AAAI 2025 Papers — Page 3
AAAI Conference on Artificial Intelligence · 3028 papers
Answering Conjunctive Queries with Safe Negation and Inequalities over RDFS Knowledge Bases
Gianluca Cima (Sapienza University of Rome), Antonella Poggi (Sapienza University of Rome)
🎯 What it does: This paper studies the combined complexity of using safe negation and inequalities in DL-Lite RDFS ontologies, providing complete upper and lower bounds.
Anti-Diffusion: Preventing Abuse of Modifications of Diffusion-Based Models
Li Zheng (University of Macau), Jinyu Tian (Macau University of Science and Technology)
Safty and PrivacyTransformerPrompt EngineeringDiffusion modelImage
🎯 What it does: Proposes the Anti-Diffusion privacy protection system, which adds small adversarial noise to images before release to prevent their misuse by diffusion model-based tuning and editing methods.
AnyTalk: Multi-modal Driven Multi-domain Talking Head Generation
Yu Wang (International Digital Economy Academy), Yu Li (International Digital Economy Academy)
GenerationData SynthesisOptical FlowImageVideoMultimodality
🎯 What it does: A unified cross-domain lip-sync animation framework called AnyTalk is proposed, which can achieve lip-sync animation between different styles (such as real humans, cartoon animals, and Disney characters) without paired data.
Anytime Multi-Agent Path Finding with an Adaptive Delay-Based Heuristic
Thomy Phan (University of Southern California), Sven Koenig (University of California, Irvine)
OptimizationReinforcement LearningAgentic AIBenchmark
🎯 What it does: This paper proposes a new single destruction heuristic called ADDRESS, which utilizes constrained Thompson Sampling to select the agent with the highest delay as a seed, improving the LNS neighborhood generation method of traditional MAPF-LNS.
Anywhere: A Multi-Agent Framework for User-Guided, Reliable, and Diverse Foreground-Conditioned Image Generation
Xie Tianyidan (Nanjing University), Zili Yi (Nanjing University)
GenerationVision Language ModelDiffusion modelImage
🎯 What it does: A multi-agent framework (Anywhere) is proposed for foreground conditional image generation, which maintains foreground integrity, enhances background diversity, and ensures text consistency.
AoP-SAM: Automation of Prompts for Efficient Segmentation
Yi Chen (Korea Advanced Institute of Science and Technology), Joo-Young Kim (Korea Advanced Institute of Science and Technology)
SegmentationTransformerPrompt EngineeringImage
🎯 What it does: This paper proposes AoP-SAM, which automatically generates key point prompts required for SAM by learning a Prompt Predictor, achieving efficient image segmentation without manual prompting.
APAR: Modeling Irregular Target Functions in Tabular Regression via Arithmetic-Aware Pre-Training and Adaptive-Regularized Fine-Tuning
Hong-Wei Wu (National Yang Ming Chiao Tung University), Wen-Chih Peng (National Yang Ming Chiao Tung University)
TransformerTabular
🎯 What it does: By using arithmetic-aware pre-training and adaptive regularization fine-tuning, we improve the table regression model to address the overfitting problem caused by irregular objective functions.
APIRL: Deep Reinforcement Learning for REST API Fuzzing
Myles Foley (Imperial College London), Sergio Maffeis (Imperial College London)
TransformerReinforcement LearningText
🎯 What it does: A REST API fuzz testing framework called APIRL based on deep reinforcement learning is proposed, which automatically generates and mutates HTTP requests to find server-side errors.
APKGC: Noise-enhanced Multi-Modal Knowledge Graph Completion with Attention Penalty
Yue Jian (Hubei University), Xiaoju Hou (Guangdong Industry Polytechnic University)
Graph Neural NetworkTransformerLarge Language ModelVision Language ModelMultimodalityGraph
🎯 What it does: The paper proposes the APKGC method, which uses attention penalty and adaptive noise sampling to complete multimodal knowledge graph completion.
Apollo-Forecast: Overcoming Aliasing and Inference Speed Challenges in Language Models for Time Series Forecasting
Tianyi Yin (Tongji University), Weiming Shen (Tongji University)
TransformerLarge Language ModelTime Series
🎯 What it does: The Apollo-Forecast framework is proposed to address the issues of aliasing distortion and slow inference in time series forecasting.
Approximate Bilevel Difference Convex Programming for Bayesian Risk Markov Decision Processes
Yifan Lin (Georgia Institute of Technology), Enlu Zhou (Georgia Institute of Technology)
OptimizationReinforcement LearningTabular
🎯 What it does: This paper proposes an infinite-horizon Bayesian Risk Markov Decision Process (BR-MDP) and provides a solution framework based on Approximate Bi-level Differential Convex Programming (ABDCP), outputting an optimal policy that can be represented as a finite state controller.
Approximate State Abstraction for Markov Games
Hiroki Ishibashi (University of Electro-Communications), Atsushi Iwasaki (University of Electro-Communications)
OptimizationReinforcement LearningTabular
🎯 What it does: This paper studies approximate state abstraction in two-player zero-sum Markov games (TZMG), proposes a clustering method based on minimax value, and provides an upper bound on the distance (duality gap) between the equilibrium strategies after abstraction and those of the original game.
Approximated Variational Bayesian Inverse Reinforcement Learning for Large Language Model Alignment
Yuang Cai (Beijing University of Posts and Telecommunications), Qinhong Lin (Beijing University of Posts and Telecommunications)
Reinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningText
🎯 What it does: Proposes an Approximate Variational Alignment (AVA) method based on Bayesian Inverse Reinforcement Learning for alignment tasks of large language models.
Approximating Metric Magnitude of Point Sets
Rayna Andreeva (University of Edinburgh), Rik Sarkar (University of Edinburgh)
OptimizationTabular
🎯 What it does: This study proposes various algorithms for fast approximation of the size of metric spaces (Magnitude), including convex optimization frameworks, iterative normalization, greedy subset selection, and discrete center hierarchy, and applies them to neural network regularization and clustering tasks.
Approximating Optimal Labelings for Temporal Connectivity
Daniele Carnevale (Gran Sasso Science Institute), Martin Olsen (Aarhus University)
OptimizationGraph Neural NetworkGraph
🎯 What it does: The study explores how to assign the minimum number of labels to an undirected graph to achieve temporal connectivity under a given maximum allowable age a, clarifying the complexity and approximability of the problem.
APR-RD: Complemental Two Steps for Self-Supervised Real Image Denoising
Hyunjun Kim (Seoul National University), Nam Ik Cho (Seoul National University)
RestorationKnowledge DistillationConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: A two-stage self-supervised denoising framework APR-RD is proposed: first, the Adjacent Pixel Replacer (APR) generates decorrelated noise pairs without downsampling, and then the Recharged Distillation (RD) enhances detail recovery through multi-target distillation, all trained on a single noisy image.
AQUAFace: Age-Invariant Quality Adaptive Face Recognition for Unconstrained Selfie vs ID Verification
Shivang Agarwal (Indian Institute of Technology Jodhpur), Sangeeth Reddy Battu (Swiggy)
RecognitionOptimizationConvolutional Neural NetworkGenerative Adversarial NetworkContrastive LearningImage
🎯 What it does: A unified loss function AQUAFace is proposed, which combines age and image quality information to recognize the pairing of self-portraits and ID photos, enhancing robustness through the joint optimization of contrastive loss and identity loss.
Arbitrary Reading Order Scene Text Spotter with Local Semantics Guidance
Jiahao Lyu (Institute of Information Engineering Chinese Academy of Science), Yu Zhou (Nankai University)
RecognitionObject DetectionConvolutional Neural NetworkTransformerImageText
🎯 What it does: A scene text detection and recognition method called LSGSpotter is proposed, which locates and recognizes text in any reading order simultaneously through an autoregressive approach.
Architecture-Aware Learning Curve Extrapolation via Graph Ordinary Differential Equation
Yanna Ding (Rensselaer Polytechnic Institute), Jianxi Gao (Rensselaer Polytechnic Institute)
OptimizationNeural Architecture SearchGraph Neural NetworkAuto EncoderTabularTime SeriesOrdinary Differential Equation
🎯 What it does: A framework-aware learning curve extrapolation model LC-GODE has been developed, which utilizes graph neural networks to encode network structures and embeds them into neural ODEs to predict the learning curves of different neural networks during the early training phase using variational inference.
Are Expressive Models Truly Necessary for Offline RL?
Guan Wang (Tsinghua University), Xianyuan Zhan (Shanghai AI Laboratory)
Reinforcement LearningTabularBenchmark
🎯 What it does: This paper proposes a Recursive Step Planning (RSP) method that uses a shallow two-layer MLP to predict sub-goals and executes actions through a goal-conditioned policy, addressing the long-term error problem in offline reinforcement learning.
Are Key-Phrases All That Reviewers Care About? A Comprehensive Benchmarking of Reviewer Matchmaking Systems
Sourish Dasgupta (Dhirubhai Ambani Institute of Information and Communication Technology), Anil K. Roy (Dhirubhai Ambani Institute of Information and Communication Technology)
Recommendation SystemTransformerSupervised Fine-TuningTextBenchmark
🎯 What it does: A comprehensive benchmark test of existing Reviewer Matchmaking (RM) systems is conducted, systematically comparing two types of models: Document Representation (DR) and Keyword Extraction (KPE). The shortcomings of traditional evaluation metrics (Precision@K, Kendall Loss) are pointed out, and the use of correlation coefficients (Pearson, Spearman, Kendall) is proposed as a reliable evaluation method.
Argumentative Large Language Models for Explainable and Contestable Claim Verification
Gabriel Freedman (Imperial College London), Francesca Toni (Imperial College London)
Explainability and InterpretabilityTransformerLarge Language ModelPrompt EngineeringTextChain-of-Thought
🎯 What it does: This paper proposes an 'Argument LLM (ArgLLM)' that combines large language models (LLM) with a formal argumentation framework (QBAF), achieving interpretable and debatable decision outputs by constructing argument trees that support and refute.
ARNet: Self-Supervised FG-SBIR with Unified Sample Feature Alignment and Multi-Scale Token Recycling
Jianan Jiang (Hunan University), Di Wu (University of Warwick)
RetrievalTransformerContrastive LearningImage
🎯 What it does: A self-supervised FG-SBIR framework ARNet has been developed, which achieves intra- and inter-sample feature alignment using a dual-weight shared encoder, and enhances fine-grained image retrieval performance by reusing discarded patch tokens in ViT through a multi-scale token recovery module.
ARTICLE: Annotator Reliability Through In-Context Learning
Sujan Dutta (Rochester Institute of Technology), Ashiqur R. KhudaBukhsh (Rochester Institute of Technology)
TransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: This paper proposes the ARTICLE framework, which utilizes context learning from large language models to evaluate the self-consistency of annotators, thereby identifying reliable annotators and modeling the perception of aggression among different political groups.
As Pseudo-Label Free as Possible: Leveraging Adaptive Feature Generation for Sparsely Annotated Object Detection
Shuilian Yao (Dalian University of Technology), Wei Zhuo (Shenzhen University)
Object DetectionAuto EncoderContrastive LearningImage
🎯 What it does: This paper proposes the AdaptFG model, which enhances feature diversity and avoids pseudo-label noise in sparse annotation object detection through adaptive feature generation, thereby improving detection performance.
AS-Det: Active Sampling for Adaptive 3D Object Detection in Point Clouds
Ziheng Ding (Fudan University), Rui Feng (Fudan University)
Object DetectionAutonomous DrivingPoint Cloud
🎯 What it does: This paper proposes an adaptive sampling single-stage 3D detector AS-Det, which can achieve high-precision detection on various point clouds (LiDAR and 4D radar).
ASER: Activation Smoothing and Error Reconstruction for Large Language Model Quantization
Weibo Zhao (Alibaba Cloud Computing), Yong Li (Alibaba Cloud Computing)
TransformerLarge Language ModelText
🎯 What it does: This study investigates the low-rank characteristics of low-bit quantization errors in LLMs and proposes the ASER algorithm, which achieves efficient post-training quantization through activation smoothing and error reconstruction.
ASP-Driven Emergency Planning for Norm Violations in Reinforcement Learning
Sebastian Adam (Vienna University of Technology), Thomas Eiter (Vienna University of Technology)
Reinforcement Learning
🎯 What it does: A framework based on Answer Set Programming (ASP) is proposed, which corrects the behavior of reinforcement learning (RL) agents by generating emergency action plans (policy fixes) when ethical violations are detected during execution, thus achieving compliance without retraining.
Aspect Enhancement and Text Simplification in Multimodal Aspect-Based Sentiment Analysis for Multi-Aspect and Multi-Sentiment Scenarios
Linlin Zhu (Xi'an Jiaotong University), Liang He (Xi'an Jiaotong University)
ClassificationTransformerVision Language ModelTextMultimodality
🎯 What it does: The AETS model is proposed, which simultaneously completes multi-faceted extraction, sentiment classification, and joint analysis in multimodal sentiment analysis through enhancement and text simplification techniques.
Assessing Modality Bias in Video Question Answering Benchmarks with Multimodal Large Language Models
Jean Park (University of Pennsylvania), Kevin Johnson (University of Pennsylvania)
TransformerLarge Language ModelVideoMultimodalityBenchmark
🎯 What it does: Proposes a Modal Importance Score (MIS) and uses a multimodal large language model to automatically assess the modal bias of video question-answering benchmarks.
Assessing Pre-Trained Models for Transfer Learning Through Distribution of Spectral Components
Tengxue Zhang (East China Normal University), Bin Yang (East China Normal University)
ClassificationObject DetectionDomain AdaptationConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: A pre-training model evaluation method based on the distribution of spectral component features (DISCO) is proposed, which uses SVD to decompose features into different spectral components and calculates the singular value ratios and their performance in downstream tasks, thereby quickly predicting model transfer performance.
Assessing the Creativity of LLMs in Proposing Novel Solutions to Mathematical Problems
Junyi Ye (New Jersey Institute of Technology), Guiling Wang (New Jersey Institute of Technology)
Large Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: A benchmark called CREATIVEMATH was established to evaluate the ability of LLMs to generate novel solutions to mathematical problems.
Association Pattern-enhanced Molecular Representation Learning
Lingxiang Jia (Zhejiang University), Zunlei Feng (Zhejiang University)
Representation LearningDrug DiscoveryGraph Neural NetworkTransformerGraph
🎯 What it does: An association pattern-aware information enhancement plugin, APMP, is proposed, which improves the atomic representations of existing molecular base models by sampling, filtering, and utilizing high-confidence paths for message passing in molecular graphs.
Asymmetric Cross-Modal Hashing Based on Formal Concept Analysis
Yinan Li (Central South University), Zhan Yang (Central South University)
RetrievalSupervised Fine-TuningMultimodality
🎯 What it does: This paper proposes an Asymmetric Cross-Modal Hashing Framework based on Formal Concept Analysis (ACHFCA), which maps different modal data into binary codes through flash projection three-layer semantic enhancement descriptors.
Asymmetric Hierarchical Difference-aware Interaction Network for Event-guided Motion Deblurring
Wen Yang (Xidian University), Guangming Shi (Xidian University)
RestorationOptical FlowImageVideo
🎯 What it does: An Asynchronous Hierarchical Differential Interaction Network (AHDINet) is proposed for event-driven motion deblurring.
Asymmetric Learning for Spectral Graph Neural Networks
Fangbing Liu (Australian National University), Qing Wang (Australian National University)
OptimizationGraph Neural NetworkGraph
🎯 What it does: An in-depth analysis of the optimization process of spectral graph convolutional networks is conducted, and a method to improve the optimization landscape through gradient asymmetric preprocessing is proposed.
Asymmetric Reinforcing Against Multi-Modal Representation Bias
Xiyuan Gao (Tianjin University), Qinghua Hu (Tianjin University)
Representation LearningContrastive LearningVideoMultimodality
🎯 What it does: A heterogeneous enhancement method based on mutual information, ARM, is designed to dynamically balance the contribution differences in multimodal learning and prevent dominant modality forgetting.
Asymmetric Visual Semantic Embedding Framework for Efficient Vision-Language Alignment
Yang Liu (Sichuan University), Jiancheng Lv (Sichuan University)
RetrievalTransformerVision Language ModelImageTextMultimodality
🎯 What it does: This paper proposes the Heterogeneous Visual-Semantic Embedding (AVSE) framework, which obtains multi-view image features through radial bias sampling and achieves efficient text-image matching using asymmetric embedding optimal matching.
Asymptotic Extinction in Large Coordination Games
Desmond Chan (King's College London), Carmine Ventre (King's College London)
Reinforcement Learning
🎯 What it does: This study investigates the exploration-exploitation trade-off of Q-Learning in large-scale multi-player coordination games, clarifying the relationship between the critical exploration rate and the number of players as well as the correlation of payoffs, and revealing the so-called 'asymptotic extinction' phenomenon.
Asymptotic Unbiased Sample Sampling to Speed Up Sharpness-Aware Minimization
Jiaxin Deng (Beijing University of Technology), Guodong Guo
ClassificationPose EstimationOptimizationComputational EfficiencyImage
🎯 What it does: An adaptive subsample sampling method based on gradient norm is proposed, significantly reducing the number of forward and backward passes during Sharpness-Aware Minimization (SAM) training and improving training speed.
AsyncDSB: Schedule-Asynchronous Diffusion Schrödinger Bridge for Image Inpainting
Zihao Han (Harbin Institute of Technology), Kolaye
RestorationDiffusion modelImage
🎯 What it does: This paper proposes an image restoration method based on asynchronous noise scheduling called AsyncDSB, which utilizes pixel gradient information to assign different noise time points to each pixel, achieving high-frequency pixel restoration first and low-frequency restoration later.
Asynchronous Distributed Gaussian Process Regression
Zewen Yang (Technical University of Munich), Sandra Hirche (Technical University of Munich)
Tabular
🎯 What it does: Proposes an asynchronous distributed Gaussian process regression method called AsyncDGP for real-time learning and safe control;
Asynchronous Federated Clustering with Unknown Number of Clusters
Yunfan Zhang (Guangdong University of Technology), Yiu-ming Cheung (Hong Kong Baptist University)
Federated LearningTabular
🎯 What it does: This paper proposes an Asynchronous Federated Clustering method (AFCL) that can automatically learn the global cluster distribution and complete cluster number estimation under conditions of unsynchronized communication, highly heterogeneous client data, and unknown number of clusters.
AtomNet: Designing Tiny Models from Operators Under Extreme MCU Constraints
Zhiwei Dong (University of Science and Technology Beijing), Xu-Cheng Yin (University of Science and Technology Beijing)
Object DetectionOptimizationComputational EfficiencyNeural Architecture SearchConvolutional Neural NetworkImage
🎯 What it does: Under the extreme resource constraints of MCUs (SRAM, Flash, latency), we propose AtomNet: first, we construct AtomDB through large-scale operator performance measurements, providing hardware-friendly design guidelines based on three-dimensional metrics of Flash, SRAM, and latency; then, we transfer the guidelines to block-level analysis, introducing a hybrid block pattern (SEP/ConvNext/MBConv) to address the peak SRAM usage in the early stages and the saturation of Flash in the later stages; on this hardware-oriented search space, we use zero-cost MACs cost search to obtain AtomNet; finally, we validate the deployment using INT8 quantization on the STM32 series MCUs with the official inference library.
Attack on Prompt: Backdoor Attack in Prompt-Based Continual Learning
Trang Nguyen (VinAI Research), Nhat Ho (The University of Texas at Austin)
Adversarial AttackPrompt EngineeringImage
🎯 What it does: Designed and validated a label-free, data-limited backdoor attack framework AOP against attacker vendors in prompt-based continual learning, demonstrating how to achieve efficient backdoor injection through prompt selection, static-dynamic trigger optimization, and BCE loss.
Attack-in-the-Chain: Bootstrapping Large Language Models for Attacks Against Black-Box Neural Ranking Models
Yu-An Liu (Chinese Academy of Sciences), Xueqi Cheng (Chinese Academy of Sciences)
RetrievalAdversarial AttackTransformerLarge Language ModelPrompt EngineeringTextChain-of-Thought
🎯 What it does: A chain reasoning framework based on large language models, Attack-in-the-Chain, is proposed for attacking neural ranking models under black-box conditions.
Attack-inspired Calibration Loss for Calibrating Crack Recognition
Zhuangzhuang Chen (Shenzhen University), Jianqiang Li (Shenzhen University)
ClassificationRecognitionAnomaly DetectionAdversarial AttackConvolutional Neural NetworkImage
🎯 What it does: Proposed an Adversarial Attack-based Calibration Loss (AICL) to achieve confidence calibration for crack recognition models;
AttackBench: Evaluating Gradient-based Attacks for Adversarial Examples
Antonio Emanuele Cinà (University of Genoa), Fabio Roli (University of Genoa)
Adversarial AttackImageBenchmark
🎯 What it does: The AttackBench benchmark framework is proposed to fairly evaluate gradient attacks under a unified model, dataset, and query budget, covering 102 implementations through experiments and constructing a complete leaderboard.
Attend and Enrich: Enhanced Visual Prompt for Zero-Shot Learning
Man Liu (Beijing Jiaotong University), Yao Zhao (National University of Singapore)
ClassificationRepresentation LearningTransformerPrompt EngineeringImage
🎯 What it does: AENet is proposed, which enhances visual representation in zero-shot learning through concept alignment attention and semantic enhancement prompts.
Attention Bootstrapping for Multi-Modal Test-Time Adaptation
Yusheng Zhao (Peking University), Ming Zhang (Peking University)
Domain AdaptationTransformerVideoMultimodalityAudio
🎯 What it does: This paper proposes a multi-modal adaptive method called ABPEM, which utilizes attention guidance and entropy minimization to achieve online adaptation of the model under unlabeled test data.
Attention-Driven GUI Grounding: Leveraging Pretrained Multimodal Large Language Models Without Fine-Tuning
Hai-Ming Xu (Australian Institute for Machine Learning), Lingqiao Liu (University of Wollongong)
RecognitionObject DetectionTransformerLarge Language ModelPrompt EngineeringImageTextMultimodality
🎯 What it does: A no-fine-tuning Attention-driven GUI Grounding (TAG) method is proposed, which achieves precise localization of GUI elements by leveraging the attention mechanism of a pre-trained multimodal large language model.
Attention-Imperceptible Backdoor Attacks on Vision Transformers
Zhishen Wang (Institute of Information Engineering, Chinese Academy of Sciences), Lihua Jing (Institute of Information Engineering, Chinese Academy of Sciences)
ClassificationAdversarial AttackTransformerImage
🎯 What it does: An attention-invisible backdoor attack method AIBA is proposed on Vision Transformer (ViT);
Attentive Eraser: Unleashing Diffusion Model’s Object Removal Potential via Self-Attention Redirection Guidance
Wenhao Sun (Zhejiang Gongshang University), Jingqun Tang (ByteDance Inc.)
RestorationGenerationDiffusion modelImage
🎯 What it does: Proposes a method called 'Attentive Eraser' that does not require fine-tuning for object removal in pre-trained diffusion models;
Attribute Inference Attacks for Federated Regression Tasks
Francesco Diana (Universite Cote d'Azur), Eoin Thomas (Amadeus)
Federated LearningSafty and PrivacyTabular
🎯 What it does: This paper proposes an Attribute Inference Attack (AIA) method for regression tasks within the framework of federated learning, exploring the feasibility of model-based attacks in this scenario.
Attribution Analysis Meets Model Editing: Advancing Knowledge Correction in Vision Language Models with VisEdit
Qizhou Chen (East China Normal University), Tingting Liu (Exacity Inc.)
Explainability and InterpretabilityKnowledge DistillationTransformerLarge Language ModelVision Language ModelImageTextMultimodality
🎯 What it does: This study investigates the impact of visual representations on text prediction in Vision-Language Large Models (VLLM) and proposes a new model editor, VisEdit, to correct knowledge errors in VLLM without retraining the entire model.
Attributive Reasoning for Hallucination Diagnosis of Large Language Models
Yuyan Chen (Fudan University), Yanghua Xiao (Zhejiang University)
TransformerLarge Language ModelText
🎯 What it does: This paper constructs a hallucination attribution framework based on internal states and proposes Differential Penalty Decoding (DPD) to reduce hallucinated outputs from LLMs.
Audio Entailment: Assessing Deductive Reasoning for Audio Understanding
Soham Deshmukh (Carnegie Mellon University), Bhiksha Raj (Microsoft Research)
ClassificationRecognitionTransformerLarge Language ModelPrompt EngineeringContrastive LearningMultimodalityAudio
🎯 What it does: This paper proposes the Audio Entailment task to evaluate the reasoning capabilities of audio-language models (ALM); it constructs two datasets, ACE and CLE, and benchmarks existing contrastive learning and next-word prediction ALMs under zero-shot and linear probe settings; it introduces the intermediate step 'caption-before-reason', which significantly enhances reasoning performance.
Audio-Visual Adaptive Fusion Network for Question Answering Based on Contrastive Learning
Xujian Zhao (Southwest University of Science and Technology), Peiquan Jin (University of Science and Technology of China)
RetrievalContrastive LearningVideoMultimodalityAudio
🎯 What it does: This paper proposes an audio-visual adaptive fusion network (AVAF-Net) based on contrastive learning for audio-visual question answering in dynamic music videos.
AudioGenX: Explainability on Text-to-Audio Generative Models
Hyunju Kang (Sungkyunkwan University), Hogun Park (NCSOFT)
GenerationData SynthesisExplainability and InterpretabilityTransformerTextAudio
🎯 What it does: AudioGenX is proposed - an explainable method based on factual and counterfactual reasoning to explain the impact of text inputs on audio generated by text-to-audio generation models.
Auditing and Enforcing Conditional Fairness via Optimal Transport
Mohsen Ghassemi (J.P.Morgan), Manuela Veloso (J.P.Morgan)
ClassificationOptimizationTabular
🎯 What it does: This paper proposes two optimal transport-based Conditional Demographic Disparity (CDD) metrics and designs two regularization methods, FairBiT and FairLeap, based on them, to regulate model performance while satisfying Conditional Demographic Parity (CDP).
Augmenting Math Word Problems via Iterative Question Composing
Haoxiong Liu (Institute for Interdisciplinary Information Sciences), Andrew C Yao (Institute for Interdisciplinary Information Sciences)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper enhances the mathematical reasoning ability of basic LLMs by constructing a new mathematical problem dataset called MMIQC and proposing the Iterative Question Composing (IQC) method.
Augmenting Online Algorithms for Knapsack Problem with Total Weight Information
Binghan Wu (University of Sydney), Bing Bing Zhou (University of Sydney)
Optimization
🎯 What it does: The paper proposes and analyzes three online knapsack algorithms that utilize total weight information: the Known Weight Algorithm (KWA) under exact weight, the Predicted Weight Algorithm (PWA) based on machine learning to predict weight, and the Limited Weight Algorithm (LWA) when the total weight is less than twice the capacity.
Augmenting Sequential Recommendation with Balanced Relevance and Diversity
Yizhou Dang (Northeastern University), Xingwei Wang (Northeastern University)
Recommendation SystemRecurrent Neural NetworkSequential
🎯 What it does: A pluggable enhancement plugin BASRec is proposed in sequential recommendation to generate new training samples while maintaining semantic relevance and introducing diversity.
AugRefer: Advancing 3D Visual Grounding via Cross-Modal Augmentation and Spatial Relation-based Referring
Xinyi Wang (University of Science and Technology of China), Xun Yang (University of Science and Technology of China)
RecognitionObject DetectionTransformerLarge Language ModelTextPoint Cloud
🎯 What it does: AugRefer is proposed in the 3D visual localization task, generating diverse text-3D pairs through cross-modal augmentation and designing a language-space adaptive decoder to enhance localization accuracy.
AUTE: Peer-Alignment and Self-Unlearning Boost Adversarial Robustness for Training Ensemble Models
Lifeng Huang (South China Agricultural University), Qiong Huang (South China Agricultural University)
OptimizationAdversarial AttackConvolutional Neural NetworkImage
🎯 What it does: A novel method for enhancing adversarial robustness, AUTE, is designed, which combines Peer-Alignment and Self-Unlearning techniques to train ensemble models.
Auto Encoding Neural Process for Multi-interest Recommendation
Yiheng Jiang (Jilin University), Chaozhuo Li (University of Macau)
Recommendation SystemAuto EncoderTabular
🎯 What it does: This paper proposes a multi-interest recommendation framework based on neural processes, NP-Rec, which utilizes distributed function learning to model users' diverse and dynamically changing preferences, and provides uncertainty estimates.
Auto-Regressive Diffusion for Generating 3D Human-Object Interactions
Zichen Geng (University of Western Australia), Ajmal Saeed Mian (Commonwealth Scientific and Industrial Research Organization)
GenerationData SynthesisRobotic IntelligenceDiffusion modelContrastive LearningPoint Cloud
🎯 What it does: This paper proposes an autoregressive diffusion model called ARDHOI, which achieves text-driven 3D human-object interaction sequence generation using continuous HOI labels.
Auto-Regressive Moving Diffusion Models for Time Series Forecasting
Jiaxin Gao (Shanghai Jiao Tong University), Yuntian Chen (Ningbo Institute of Digital Twin, Eastern Institute of Technology)
Diffusion modelTime SeriesFinance Related
🎯 What it does: The Auto-Regressive Moving Diffusion (ARMD) model is proposed, treating the future sequence of a time series as the initial state of a diffusion process and the historical sequence as the final state. It utilizes a sliding mechanism to generate continuous intermediate states, achieving unconditional continuous diffusion time series forecasting.
AutoFEA: Enhancing AI Copilot by Integrating Finite Element Analysis Using Large Language Models with Graph Neural Networks
Shifu Hou (University of Notre Dame), Yanfang Ye (University of Notre Dame)
AI Code AssistantGraph Neural NetworkTransformerLarge Language ModelGraphRetrieval-Augmented Generation
🎯 What it does: Integrate large language models with finite element analysis (FEA) to build the AutoFEA system, achieving automatic generation of FEA input files and executing simulations, thereby reducing AI hallucinations;
Automated Creation of Reusable and Diverse Toolsets for Enhancing LLM Reasoning
Zhiyuan Ma (University of Science and Technology of China), Xin Li (Tianjin University)
Large Language ModelAgentic AITextTabular
🎯 What it does: This paper proposes a two-stage knowledge-driven tool creation and evolution framework (KTCE) that can automatically generate reusable and diverse toolsets to enhance the reasoning capabilities of large language models.
Automatic Selection of Macro-Events for Heuristic-Search Temporal Planning
Alessandro La Farciola (Fondazione Bruno Kessler), Andrea Micheli (Fondazione Bruno Kessler)
OptimizationReinforcement LearningSequential
🎯 What it does: This study investigates the concept of 'macro-event' used in heuristic time planning and proposes a statistical ranking method for automatically selecting useful macro-events from a verified set of plans.
Automatically Generating Numerous Context-Driven SFT Data for LLMs Across Diverse Granularity
Shanghaoran Quan (Beihang University)
GenerationData SynthesisTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringContrastive LearningText
🎯 What it does: This paper proposes an automatic generation framework for multi-granularity context-driven SFT data—AUGCON.
AutoMMLab: Automatically Generating Deployable Models from Language Instructions for Computer Vision Tasks
Zekang Yang (SenseTime Research and Tetras.AI), Wentao Liu (SenseTime Research and Tetras.AI)
Object DetectionSegmentationHyperparameter SearchTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringImageTextBenchmark
🎯 What it does: The AutoMMLab platform has been built to realize a complete end-to-end automated process from natural language requests to deployable computer vision models.
Autonomous Goal Detection and Cessation in Reinforcement Learning: A Case Study on Source Term Estimation
Yiwei Shi (University of Bristol), Weiru Liu
Robotic IntelligenceReinforcement Learning
🎯 What it does: In the source term estimation (STE) problem, the self-supervised goal detection and stopping (AGDC) module is combined with various reinforcement learning (RL) algorithms (DQN, PPO, DDPG) to achieve automatic task completion recognition and search termination in environments lacking clear feedback signals.
Autonomous LLM-Enhanced Adversarial Attack for Text-to-Motion
Honglei Miao (Lanzhou University), Yi Yang (Zhejiang University)
Adversarial AttackRobotic IntelligenceTransformerLarge Language ModelContrastive LearningTextMultimodality
🎯 What it does: An adaptive attack framework called ALERT-Motion based on large language models is proposed to generate covert attack texts targeting text-to-motion (T2M) models.
Autonomous Option Invention for Continual Hierarchical Reinforcement Learning and Planning
Rashmeet Kaur Nayyar (Arizona State University), Siddharth Srivastava (Arizona State University)
Robotic IntelligenceReinforcement LearningSequential
🎯 What it does: A hierarchical reinforcement learning and planning framework called CHiRP has been developed, which can automatically invent, represent, and utilize symbolic options in a continual learning environment.
Autoregressive Sequence Modeling for 3D Medical Image Representation
Siwen Wang (University of Hong Kong), Yizhou Yu (Peking University)
ClassificationSegmentationRepresentation LearningTransformerContrastive LearningImageMultimodalityBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: Using autoregressive sequence modeling methods for representation learning of three-dimensional medical images;
AutoSciLab: A Self-Driving Laboratory for Interpretable Scientific Discovery
Saaketh Desai (Sandia National Laboratories), Prasad P. Iyer (Sandia National Laboratories)
OptimizationExplainability and InterpretabilityRobotic IntelligenceAuto EncoderTabularPhysics Related
🎯 What it does: AutoSciLab is a self-driven experimental platform that combines generative models, active learning, directional autoencoders, and neural network equation learners to automatically design experiments in high-dimensional experimental spaces, quickly locate effective experiments, extract relevant low-dimensional latent variables, and derive interpretable physical equations.
AutoSGNN: Automatic Propagation Mechanism Discovery for Spectral Graph Neural Networks
Shibing Mo (Xidian University), Jing Liu (Xidian University)
Graph Neural NetworkLarge Language ModelPrompt EngineeringGraph
🎯 What it does: This paper proposes AutoSGNN, a method for automatically generating spectral graph neural network propagation mechanisms using LLM and evolutionary strategies.
AWRaCLe: All-Weather Image Restoration Using Visual In-Context Learning
Sudarshan Rajagopalan (Johns Hopkins University), Vishal M. Patel (Johns Hopkins University)
RestorationTransformerVision Language ModelImage
🎯 What it does: This paper proposes the AWRaCLe framework, which utilizes visual context learning to achieve one-time restoration of images under various adverse weather conditions such as rain, snow, and fog.
B2Opt: Learning to Optimize Black-box Optimization with Little Budget
Xiaobin Li (Xidian University), Handing Wang (Xidian University)
OptimizationTransformerTabular
🎯 What it does: This study investigates how to solve high-dimensional and expensive black-box optimization problems through learning automated optimization strategies.
Backdoor Attack on Propagation-based Rumor Detectors
Di Jin (Tianjin University), Zhen Wang (Northwestern Polytechnical University)
Adversarial AttackGraph Neural NetworkGraph
🎯 What it does: This paper proposes an injection-based backdoor attack framework (IBAttack) for rumor detection models in propagation structures, achieving covert manipulation of the model by injecting trigger nodes at key points and adaptively generating features.
Backdoor Attacks Against No-Reference Image Quality Assessment Models via a Scalable Trigger
Yi Yu (Nanyang Technological University), Alex Kot (Nanyang Technological University)
Adversarial AttackConvolutional Neural NetworkImage
🎯 What it does: A backdoor attack model for no-reference image quality assessment (NR-IQA) is proposed, which injects noise in the DCT domain using a scalable trigger to achieve arbitrary control over the model output score.
Backdoor Token Unlearning: Exposing and Defending Backdoors in Pretrained Language Models
Peihai Jiang (Xidian University), Jing Ma (Xidian University)
Anomaly DetectionTransformerLarge Language ModelText
🎯 What it does: A method for actively detecting and eliminating backdoor triggers in pre-trained language models during the training phase is proposed—Backdoor Token Unlearning (BTU);
Bagging-Expert Network for Multi-Task Learning: A Depolarization Solution in Multi-Gate Mixture-of-Experts
Gong-Duo Zhang (Ant Group), Lin Zhou (Ant Group)
Recommendation SystemMixture of ExpertsTabular
🎯 What it does: This paper proposes the Bagging-Expert Network (BEnet) model to eliminate the expert weight polarization problem in the multi-gate mixture of experts framework for multi-task learning, enhancing the effectiveness of recommendation systems.
Balanced Adaptive Subspace Collaboration for Mixed Pareto-Lexicographic Multi-Objective Problems with Priority Levels
Wenjing Hong (Shenzhen University)
Optimization
🎯 What it does: A new Balanced Adaptive Subspace Collaboration (BASC) algorithm is proposed to address mixed Pareto-Lexicographic multi-objective optimization problems (PL-MPL-MOPs) with multiple priority levels. It generates new solutions by sampling within subspaces and dynamically balances exploration and exploitation across different priority layers during the search process.
Balanced and Fair Partitioning of Friends
Argyrios Deligkas (Royal Holloway University of London), Šimon Schierreich (Czech Technical University in Prague)
Graph
🎯 What it does: The study divides agents in a friend social graph into k groups of almost equal size, aiming to ensure that the allocation meets definitions of fairness (such as EF, EFX, PROP, MMS, etc.), and analyzes existence and computational complexity under different graph structures and utility functions.
Balancing Privacy and Performance: A Many-in-One Approach for Image Anonymization
Xuemei Jia (Wuhan University), Jun Chen (A*STAR)
RecognitionSafty and PrivacyKnowledge DistillationImage
🎯 What it does: This paper proposes an image anonymization method called FRO based on multi-instance mixing, which generates anonymized images that protect privacy while maintaining task performance by recombining feature fragments of different identities extracted from a knowledge pool, without adding noise.
Batch Ensemble for Variance Dependent Regret in Stochastic Bandits
Asaf Cassel (Tel Aviv University), Yishay Mansour (Google Research)
OptimizationReinforcement Learning from Human Feedback
🎯 What it does: This paper proposes a Batch Ensemble algorithm that constructs an optimistic loss estimate by dividing the samples of each arm into several batches and taking the minimum of the batch means, thereby achieving a balance between exploration and exploitation in multi-armed bandits (MAB);
Batch Selection for Multi-Label Classification Guided by Uncertainty and Dynamic Label Correlations
Ao Zhou (Chongqing University of Posts and Telecommunications), Grigorios Tsoumakas (Aristotle University of Thessaloniki)
ClassificationImageVideoText
🎯 What it does: A multi-label batch selection method based on uncertainty is proposed, which evaluates label uncertainty using a combination of sliding window differences and entropy, and further measures sample uncertainty through dynamic label associations, thereby selecting more informative samples during training.
Battling the Non-stationarity in Time Series Forecasting via Test-time Adaptation
HyunGi Kim (Seoul National University), Sungroh Yoon (Seoul National University)
TransformerTime SeriesBenchmark
🎯 What it does: This paper proposes a test-time adaptive framework for non-stationary time series forecasting, TAFAS, which can actively utilize partially observed true values to adjust the pre-trained model during the inference phase.
BayesCNS: A Unified Bayesian Approach to Address Cold Start and Non-Stationarity in Search Systems at Scale
Randy Ardywibowo (Apple), Sankalp Nayak (Apple)
Recommendation SystemReinforcement LearningTabular
🎯 What it does: To address the cold start and non-stationarity issues in search systems, a unified Bayesian online learning framework called BayesCNS is proposed to estimate and continuously update the prior distribution of user-item interactions, thereby enhancing the display effectiveness of new items and the overall success rate of the system.
Bayesian Low-Rank Learning (Bella): A Practical Approach to Bayesian Neural Networks
Bao Gia Doan (University of Adelaide), Ehsan Abbasnejad (Idiap Research Institute)
ClassificationOptimizationComputational EfficiencyTransformerVision Language ModelImage
🎯 What it does: This paper proposes a Bayesian low-rank learning framework named Bella, which utilizes low-rank adapters to generate multi-particle approximations on pre-trained models, thereby achieving efficient inference for Bayesian neural networks.
Bayesian Optimization for Unknown Cost-Varying Variable Subsets with No-Regret Costs
Vu Viet Hoang (FPT Software), Hung The Tran (Hanoi University of Science and Technology)
OptimizationTabular
🎯 What it does: This paper proposes a Bayesian Optimization with Cost-varying Variable Subsets (BOCVS) algorithm aimed at unknown and random costs, employing an exploration-exploitation phase strategy. In each iteration, the algorithm estimates the objective function using Gaussian processes and combines UCB/LCB to determine the optimal control subset and its values.
Bayesian Persuasion with Externalities: Exploiting Agent Types
Jonathan Shaki (Bar-Ilan University), Sarit Kraus (Bar-Ilan University)
OptimizationReinforcement Learning
🎯 What it does: In the multi-agent Bayesian persuasion framework with externality effects, the author provides methods for constructing and solving optimal strategies for three types of information transmission channels: public, semi-private, and private.
BearLLM: A Prior Knowledge-Enhanced Bearing Health Management Framework with Unified Vibration Signal Representation
Haotian Peng (Shenyang Institute of Automation, Chinese Academy of Sciences), Wei Wang (Shenyang Institute of Automation, Chinese Academy of Sciences)
Anomaly DetectionTransformerLarge Language ModelSupervised Fine-TuningMultimodalityTime Series
🎯 What it does: This paper presents BearLLM, an integrated multi-task bearing health management framework that achieves anomaly detection, fault diagnosis, maintenance recommendations, and risk analysis through a unified vibration signal representation and large language models.
BEE: Metric-Adapted Explanations via Baseline Exploration-Exploitation
Oren Barkan (Open University), Noam Koenigstein (Tel Aviv University)
Explainability and InterpretabilityConvolutional Neural NetworkTransformerReinforcement LearningImage
🎯 What it does: A method named BEE is proposed, which is based on baseline exploration-exploitation for path integral interpretation. It automatically learns and samples baseline distributions suitable for specific evaluation metrics, thereby generating explanation maps that better align with the metrics.
BeFA: A General Behavior-driven Feature Adapter for Multimedia Recommendation
Qile Fan (Nanjing University of Posts and Telecommunications), Guanming Lu (Nanjing University of Posts and Telecommunications)
Recommendation SystemTransformerContrastive LearningMultimodality
🎯 What it does: This paper proposes a visual attribution analysis method to diagnose the defects of content features in multimodal recommendation, and designs a general behavior-driven feature adapter (BeFA) based on behavioral information to reconstruct content features.
Behavior Importance-Aware Graph Neural Architecture Search for Cross-Domain Recommendation
Chendi Ge (Tsinghua University), Wenwu Zhu (Tsinghua University)
Recommendation SystemNeural Architecture SearchGraph Neural NetworkGraph
🎯 What it does: The BiGNAS framework is proposed, which jointly optimizes the GNN architecture and the importance of source domain behavior in cross-domain recommendation.
Behaviour Preference Regression for Offline Reinforcement Learning
Padmanaba Srinivasan (Imperial College London), William Knottenbelt (Imperial College London)
Reinforcement LearningTabular
🎯 What it does: An offline reinforcement learning method based on Behavior Preference Regression (BPR) is proposed, which directly learns the policy through least squares regression, maximizing rewards while maintaining consistency with the behavior policy.