AAAI 2025 Papers — Page 6
AAAI Conference on Artificial Intelligence · 3028 papers
Compose with Me: Collaborative Music Inpainter for Symbolic Music Infilling
Zhejing Hu (Hong Kong Polytechnic University), Bruce X.B. Yu (Zhejiang University University of Illinois Urbana-Champaign Institute)
GenerationTransformerContrastive LearningSequentialAudio
🎯 What it does: A collaborative music filling model CMI is designed, implementing a Human-in-the-Loop (HITL) framework for symbolic music filling, capable of local filling and iterative improvement based on user-marked missing segments.
ComprehendEdit: A Comprehensive Dataset and Evaluation Framework for Multimodal Knowledge Editing
Yaohui Ma (Harbin Institute of Technology), Zhiheng Ma (Shenzhen Institutes of Advanced Technology)
Knowledge DistillationTransformerLarge Language ModelVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Proposed the ComprehendEdit benchmark and two new evaluation metrics, KGI/KPI, and designed a Hierarchical Context Editing (HICE) method to improve knowledge editing in multimodal large language models.
Comprehensive Multi-Modal Prototypes Are Simple and Effective Classifiers for Vast-Vocabulary Object Detection
Yitong Chen (Fudan University), Yu-Gang Jiang (Fudan University)
Object DetectionTransformerContrastive LearningImageTextMultimodality
🎯 What it does: Prova is proposed, a multimodal prototype classifier for large vocabulary object detection, which can significantly improve detection accuracy.
Compress to One Point: Neural Collapse for Pre-Trained Model-Based Class-Incremental Learning
Kun Wei (Xidian University), Cheng Deng (Xidian University)
ClassificationTransformerSupervised Fine-TuningImage
🎯 What it does: This paper proposes an incremental learning method based on pre-trained models (PTMCIL), which achieves feature adaptation and inter-class discrimination for different tasks by freezing the Equiangular Tight Frame (ETF) classifier and combining task-related adaptation, feature compression, and structural alignment.
Compressing Streamable Free-Viewpoint Videos to 0.1 MB per Frame
Luyang Tang (Peking University), Ronggang Wang (Peking University)
CompressionGaussian SplattingVideo
🎯 What it does: A streaming FVV compression framework iFVC based on 3D Gaussians is proposed, capable of online training, real-time rendering, and low storage costs;
Computationally Hard Problems Are Hard for QBF Proof Systems Too
Agnes Schleitzer (Friedrich Schiller University Jena), Olaf Beyersdorff (Friedrich Schiller University Jena)
🎯 What it does: A general method is proposed to transform computationally difficult problems in polynomial hierarchies (such as Succinct k-Radius, k-Clique Colouring, ALL-EQUAL ∃∀ 3 SAT) into QBF, and the corresponding QBF forms are provided. Using this method, three families of QBF are constructed, and it is proven that these QBFs require exponential proof lengths in the Q-Res and QU-Res proof systems.
Computing Game Symmetries and Equilibria That Respect Them
Emanuel Tewolde (Carnegie Mellon University), Vincent Conitzer (Carnegie Mellon University)
Optimization
🎯 What it does: The study investigates the problem of identifying and utilizing symmetry in normal form games, establishes its equivalence with graph automorphism and graph isomorphism problems, and analyzes the complexity of solving Nash equilibria under symmetry constraints.
Computing Perfect Bayesian Equilibria in Sequential Auctions with Verification
Vinzenz Thoma (ETH Zurich), Sven Seuken (University of Zurich)
OptimizationReinforcement LearningSequential
🎯 What it does: This paper proposes an algorithm for calculating pure strategy ε-perfect Bayesian equilibrium (ε-PBE) in sequential auctions under continuous action and value spaces, incorporating a verification phase to assess strategy errors.
CoMT: A Novel Benchmark for Chain of Multi-modal Thought on Large Vision-Language Models
Zihui Cheng (Central South University), Libo Qin (Harbin Institute of Technology)
TransformerLarge Language ModelVision Language ModelImageTextMultimodalityBenchmarkChain-of-Thought
🎯 What it does: The CoMT benchmark is proposed to evaluate the performance of large visual-language models in chain-based multimodal reasoning, requiring outputs that include both textual and visual (image) reasoning steps.
Concept Conductor: Orchestrating Multiple Personalized Concepts in Text-to-Image Synthesis
Zebin Yao (Beijing University of Posts and Telecommunications), Xiaojie Wang (Beijing University of Posts and Telecommunications)
GenerationData SynthesisTransformerDiffusion modelImageText
🎯 What it does: A training-free framework named Concept Conductor is designed to address the issues of attribute leakage and layout confusion in multi-concept text-to-image generation.
Concept Matching with Agent for Out-of-Distribution Detection
Yuxiao Lee (Jilin University), Yi Chang (Jilin University)
Anomaly DetectionTransformerAgentic AIVision Language ModelContrastive LearningImage
🎯 What it does: Proposes the Concept Matching and Agent (CMA) method in an untrained zero-shot environment, using neutral text agents to construct a triangular relationship that improves OOD detection performance.
ConceptSearch: Towards Efficient Program Search Using LLMs for Abstraction and Reasoning Corpus (ARC)
Kartik Singhal (Indraprastha Institute of Information Technology Delhi), Gautam Shroff (Indraprastha Institute of Information Technology Delhi)
OptimizationComputational EfficiencyAI Code AssistantConvolutional Neural NetworkTransformerLarge Language ModelContrastive LearningText
🎯 What it does: A concept-based function search algorithm called ConceptSearch is proposed, which utilizes LLM to generate programs and guides the search through concept scoring to solve the ARC task.
Concurrent Planning and Execution in Lifelong Multi-Agent Path Finding with Delay Probabilities
Yue Zhang (Monash University), Peter J. Stuckey (Monash University)
OptimizationRobotic Intelligence
🎯 What it does: This paper proposes a new parallel planning and execution framework, PIE-D, to maintain synchronization between planning and execution and maximize throughput in the lifelong multi-agent pathfinding (LMAPF) problem with execution delays.
Conditional Diffusion Models Based Conditional Independence Testing
Yanfeng Yang (East China Normal University), Renming Zhang (Boston University)
Diffusion modelScore-based ModelTabular
🎯 What it does: A conditional independence testing method based on conditional diffusion models is proposed, utilizing a conditional randomization testing framework to generate approximate samples of $X|Z$ and estimating conditional mutual information with a classifier.
Conditional Feature Importance with Generative Modeling Using Adversarial Random Forests
Kristin Blesch (Leibniz Institute for Prevention Research and Epidemiology), Marvin N. Wright
Explainability and InterpretabilityAdversarial AttackGenerative Adversarial NetworkTabularTime Series
🎯 What it does: The cARFi method is proposed, which utilizes Adversarial Random Forests (ARF) to evaluate conditional feature importance for any subset of features without model fitting and with low hyperparameter tuning.
Conditional Latent Coding with Learnable Synthesized Reference for Deep Image Compression
Siqi Wu (University of Missouri), Zhihai He (University of Science and Technology of China)
CompressionConvolutional Neural NetworkTransformerImage
🎯 What it does: A conditional latent coding (CLC) framework based on an external image dictionary is proposed, which dynamically generates latent references to achieve deep image compression.
ConDo: Continual Domain Expansion for Absolute Pose Regression
Zijun Li (Xiamen University), Cheng Wang (Xiamen University)
Pose EstimationDomain AdaptationKnowledge DistillationTransformerSimultaneous Localization and MappingImage
🎯 What it does: This paper proposes the Continual Domain Expansion (ConDo) method, which continuously updates the absolute pose regression model using unlabeled inference images after deployment, and obtains supervisory signals from scene-independent localization methods through knowledge distillation, constructing a large-scale benchmark aimed at long-term environmental changes.
ConDSeg: A General Medical Image Segmentation Framework via Contrast-Driven Feature Enhancement
Mengqi Lei (China University of Geosciences), Xin Wang (Baidu Inc)
SegmentationConvolutional Neural NetworkTransformerImageBiomedical Data
🎯 What it does: A general medical image segmentation framework named ConDSeg is proposed, addressing soft boundaries and co-occurrence phenomena, and improving segmentation quality in low-light and low-contrast environments.
Confidence Estimation for Error Detection in Text-to-SQL Systems
Oleg Somov (Artificial Intelligence Research Institute), Elena Tutubalina (Sber Artificial Intelligence)
ClassificationAnomaly DetectionTransformerLarge Language ModelTextBiomedical DataElectronic Health Records
🎯 What it does: Proposed the incorporation of a rejectable selective classifier in the Text-to-SQL system to leverage model uncertainty for detecting erroneous generations and unanswerable questions.
Configuring Data Augmentations to Reduce Variance Shift in Positional Embedding of Vision Transformers
Bum Jun Kim (Pohang University of Science and Technology), Sang Woo Kim (Pohang University of Science and Technology)
ClassificationSegmentationTransformerImage
🎯 What it does: This paper studies the issue of variance shift in position encoding when using data augmentation during the training of Vision Transformer (ViT), and provides corresponding configuration suggestions.
ConfigX: Modular Configuration for Evolutionary Algorithms via Multitask Reinforcement Learning
Hongshu Guo (South China University of Technology), Yue-Jiao Gong (Singapore Management University)
OptimizationTransformerReinforcement LearningTabular
🎯 What it does: A MetaBBO framework named ConfigX has been constructed, capable of learning a universal configuration model for automatically configuring parameters across various evolutionary algorithms and black-box optimization problems.
Conformal Inference of Individual Treatment Effects Using Conditional Density Estimates
Baozhen Wang (Binghamton University), Xingye Qiao (Binghamton University)
TabularSequential
🎯 What it does: This paper proposes a new method for interval inference of individual treatment effects (ITE) using conditional density estimation.
Conformal Prediction for Partial Label Learning
Xiuwen Gong (University of Technology Sydney), Guandong Xu (Education University of Hong Kong)
ClassificationSupervised Fine-TuningImage
🎯 What it does: A CP-PLL method is proposed to quantify model uncertainty and provide reliable confidence guarantees within the framework of Partial Label Learning (PLL) using Conformal Prediction.
Conformal Thresholded Intervals for Efficient Regression
Rui Luo (City University of Hong Kong), Zhixin Zhou (Alpha Benito Research)
Tabular
🎯 What it does: This paper proposes a new conformal prediction method - Conformal Thresholded Intervals (CTI), which estimates the length of conditional interval ranges through multi-output quantile regression and constructs a minimal prediction set based on a threshold to ensure coverage.
Conformalized Interval Arithmetic with Symmetric Calibration
Rui Luo (City University of Hong Kong), Zhixin Zhou (Alpha Benito Research)
TabularTime Series
🎯 What it does: This paper studies a new method for predicting confidence intervals for the sum/average of unknown label sets—Symmetric Calibration of Synthetic Interval Arithmetic (CIA).
Confounding-Robust Deferral Policy Learning
Ruijiang Gao (University of Texas at Dallas), Mingzhang Yin (University of Florida)
Recommendation SystemOptimizationReinforcement Learning from Human FeedbackReinforcement LearningTabularBiomedical DataFinance Related
🎯 What it does: This paper studies strategy learning in human-machine collaboration in the presence of unobserved confounding and proposes the 'Confounding-Robust Deferral Policy Learning' method.
ConFREE: Conflict-free Client Update Aggregation for Personalized Federated Learning
Hao Zheng (Central South University), Boyu Wang (Central South University)
Federated LearningImage
🎯 What it does: Proposes the ConFREE method, which constructs conflict elimination guiding vectors in personalized federated learning and seeks optimal updates within their neighborhoods to achieve global aggregation without conflicts, thereby reducing negative transfer.
ConSense: Continually Sensing Human Activity with WiFi via Growing and Picking
Rong Li (Soochow University), Juncheng Jia (Soochow University)
RecognitionTransformerTime Series
🎯 What it does: A lightweight, example-free incremental learning framework called ConSense is proposed, specifically for continuous human activity recognition using WiFi signals.
Consistency of Compositional Generalization Across Multiple Levels
Chuanhao Li (Beijing Institute of Technology), Yunde Jia (Shenzhen MSU-BIT University)
OptimizationMeta LearningLarge Language ModelVideoText
🎯 What it does: A multi-layer optimization framework based on meta-learning is proposed, which gradually trains the model in a sequence from simple to complex by introducing multiple meta-weight networks to achieve consistent combinatorial generalization across multiple levels (phrase-phrase, phrase-word, word-word).
Consistent Query Answering over Existential Rules with Open and Closed Predicates
Lorenzo Marconi (Sapienza University of Rome), Riccardo Rosati (Sapienza University of Rome)
🎯 What it does: Under the existence rules, the problem of consistent query answers (CQA) in knowledge bases with both open and closed predicates is studied, a new framework is proposed, and the data complexity under different classes of rules is analyzed.
Constant-Factor Distortion Mechanisms for k-Committee Election
Haripriya Pulyassary (Cornell University), Chaitanya Swamy (University of Waterloo)
Optimization
🎯 What it does: This paper studies the k-committee election problem in a given metric space, using only a limited number of value queries to minimize the Top ℓ cost (i.e., the sum of distances of the top ℓ voters), and proposes a mechanism that can achieve constant factor distortion based on this.
Constrained Fair and Efficient Allocations
Benjamin Cookson (University of Toronto), Nisarg Shah (University of Toronto)
Optimization
🎯 What it does: This paper studies how the maximum Nash welfare (MNW) allocation can simultaneously satisfy the fairness efficiency properties of 1/2-EF1 and PO under various feasibility constraints (such as arbitrary matroids, base-exchangeable matroids, partition matroids with lower bounds, goods replication, and equilibrium constraints), and proposes a randomized algorithm to achieve the optimal 'both-and' scheme (BoBW) under these constraints.
Constrained Generative Modeling with Manually Bridged Diffusion Models
Saeid Naderiparizi (University of British Columbia), Frank Wood (University of British Columbia)
GenerationData SynthesisAutonomous DrivingTransformerDiffusion modelTabularStochastic Differential Equation
🎯 What it does: A new 'manual bridge' framework is proposed for generating in constrained spaces within diffusion models, achieving multi-constraint models through the combination of multiple constraints.
Constrained Offline Black-Box Optimization via Risk Evaluation and Management
Yiyi Zhu (East China Normal University), Hong Qian (East China Normal University)
OptimizationSupervised Fine-TuningContrastive LearningTabular
🎯 What it does: The COOREM method is proposed to address the dual challenges of OOD and constraint risks in constrained offline black-box optimization, achieving dynamic assessment and management of risks.
Constraint-Adaptive Policy Switching for Offline Safe Reinforcement Learning
Yassine Chemingui (Washington State University), Jana Doppa (Washington State University)
Safty and PrivacyReinforcement LearningTabularBenchmark
🎯 What it does: A framework called Constraint-Adaptive Policy Switching (CAPS) is proposed for offline safe reinforcement learning, which can dynamically switch between multiple strategies based on different cost thresholds during deployment.
Constructing Fair Latent Space for Intersection of Fairness and Explainability
Hyungjun Joo (Seoul National University), Jungwoo Lee (Seoul National University)
GenerationExplainability and InterpretabilityDiffusion modelAuto EncoderImage
🎯 What it does: Add reversible network modules to pre-trained generative models to construct a fair latent space, achieving fair interpretability for each sample.
Content-aware Balanced Spectrum Encoding in Masked Modeling for Time Series Classification
Yudong Han (Shandong University), Weili Guan
ClassificationRepresentation LearningTransformerTime SeriesElectrocardiogram
🎯 What it does: This paper proposes a dual-branch masked time series modeling framework, incorporating a Content-Aware Balanced Spectrum Decoder (CBD) to enhance the quality of time series representation learning.
Content-free Logical Modification of Large Language Model by Disentangling and Modifying Logic Representation
Xin Wu (South China University of Technology), Yi Cai (South China University of Technology)
RecognitionGenerationTransformerLarge Language ModelContrastive LearningText
🎯 What it does: A logical control framework (LCF) is proposed, which splits the hidden layers of LLM into content and logic spaces, and transfers representations to valid areas in the logic space through contrastive learning, thereby enhancing the logical correctness of generated conclusions.
Context-aware Graph Neural Network for Graph-based Fraud Detection with Extremely Limited Labels
Pengbo Li (Shanghai University), Xiangfeng Luo (Shanghai University)
Recommendation SystemAnomaly DetectionOptimizationGraph Neural NetworkContrastive LearningGraphFinance Related
🎯 What it does: This paper proposes a Context-aware Graph Neural Network (CGNN) that utilizes graph-structured data with very few labels for fraud detection, enhancing model performance through semantic splitting of neighbor features, denoising attention, and consistency regularization.
Context-aware Inductive Knowledge Graph Completion with Latent Type Constraints and Subgraph Reasoning
Muzhi Li (Chinese University of Hong Kong), Irwin King (Chinese University of Hong Kong)
Graph Neural NetworkLarge Language ModelSupervised Fine-TuningPrompt EngineeringGraph
🎯 What it does: This paper proposes a context-aware knowledge graph completion method based on large language models called CATS.
ContextHOI: Spatial Context Learning for Human-Object Interaction Detection
Mingda Jia (Peking University), Yun Zheng (Alibaba Group)
Object DetectionTransformerVision Language ModelImage
🎯 What it does: A dual-branch framework called ContextHOI is proposed, which learns scene context and instance features through context and instance branches respectively, and integrates them for human-object interaction detection; introduces spatial contrast constraints and semantically guided context exploration, utilizing VLM priors to enhance context learning; constructs a new HICO-DET (ambiguous) subset to evaluate the robustness of the model under unclear instance conditions.
Contextual Structure Knowledge Transfer for Graph Neural Networks
Zhiyuan Yu (Nanjing University), Sanglu Lu (Fudan University)
ClassificationDomain AdaptationGraph Neural NetworkGraph
🎯 What it does: To address the issue of homogeneity shift in graph transfer learning, we propose the Contextual Structural Graph Neural Network (CS-GNN), which achieves cross-domain structural knowledge transfer and node classification by constructing local EGO networks, quantifying structural diversity based on feature smoothing moment, employing a contextual attention mechanism, and implementing group fairness loss.
ContextualStory: Consistent Visual Storytelling with Spatially-Enhanced and Storyline Context
Sixiao Zheng (Fudan University), Yanwei Fu (Fudan University)
GenerationTransformerDiffusion modelImageVideo
🎯 What it does: Proposes the ContextualStory framework for visual storytelling, capable of generating coherent story frames and supporting story continuation.
Continual Learning Using a Kernel-Based Method Over Foundation Models
Saleh Momeni (University of Illinois Chicago), Bing Liu (University of Illinois Chicago)
ClassificationDomain AdaptationSupervised Fine-TuningImageText
🎯 What it does: Utilizing fixed base model features, combined with kernel functions and random Fourier features, class incremental learning is achieved through linear discriminant analysis as new tasks are continuously added.
Continual Unsupervised Generative Modelling via Online Optimal Transport
Fei Ye (University of Electronic Science and Technology of China), Kun Zhang (Carnegie Mellon University)
GenerationData SynthesisDiffusion modelImage
🎯 What it does: This paper proposes an online unsupervised generative modeling framework that utilizes Sinkhorn distance-driven dual dynamic memory (SDDM) to select and maintain short-term and long-term samples, thereby achieving a task-free generative model in continual learning.
Contract-based Design and Verification of Multi-Agent Systems with Quantitative Temporal Requirements
Rafael Dewes (CISPA Helmholtz Center for Information Security), Rayna Dimitrova (CISPA Helmholtz Center for Information Security)
TabularBenchmark
🎯 What it does: This paper proposes a contract-based framework for the hierarchical design and verification of multi-agent systems (MAS) under quantitative temporal requirements (LTL[F]); it achieves the decomposition of overall quantitative specifications, collaborative constraints, and best-effort satisfaction through the definition of Good-Enough Deconstructive Contracts (GEDC); an automated verification method and implementation based on GNBA are also provided.
Contradicted in Reliable, Replicated in Unreliable: Dual-Source Reference for Fake News Early Detection
Yifan Feng (Shanghai University), Zhongming Han (Beijing Technology and Business University)
ClassificationAnomaly DetectionTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: The RUSR method is proposed, which evaluates support and semantic heterogeneity by constructing publication backgrounds of credible and non-credible sources, achieving early truth determination of news.
Contrasting Adversarial Perturbations: The Space of Harmless Perturbations
Lu Chen (Shanghai Jiao Tong University), Yuan Luo (Shanghai Jiao Tong University)
Adversarial AttackConvolutional Neural NetworkImage
🎯 What it does: The paper proposes and validates the concept of 'harmless perturbation', which states that there exists a continuous perturbation subspace in deep neural networks, where any perturbation from this space, regardless of its magnitude, will not change the network output.
Contrastive Auxiliary Learning with Structure Transformation for Heterogeneous Graphs
Wei Du (Jilin University), Ying Li (Jilin University)
Representation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: This paper proposes an auxiliary contrastive learning model CALHG that combines edge perturbation, graph diffusion, and category-guided multi-view contrastive learning to learn node representations on heterogeneous graphs with no or few features.
Contrastive Functional Principal Component Analysis
Eric Zhang (University of North Carolina at Chapel Hill), Didong Li (University of North Carolina at Chapel Hill)
Contrastive LearningTime SeriesFinance Related
🎯 What it does: Proposes Contrastive Functional Principal Component Analysis (CFPCA) to identify low-dimensional structures unique to the foreground group relative to the background group, reducing common variation;
Contrastive Multi-view Subspace Clustering via Tensor Transformers Autoencoder
Qianqian Wang (Xidian University), Quanxue Gao (Xidian University)
TransformerAuto EncoderContrastive LearningImage
🎯 What it does: A multi-view subspace clustering method based on tensor transformer autoencoders (TTAE) is proposed, achieving view complementarity and discriminative feature learning through cross-view attention, autoencoders, contrastive learning, and self-expression layers;
Contrastive Representation for Interactive Recommendation
Jingyu Li (Tianjin University), Guoli Wu (Tianjin University)
Recommendation SystemReinforcement LearningContrastive Learning
🎯 What it does: This paper proposes an interactive recommendation framework CRIR based on contrastive learning, which utilizes user state representations generated by interest weights and enhances representation quality through Preference Ranking Contrastive Learning (PRCL), thereby improving the sample efficiency of the DRL recommender.
Controllable 3D Dance Generation Using Diffusion-Based Transformer U-Net
Puyuan Guo (Beijing University of Posts and Telecommunications), Ya Li (Beijing University of Posts and Telecommunications)
GenerationData SynthesisPose EstimationTransformerDiffusion modelVideo
🎯 What it does: A controllable 3D dance generation framework based on diffusion models is proposed, with the core being the Transformer-U-Net network. A control network is built on this foundation, allowing the generation of 3D dance to be guided by 2D keypoint sequences.
Controllable Distortion-Perception Tradeoff Through Latent Diffusion for Neural Image Compression
Chuqin Zhou (Shanghai Jiao Tong University), Wenjun Zhang (Shanghai Jiao Tong University)
CompressionDiffusion modelAuto EncoderImage
🎯 What it does: This paper proposes a pluggable adaptive latent fusion module that can achieve controllable reconstruction with a distortion-perception balance at the decoding end by adjusting parameters while keeping the original bitstream unchanged.
Controllable Protein Sequence Generation with LLM Preference Optimization
Xiangyu Liu (Nanjing University), Wei Hu (Nanjing University)
GenerationOptimizationDrug DiscoveryTransformerLarge Language ModelSupervised Fine-TuningTextBiomedical Data
🎯 What it does: The CtrlProt method is proposed, achieving controllable protein sequence generation through prefix fine-tuning of a pre-trained protein LLM and multi-list priority optimization.
Controlling Equational Reasoning in Large Language Models with Prompt Interventions
Jordan Meadows (University of Manchester), André Freitas (Idiap Research Institute)
TransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: A 30k fine-grained equation derivation dataset was constructed using a symbolic data generation framework, and large language models were fine-tuned and evaluated through various prompt interventions (variable renaming, expression swapping, target replacement, step deletion) to explore the impact of interventions on the distribution of reasoning errors.
Controlling Large Language Models Through Concept Activation Vectors
Hanyu Zhang (Chinese Academy of Sciences), Qing He (Chinese Academy of Sciences)
TransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: The GCAV framework is proposed, which modifies the activation layer during the inference phase using Concept Activation Vectors (CAV) to achieve fine-grained and quantifiable control over LLM outputs.
Convergence Analysis of Federated Learning Methods Using Backward Error Analysis
Jinwoo Lim (Seoul National University), Soo-Mook Moon (Seoul National University)
OptimizationFederated LearningImage
🎯 What it does: This paper utilizes backward error analysis to theoretically derive and empirically validate the implicit regularization of federated learning algorithms (FedAvg, FedSAM, SCAFFOLD) under non-IID distributions.
Convergence Rate in a Nonlinear Two-Time-Scale Stochastic Approximation with State (Time)-Dependence
Zixi Chen (Peking University), Ruixun Zhang (Peking University)
OptimizationStochastic Differential Equation
🎯 What it does: This paper studies the convergence rate of two-time-scale stochastic approximation under state or time-dependent noise conditions, proving that polynomial or even exponential convergence can be achieved when the noise variance decreases with state/time.
ConVis: Contrastive Decoding with Hallucination Visualization for Mitigating Hallucinations in Multimodal Large Language Models
Yeji Park (Sogang University), Buru Chang (Korea University)
GenerationData SynthesisTransformerLarge Language ModelVision Language ModelContrastive LearningImageTextMultimodalityBenchmark
🎯 What it does: A training-free contrastive decoding method called ConVis is proposed, which visualizes hallucinations in text-to-image models and suppresses the generation of multimodal LLM hallucinations during decoding by contrasting the distribution of the original image with the reconstructed image.
Cooperative Policy Agreement: Learning Diverse Policy for Offline MARL
Yihe Zhou (Zhejiang University), Shunyu Liu (Nanyang Technological University)
Reinforcement Learning
🎯 What it does: This paper proposes an offline multi-agent reinforcement learning method called Cooperative Policy Agreement (CPA) to address the issue of policy mismatch caused by data from different sources and to enhance policy diversity.
CoPEFT: Fast Adaptation Framework for Multi-Agent Collaborative Perception with Parameter-Efficient Fine-Tuning
Quanmin Wei (Southwest Jiaotong University), Xiao Wu (Wuhan University of Technology)
Object DetectionDomain AdaptationComputational EfficiencyConvolutional Neural NetworkSupervised Fine-TuningPrompt EngineeringPoint Cloud
🎯 What it does: A lightweight CoPEFT framework is proposed to achieve rapid adaptation of trained multi-agent collaborative perception models in newly deployed environments.
CoPRA: Bridging Cross-domain Pretrained Sequence Models with Complex Structures for Protein-RNA Binding Affinity Prediction
Rong Han (Tsinghua University), Ting Chen (Tsinghua University)
Drug DiscoveryProtein Structure PredictionTransformerContrastive LearningBiomedical Data
🎯 What it does: This study proposes the CoPRA model, which integrates protein language models, RNA language models, and the three-dimensional structural information of protein-RNA complexes to predict protein-RNA binding affinity.
CoRA: Collaborative Information Perception by Large Language Model’s Weights for Recommendation
Yuting Liu (Northeastern University), Xingwei Wang (Northeastern University)
Recommendation SystemTransformerLarge Language ModelTabular
🎯 What it does: Proposes the CoRA framework, which integrates collaborative filtering information in the LLM parameter space to achieve personalized recommendations without fine-tuning and without additional tokens.
Core Knowledge Learning Framework for Graph
Bowen Zhang (Shenzhen Technology University), Hu Huang (University of Science and Technology of China)
Domain AdaptationMeta LearningGraph Neural NetworkGraph
🎯 What it does: The paper proposes a Core Knowledge Learning (CKL) framework to extract the most valuable subgraphs from graphs for the tasks of graph domain adaptation and few-shot learning.
Core-to-Global Reasoning for Compositional Visual Question Answering
Hao Zhou (Naval University of Engineering), Zhangqi Jiang (National University of Defense Technology)
RecognitionObject DetectionRecurrent Neural NetworkVision Language ModelImageTextMultimodality
🎯 What it does: A core-to-global reasoning (CTGR) framework is proposed for compositional visual question answering, which first extracts core semantic features from images and questions, then aligns and suppresses redundant information through information filtering, and finally achieves multi-modal semantic fusion and predicts answers using a core-to-global fusion mechanism.
CoRe: Context-Regularized Text Embedding Learning for Text-to-Image Personalization
Feize Wu (Sun Yat-sen University), Xudong Mao (The Hong Kong Polytechnic University)
GenerationData SynthesisTransformerDiffusion modelContrastive LearningImageText
🎯 What it does: A new text embedding learning method called CoRe is proposed, which enhances personalized text alignment and identity preservation from text to image through contextual regularization.
Correcting Large Language Model Behavior via Influence Function
Han Zhang (Harbin Institute of Technology), Ruifeng Xu (Harbin Institute of Technology)
OptimizationTransformerLarge Language ModelSupervised Fine-TuningContrastive LearningText
🎯 What it does: A two-stage method named LANCET is proposed, which automatically detects and corrects inappropriate behaviors of large language models (LLMs) in the face of dynamic human preferences without the need for human intervention;
Correlation-Attention Masked Temporal Transformer for User Identity Linkage Using Heterogeneous Mobility Data
Ziang Yan (Ocean University of China), Junyu Dong (Ocean University of China)
TransformerTime SeriesSequential
🎯 What it does: This paper proposes the MT-Link framework for cross-platform user identity linking, utilizing spatiotemporal co-occurrence information.
CoSDA: Enhancing the Robustness of Inversion-based Generative Image Watermarking Framework
Han Fang (National University of Singapore), Ee-Chien Chang (National University of Singapore)
GenerationData SynthesisDiffusion modelImage
🎯 What it does: Embed watermarks in the initial latent variables of the diffusion model and reduce the error of the reverse process through compensation sampling and drift alignment mechanisms, thereby enhancing the robustness of the generated images' watermarks.
COSEE: Consistency-Oriented Signal-Based Early Exiting via Calibrated Sample Weighting Mechanism
Jianing He (Tongji University), Duoqian Miao (Tongji University)
ClassificationComputational EfficiencyTransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: This paper proposes the COSEE framework, which simulates different acceleration thresholds during the training phase and generates sample weights for each layer, allowing the internal classifier to focus on samples that may exit at that layer during training, thus achieving consistency between training and inference and supporting adjustable rates.
CoT4Rec: Revealing User Preferences Through Chain of Thought for Recommender Systems
Weiqi Yue (Hangzhou Dianzi University), Jian Wan (Hangzhou Dianzi University)
Recommendation SystemLarge Language ModelTabularChain-of-Thought
🎯 What it does: Designed and implemented a user preference analysis and dual-stage recommendation framework CoT4Rec based on chain-of-thought (CoT) reasoning, utilizing LLM to generate user preferences and construct interpretable recommendation paths.
Counterexample Guided Program Repair Using Zero-Shot Learning and MaxSAT-based Fault Localization
Pedro Orvalho (University of Oxford), Vasco M. Manquinho (Universidade de Lisboa)
OptimizationAI Code AssistantTransformerLarge Language ModelTextBenchmark
🎯 What it does: This paper proposes a zero-shot learning method that combines MaxSAT fault localization with LLM to improve automatic program repair for gate-level programming tasks.
Counterfactual Debiasing for Physical Audiovisual Commonsense Reasoning
Daoming Zong (SenseTime Research), Shuaiyu Wang (Fudan University)
ClassificationRecognitionObject DetectionTransformerContrastive LearningVideoMultimodalityAudio
🎯 What it does: A model-agnostic adversarial causal reasoning framework CF-PACR is proposed to eliminate visual bias in physical audio-visual commonsense reasoning models and enhance performance.
Counterfactual Identification Under Monotonicity Constraints
Aurghya Maiti (Columbia University), Elias Bareinboim (Columbia University)
Tabular
🎯 What it does: This paper proposes a method for identifying arbitrary counterfactual queries under causal graphs with monotonicity constraints, providing the monotonicity reduction lemma and the M-ID algorithm.
Counterfactual Online Learning for Open-Loop Monte-Carlo Planning
Thomy Phan (University of Southern California), Sven Koenig (University of California)
OptimizationReinforcement Learning
🎯 What it does: This paper proposes an open Monte Carlo Tree Search algorithm based on causal multi-armed bandits (with unobserved confounding variables) — CORAL, for online planning in partially observable environments.
Counterfactual Task-augmented Meta-learning for Cold-start Sequential Recommendation
Zhiqiang Wang (Shanxi University), Kaixuan Yao (Shanxi University of Finance and Economics)
Recommendation SystemMeta LearningSequential
🎯 What it does: This paper addresses the cold start sequential recommendation problem by proposing a counterfactual task-enhanced meta-learning framework to simulate users' potential behaviors and quickly adapt to new users.
Counting and Reasoning with Plans
David Speck (University of Basel), Augusto B. Corrêa (University of Oxford)
🎯 What it does: This paper proposes a planning space framework for counting and reasoning, capable of quantitative analysis of multi-step plans.
Coupling-based Convergence Diagnostic and Stepsize Scheme for Stochastic Gradient Descent
Xiang Li (University of Wisconsin Madison), Qiaomin Xie (University of Wisconsin Madison)
OptimizationConvolutional Neural NetworkImageTabular
🎯 What it does: We propose a convergence diagnosis method for constant step size SGD based on Markov chain coupling, and design an adaptive step size adjustment scheme based on this.
CP-DETR: Concept Prompt Guide DETR Toward Stronger Universal Object Detection
Qibo Chen (China Mobile Research and Innovation Institute), Jianzhong Chen (China Mobile Research and Innovation Institute)
Object DetectionTransformerPrompt EngineeringImageText
🎯 What it does: We propose CP-DETR, a universal object detection model achieved through concept prompts (text, visual, optimization), which combines strong zero-shot inference with performance competitive with specialized models.
CP-Guard: Malicious Agent Detection and Defense in Collaborative Bird’s Eye View Perception
Senkang Hu (City University of Hong Kong), Sam Kwong (City University of Hong Kong)
SegmentationAutonomous DrivingAdversarial AttackConvolutional Neural NetworkSupervised Fine-TuningPoint Cloud
🎯 What it does: This study investigates the robustness of cooperative perception (CP) in multi-agent systems when facing malicious agent attacks, proposing the CP-Guard defense framework, which includes two main components: Probability-Agnostic Sampling Consensus (PASAC) and Cooperative Consistency Loss (CCLoss) for detecting and eliminating malicious agents.
Cradle-VAE: Enhancing Single-Cell Gene Perturbation Modeling with Counterfactual Reasoning-based Artifact Disentanglement
Seungheun Baek (Korea University), Jaewoo Kang (AIGEN Sciences)
GenerationData SynthesisAuto EncoderBiomedical Data
🎯 What it does: This paper proposes CRADLE-VAE, a causal modeling framework based on variational autoencoders, which utilizes adversarial and counterfactual reasoning to achieve decoupling of technical artifacts and realignment of baseline states, thereby enhancing the generation quality and reliability of single-cell gene perturbation predictions.
CraftFactory: A Conditioned Control Policy Benchmark for Compositional Generalization
Jinbing Hou (Polixir Technologies), Jian Zhao (Polixir Technologies)
Robotic IntelligenceMeta LearningReinforcement Learning from Human FeedbackSequentialBenchmark
🎯 What it does: The CraftFactory benchmark is proposed for evaluating the compositional generalization of conditional control strategies in the interactive control environment of Minecraft.
CrAM: Credibility-Aware Attention Modification in LLMs for Combating Misinformation in RAG
Boyi Deng (University of Science and Technology of China), Fuli Feng (University of Science and Technology of China)
GenerationRetrievalTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: The CrAM (Credibility-aware Attention Modification) method is proposed, which identifies important attention heads and dynamically adjusts their attention weights based on document credibility within the Retrieval-Augmented Generation (RAG) framework, thereby reducing the negative impact of low-credibility documents on the generation results of large language models (LLMs).
Creating Coherence in Federated Non-Negative Matrix Factorization
Sebastian Dalleiger (KTH Royal Institute of Technology), Aristides Gionis (KTH Royal Institute of Technology)
Recommendation SystemOptimizationFederated LearningSafty and PrivacyExplainability and InterpretabilityTabularSequential
🎯 What it does: This paper proposes an alignment-aware federated non-negative matrix factorization (FedNMF) framework, which achieves interpretable latent factor learning across multiple clients by introducing drift alignment, proximal optimization, and matrix centroid aggregation.
CREST: An Efficient Conjointly-trained Spike-driven Framework for Event-based Object Detection Exploiting Spatiotemporal Dynamics
Ruixin Mao (University of Electronic Science and Technology of China), Jun Zhou (University of Electronic Science and Technology of China)
Object DetectionSpiking Neural NetworkVideo
🎯 What it does: A collaborative training and full peak-driven framework CREST based on event cameras is proposed for efficient event-driven object detection.
CriSPO: Multi-Aspect Critique-Suggestion-guided Automatic Prompt Optimization for Text Generation
Han He (Amazon Web Services Artificial Intelligence Labs), Katrin Kirchhoff (Amazon Web Services Artificial Intelligence Labs)
GenerationOptimizationTransformerLarge Language ModelPrompt EngineeringTextChain-of-Thought
🎯 What it does: A multi-dimensional criticism-suggestion driven automatic prompt optimization framework named CriSPO is proposed, further introducing AST suffix tuning to achieve joint optimization of multiple metrics;
Critical Forgetting-Based Multi-Scale Disentanglement for Deepfake Detection
Kai Li (Sun Yat-sen University), Xiaochun Cao (Sun Yat-sen University)
ClassificationRecognitionConvolutional Neural NetworkAuto EncoderContrastive LearningImageVideo
🎯 What it does: A multi-scale decoupling framework is proposed, incorporating a key forgetting mechanism to enhance the generalization ability of deepfake detection.
Cross-Domain Trajectory Association Based on Hierarchical Spatiotemporal Enhanced Attention Hypergraph
Chenlong Wu (Tiangong University), Jin Hao (Boya Triz Technology Co., Ltd.)
Recommendation SystemGraph Neural NetworkGraphTime Series
🎯 What it does: A cross-domain trajectory association model called StarNet is proposed, which integrates local space-time enhanced graph neural networks, global space-time enhanced hypergraph networks, and a fusion association module based on the Kolmogorov-Arnold network to achieve user identity linking across platforms.
Cross-Lingual Text-Rich Visual Comprehension: An Information Theory Perspective
Xinmiao Yu (Harbin Institute of Technology), Bing Qin (Harbin Institute of Technology)
Knowledge DistillationTransformerVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Construct a cross-language text-rich visual question answering benchmark XT-VQA and propose the MVCL-MI method to enhance the understanding and answering capabilities of large visual language models on cross-language text images.
Cross-Modal Few-Shot Learning with Second-Order Neural Ordinary Differential Equations
Yi Zhang (Southern University of Science and Technology), Angelica I Aviles-Rivero (Tsinghua University)
ClassificationRecognitionDomain AdaptationMeta LearningContrastive LearningImageTextMultimodalityOrdinary Differential Equation
🎯 What it does: This study proposes the SONO method, which utilizes second-order neural ordinary differential equations and text-to-image enhancement to achieve cross-modal few-shot learning.
Cross-modal Multi-task Learning for Multimedia Event Extraction
Jianwei Cao (National University of Defense Technology), Xiang Zhao (National University of Defense Technology)
Object DetectionKnowledge DistillationRepresentation LearningTransformerPrompt EngineeringImageTextMultimodality
🎯 What it does: A cross-modal multi-task learning framework X-MTL is proposed and implemented, jointly completing multimedia event extraction through four tasks (trigger detection, argument extraction, verb classification, role classification) by processing different modal inputs with a shared encoder.
Cross-Modal Stealth: A Coarse-to-Fine Attack Framework for RGB-T Tracker
Xinyu Xiang (Wuhan University), Jiayi Ma (Wuhan University)
Object TrackingAdversarial AttackGenerative Adversarial NetworkImageMultimodality
🎯 What it does: Designed and implemented a coarse-fine hierarchical adversarial patch attack framework for RGB-T cross-modal trackers, achieving covert attacks across modalities.
Cross-modulated Attention Transformer for RGBT Tracking
Yun Xiao (Anhui University), Cong Liu (iFLYTEK Company)
Object TrackingTransformerMultimodality
🎯 What it does: This paper proposes a Cross-Modal Modulation Attention Transformer (CAFormer), which simultaneously performs self-attention and cross-attention within a single attention module, achieving unified feature extraction and fusion of RGB and TIR modalities.
Cross-PCR: A Robust Cross-Source Point Cloud Registration Framework
Guiyu Zhao (Beijing Institute of Technology), Hongbin Ma (Beijing Institute of Technology)
Pose EstimationOptimizationPoint Cloud
🎯 What it does: A cross-source point cloud registration framework called Cross-PCR is proposed to address the registration failure issues caused by inconsistent density and difficulties in feature matching in cross-source point clouds.
Cross-Silo Feature Space Alignment for Federated Learning on Clients with Imbalanced Data
Zhuang Qi (Shandong University), Xiangxu Meng (Inspur)
Federated LearningRepresentation LearningContrastive LearningImage
🎯 What it does: A cross-model feature space alignment method called FedFSA is proposed to address the issue of feature space inconsistency caused by client data imbalance in federated learning.
Cross-Spectral Gaussian Splatting with Spatial Occupancy Consistency
Haipeng Guo (Harbin Institute of Technology), Junbao Li (Harbin Institute of Technology)
GenerationOptimizationComputational EfficiencyNeural Radiance FieldGaussian SplattingPoint Cloud
🎯 What it does: The SOC-GS method is proposed for real-time rendering of cross-spectral scenes.
Cross-Validated Off-Policy Evaluation
Matej Cief (Brno University of Technology), Michal Kompan (Adobe Research)
OptimizationHyperparameter SearchReinforcement LearningTabular
🎯 What it does: This paper proposes a method for estimator selection and hyperparameter tuning based on cross-validation for offline policy evaluation (OPE), which allows for estimator comparison and tuning using a dataset collected from a single logging policy.
Cross-View Graph Consistency Learning for Invariant Graph Representations
Jie Chen (Sichuan University), Xi Peng (Sichuan University)
Representation LearningGraph Neural NetworkGraph
🎯 What it does: A Cross-View Graph Consistency Learning (CGCL) method is proposed, which utilizes two complementary graphs to enhance view-invariant graph representation for link prediction.
Cross-View Referring Multi-Object Tracking
Sijia Chen (Huazhong University of Science and Technology), Wenbing Tao (Huazhong University of Science and Technology)
Object DetectionObject TrackingTransformerVision Language ModelVideoTextBenchmark
🎯 What it does: This paper proposes the Cross-Perspective Reference Multi-Object Tracking (CRMOT) task and presents an end-to-end method called CRTracker to achieve object detection, tracking, and language matching across perspectives.