentation.
Aligning Composed Query with Image via Discriminative Perception from Negative Correspondences
Yifan Wang (Tsinghua University), Chun Yuan (Chinese Academy of Sciences)
RetrievalContrastive LearningImageText
🎯 What it does: Proposes the DIPNEC framework, which enhances the performance of synthesized query image retrieval through discriminative perception of negative correspondence, differential quantization alignment, and word-level alignment.
Aligning Instance Brownian Bridge with Texts for Open-Vocabulary Video Instance Segmentation
Zesen Cheng (Peking University), Jie Chen (Peking University)
Object DetectionSegmentationTransformerContrastive LearningVideoText
🎯 What it does: This paper proposes a framework for video instance segmentation using the Brownian Bridge model (BriVIS), which constructs a Brownian bridge on frame-level instance features and aligns the bridge center with text features to achieve open vocabulary video instance segmentation.
Aligning Language Models Using Follow-up Likelihood as Reward Signal
Chen Zhang (National University of Singapore), Haizhou Li (Tencent AI Lab)
TransformerLarge Language ModelReinforcement LearningText
🎯 What it does: This paper proposes a new reward mechanism called 'Follow-up Likelihood as Reward (FLR)', which uses the likelihood of user follow-up statements after generating replies as an unannotated reward signal. It constructs preference data by automating the annotation of online generated replies and further enhances model helpfulness using DAP (such as DPO).
Alignment of CNN and Human Judgments of Geometric and Topological Concepts
Neha Upadhyay (Georgia Institute of Technology), Sashank Varma (Georgia Institute of Technology)
ClassificationRecognitionConvolutional Neural NetworkImage
🎯 What it does: The study investigates the sensitivity of CNN models to human geometric and topological concepts and compares the performance of individuals from different age groups and cultural backgrounds.
Alignment-Free RGB-T Salient Object Detection: A Large-Scale Dataset and Progressive Correlation Network
Kunpeng Wang (Anhui University), Bin Luo (Anhui University)
Object DetectionTransformerImageMultimodality
🎯 What it does: This paper proposes a large-scale unaligned RGB-Thermal visual salient object detection dataset named UVT20K, and designs a Progressive Correlation Network (PCNet) based on this dataset for salient object detection in unaligned image pairs.
All You Need in Knowledge Distillation Is a Tailored Coordinate System
Junjie Zhou (Nanjing University), Jianxin Wu (Nanjing University)
ClassificationKnowledge DistillationConvolutional Neural NetworkTransformerImage
🎯 What it does: A teacher-free, flexible, and efficient knowledge distillation method called TCS is proposed, which utilizes the feature coordinate system of SSL pre-trained models to transfer dark knowledge through PCA and iterative feature selection.
All-in-One: Transferring Vision Foundation Models into Stereo Matching
Jingyi Zhou (Fudan University), Yangyang Zhang (Xiaomi Inc.)
Depth EstimationOptimizationTransformerImage
🎯 What it does: By flexibly selecting and transferring knowledge from various visual foundation models to a single stereo matching network, stronger feature representation is achieved.
Alleviate and Mining: Rethinking Unsupervised Domain Adaptation for Mitochondria Segmentation from Pseudo-Label Perspective
Yujia Chen (University of Science and Technology of China), Tianzhu Zhang (University of Science and Technology of China)
SegmentationDomain AdaptationImageBiomedical Data
🎯 What it does: Designed and implemented the R4MITO framework for unsupervised domain adaptation of mitochondrial segmentation in unlabeled target domains.
Alleviating Performance Disparity in Adversarial Spatiotemporal Graph Learning Under Zero-Inflated Distribution
Songran Bai (Institute of Automation, Chinese Academy of Sciences), Daniel Dajun Zeng (Institute of Automation, Chinese Academy of Sciences)
OptimizationAdversarial AttackGraph Neural NetworkContrastive LearningGraphTime Series
🎯 What it does: The MinGRE framework is proposed to improve the robustness of graph-based spatiotemporal learning models and reduce the performance gap of minority classes through multi-dimensional attention reweighted gradients and uncertainty-guided contrastive learning under zero-inflated distribution (ZID).
Alleviating Shifted Distribution in Human Preference Alignment through Meta-Learning
Shihan Dou (Fudan University), Xuanjing Huang (Fudan University)
Meta LearningReinforcement Learning from Human FeedbackSupervised Fine-TuningReinforcement LearningText
🎯 What it does: The MetaRM method is proposed, which alternately optimizes the reward model through meta-learning during the RLHF process to adapt to the drift in the output distribution of the policy model, maintaining the ability to distinguish new distribution samples while not losing the original preference alignment effect.
ALLVB: All-in-One Long Video Understanding Benchmark
Xichen Tan (National University of Defense Technology), Zhiping Cai (National University of Defense Technology)
TransformerLarge Language ModelVideoTextBenchmark
🎯 What it does: Proposed the ALLVB long video understanding benchmark, covering 1,376 movie long videos and 252k multiple-choice QA, encompassing 9 types of video tasks.
AlphaForge: A Framework to Mine and Dynamically Combine Formulaic Alpha Factors
Hao Shi (University of Chinese Academy of Sciences), Luis Angel Seco (University of Toronto)
Recommendation SystemOptimizationExplainability and InterpretabilityComputational EfficiencyReinforcement LearningGenerative Adversarial NetworkTabularTime SeriesFinance Related
🎯 What it does: The AlphaForge framework is proposed, which is divided into two stages: a generative predictive network is used to mine formalized Alpha factors, and then a dynamic factor combination model dynamically adjusts the weights based on the latest performance of the factors to form a Mega-Alpha signal.
ALRMR-GEC: Adjusting Learning Rate Based on Memory Rate to Optimize the Edit Scorer for Grammatical Error Correction
Zhixiao Wu (Harbin Institute of Technology), Guangming Lu (Harbin Institute of Technology)
GenerationOptimizationTransformerSupervised Fine-TuningText
🎯 What it does: This paper proposes a dynamic learning rate adjustment method based on memory rate, ALRMR-GEC, to improve the generalization performance of edit-based grammar error correction models.
ALTBI: Constructing Improved Outlier Detection Models via Optimization of Inlier-Memorization Effect
Seoyoung Cho (Sungshin Women's University), Dongha Kim (Sungshin Women's University)
Anomaly DetectionOptimizationFlow-based ModelGenerative Adversarial NetworkImageTextTabularBenchmark
🎯 What it does: A novel unsupervised anomaly detection framework, ALTBI, is proposed, which maximizes anomaly recognition by utilizing the 'in-point memory effect' of deep generative models.
Ambiguity-Restrained Text-Video Representation Learning for Partially Relevant Video Retrieval
Cheol-Ho Cho (Sungkyunkwan University), Jae-Pil Heo (Sungkyunkwan University)
RetrievalRepresentation LearningTransformerContrastive LearningVideoText
🎯 What it does: This paper proposes an ambiguity-restricted representation learning (ARL) based on uncertainty and similarity detection to address the ambiguity problem of text-video labels in partially related video retrieval.
Ambiguous Instance-Aware Contrastive Network with Multi-Level Matching for Multi-View Document Clustering
Zhenqiu Shu (Kunming University of Science and Technology), Zhengtao Yu (Kunming University of Science and Technology)
Auto EncoderContrastive LearningText
🎯 What it does: A multi-view document clustering method based on fuzzy instance-aware contrastive learning, AICN-MLM, is proposed.
Amplifier: Bringing Attention to Neglected Low-Energy Components in Time Series Forecasting
Jingru Fei (Beijing Institute of Technology), Zhendong Niu (Beijing Institute of Technology)
TransformerTime Series
🎯 What it does: Proposes energy amplification and recovery technology, and constructs an Amplifier model based on this technology to enhance the learning of low-energy spectral components, thereby improving time series forecasting.
An Algebraic Notion of Conditional Independence, and Its Application to Knowledge Representation
Jesse Heyninck (Open Universiteit)
🎯 What it does: This paper proposes a concept of conditional independence based on the Algebraic Fixed Point Theory (AFT) and applies it to various formalizations of knowledge representation, such as propositional logic, belief revision, conditional logic, and ordinary logic programming.
An Alternative Theory of Stable Revision for Nondeterministic Approximation Fixpoint Theory and the Relationships
Spencer Killen (University of Alberta), Jesse Heyninck (Open Universiteit)
🎯 What it does: A theory is proposed that uses a set of deterministic approximation operators to define non-deterministic stable revisions, thereby capturing the weak stable models and answer sets of disjunctive logic programs (DLP) at an abstract level.
An And-Sum Circuit with Signed Edges That Is More Succinct than SDD
Ryoma Onaka (NTT Corporation), Norihito Yasuda (NTT Corporation)
🎯 What it does: This paper proposes a new representation of Boolean functions - Structured Decomposable And-Sum Circuits (st-DASC), which achieves polynomial-time processing of logical negation by introducing signed + nodes, while maintaining the structured decomposable property and providing support for various queries and transformations in knowledge compilation graphs.
An Automatic Sound and Complete Abstraction Method for Generalized Planning with Baggable Types
Hao Dong (Sun Yat-sen University), Yongmei Liu (Sun Yat-sen University)
🎯 What it does: This paper proposes an automated abstract method with soundness and completeness for generating Bounded QNP (BQNP) abstractions from STRIPS domain instances containing countable types (baggable types) to solve generalized planning problems.
An Efficient and Accurate Dynamic Sparse Training Framework Based on Parameter-Freezing
Lei Li (Cleveland State University), Tianyun Zhang (Cleveland State University)
Federated LearningConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: A dynamic sparse training framework based on parameter freezing and server-side sparse mask re-tuning is proposed in federated learning, significantly reducing communication and computation costs.
An Efficient Framework for Enhancing Discriminative Models via Diffusion Techniques
Chunxiao Li (Beijing Normal University), Yao Zhu (Zhejiang University)
ClassificationConvolutional Neural NetworkTransformerDiffusion modelImage
🎯 What it does: A framework based on diffusion models (DBMEF) is proposed, which provides a 're-thinking' function for existing discriminative models without the need for additional training, significantly improving classification accuracy.
An Elite-guided Weighted Simulated Annealing Algorithm for the Clique Partitioning Problem
Baiyu Chen (Huazhong University of Science and Technology), Zhipeng Lü (Huazhong University of Science and Technology)
OptimizationGraph
🎯 What it does: An elite-guided weighted simulated annealing algorithm named EWSA is proposed and implemented to solve the clique partition problem (CPP) of complete graphs, improving search efficiency and solution quality through the alternating use of two search configurations, a weighted scoring function, and partition constraint strategies.
An Enhanced Levenberg--Marquardt Method via Gram Reduction
Chengchang Liu (Fudan University), John C.S. Lui (Chinese University of Hong Kong)
OptimizationTabular
🎯 What it does: A Gram-Reduced Levenberg-Marquardt (GRLM) method is proposed for efficiently solving nonlinear equation systems while ensuring global convergence and local superlinear convergence;
An Evaluation Framework for Product Images Background Inpainting Based on Human Feedback and Product Consistency
Yuqi Liang (Ant Group), Jianqi Bi (Ant Group)
Image TranslationRestorationSegmentationReinforcement Learning from Human FeedbackTransformerImageMultimodality
🎯 What it does: This paper proposes an automatic evaluation framework named HFPC for detecting the background appropriateness and product consistency of product images after AI background filling.
An Exemplar-based Framework for Chinese Text Recognition
Zhao Zhou (Fudan University), Cheng Jin (Fudan University)
RecognitionRetrievalTransformerText
🎯 What it does: A Chinese text recognition framework based on sample retrieval is proposed—DECTR, which first discovers character samples through a weakly supervised feature extraction network, and then retrieves similar characters from an external sample library to correct recognition errors.
An Extension-Based Argument-Ranking Semantics: Social Rankings in Abstract Argumentation
Lars Bengel (University of Hagen), Kenneth Skiba (University of Hagen)
🎯 What it does: This paper studies a new argument ranking semantics based on social ranking functions, used for fine-grained ranking of arguments in an argumentation framework.
An Information-theoretic Multi-task Representation Learning Framework for Natural Language Understanding
Dou Hu (Institute of Information Engineering, Chinese Academy of Sciences), Songlin Hu (Institute of Information Engineering, Chinese Academy of Sciences)
ClassificationRepresentation LearningTransformerLarge Language ModelText
🎯 What it does: This paper proposes an information-theoretic multi-task representation learning framework called InfoMTL, aimed at learning noise-invariant and sufficient shared and task-specific representations in multi-task scenarios.
An Item Is Worth a Prompt: Versatile Image Editing with Disentangled Control
Aosong Feng (Yale University), Leandros Tassiulas (Yale University)
Image TranslationSegmentationGenerationTransformerSupervised Fine-TuningPrompt EngineeringDiffusion modelImageTextMultimodality
🎯 What it does: The D-Edit framework is proposed, which segments the input image into several 'items', learns a unique prompt for each item, and achieves decoupling of prompts and items through grouped cross-attention, enabling various editing tasks such as text, image, mask editing, and item deletion.
An LLM-Empowered Adaptive Evolutionary Algorithm for Multi-Component Deep Learning Systems
Haoxiang Tian (Nanyang Technological University), Tianwei Zhang (Nanyang Technological University)
Anomaly DetectionAutonomous DrivingOptimizationLarge Language ModelPrompt EngineeringMultimodality
🎯 What it does: An adaptive evolutionary algorithm based on LLM (µ MOEA) is proposed for the detection of security violations in multi-component deep learning systems.
An Open-Ended Learning Framework for Opponent Modeling
Yuheng Jing (Institute of Automation Chinese Academy of Sciences), Jian Cheng (Tencent AI Lab)
TransformerReinforcement LearningSequential
🎯 What it does: This study proposes an Open-Ended Opponent Modeling framework (OEOM) and designs a Transformer-based ICRL opponent modeling method (IOM) within this framework.
An Optimal Transport-based Latent Mixer for Robust Multi-modal Learning
Fengjiiao Gong, Hongteng Xu (Renmin University of China)
ClassificationFederated LearningSafty and PrivacyRepresentation LearningAuto EncoderMultimodality
🎯 What it does: Designed and implemented an OTM mixer based on optimal transport, which performs privacy-friendly and robust feature fusion and enhancement of multi-head Wasserstein autoencoders on distributed, unaligned multimodal data;
AnalogCoder: Analog Circuit Design via Training-Free Code Generation
Yao Lai (University of Hong Kong), Ping Luo (University of Hong Kong)
OptimizationAI Code AssistantTransformerLarge Language ModelPrompt EngineeringTextChain-of-Thought
🎯 What it does: This paper presents AnalogCoder, a training-agnostic agent based on LLM that can automate circuit design by generating Python code.
Analytical-Chemistry-Informed Transformer for Infrared Spectra Modeling
Shiluo Huang (Southwestern University of Finance and Economics), Ying Mu (Zhejiang University)
TransformerTabularAgriculture Related
🎯 What it does: An Analytical-Chemistry-Informed Transformer (ACT) aimed at infrared spectrum modeling is proposed, achieving domain-invariant representation through a learnable baseline correction module and spectral attention mechanism to address the calibration transfer problem.
Anatomical Knowledge Mining and Matching for Semi-supervised Medical Multi-structure Detection
Bin Pu (Hunan University), Kenli Li (Hunan University)
Object DetectionGraph Neural NetworkImageBiomedical DataUltrasound
🎯 What it does: Developed Semi-akmm, a semi-supervised multi-structure detection method based on a teacher-student framework, which utilizes the anatomical prior knowledge contained in ultrasound images to enhance detection performance.
Anchor Learning with Potential Cluster Constraints for Multi-view Clustering
Yawei Chen (Dalian Maritime University), Yang Wang (Hefei University of Technology)
OptimizationImageVideo
🎯 What it does: This paper proposes an Anchor Learning with Potential Cluster Constraints (ALPC) method, which unifies the selection of anchors and the construction of anchor graphs into a single framework, and achieves a uniform distribution of anchors across clusters and alignment with the original data cluster centers through potential cluster constraints.
Anchor Search: A Unified Framework for Suboptimal Bidirectional Search
Sepehr Lavasani (University of Alberta), Nathan R. Sturtevant (University of Alberta)
OptimizationTabularBenchmark
🎯 What it does: Proposed the Anchor Search framework for unbounded suboptimal bidirectional heuristic search;
AnchorInv: Few-Shot Class-Incremental Learning of Physiological Signals via Feature Space-Guided Inversion
Chenqi Li (University of Oxford), Tingting Zhu (University of Oxford)
ClassificationData SynthesisConvolutional Neural NetworkTransformerTime SeriesBiomedical Data
🎯 What it does: This paper proposes the AnchorInv method, which utilizes anchor points in the feature space to guide model inversion, generating synthetic samples of previously learned categories to achieve few-shot class incremental learning and prevent catastrophic forgetting.