AAAI 2024 Papers — Page 4
AAAI Conference on Artificial Intelligence · 2331 papers
CLIP-Guided Federated Learning on Heterogeneity and Long-Tailed Data
Jiangming Shi (Xiamen University), Yanyun Qu (Xiamen University)
Federated LearningKnowledge DistillationContrastive LearningImage
🎯 What it does: This paper proposes a federated learning framework CLIP2FL for heterogeneous and long-tail data scenarios, utilizing CLIP for knowledge distillation and prototype contrastive learning to enhance the generalization ability of both server and client models.
CMDA: Cross-Modal and Domain Adversarial Adaptation for LiDAR-Based 3D Object Detection
Gyusam Chang (Korea University), Sangpil Kim (Korea University)
Object DetectionDomain AdaptationAutonomous DrivingMultimodalityPoint Cloud
🎯 What it does: A cross-modal unsupervised domain adaptation framework CMDA is proposed, which utilizes the semantic information of camera images and interacts with LiDAR BEV features, further eliminating the source-target domain gap through cross-domain adversarial self-training, enhancing the cross-domain generalization performance of LiDAR-based 3D object detection.
CMG-Net: Robust Normal Estimation for Point Clouds via Chamfer Normal Distance and Multi-Scale Geometry
Yingrui Wu (Institute of Automation, Chinese Academy of Sciences), Dong-Ming Yan (Institute of Automation, Chinese Academy of Sciences)
OptimizationPoint Cloud
🎯 What it does: A robust point cloud normal estimation network called CMG-Net based on Chamfer Normal Distance (CND) is proposed, which improves accuracy through multi-scale local feature aggregation and hierarchical geometric information fusion.
Code-Style In-Context Learning for Knowledge-Based Question Answering
Zhijie Nie (Beihang University), Xudong Liu (Beihang University)
RetrievalKnowledge DistillationTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: A training-free KBQA framework called KB-Coder is designed, which utilizes code-style contextual learning to transform logical form generation into code call sequences, and enhances zero-shot generalization through retrieval augmentation.
CoLAL: Co-learning Active Learning for Text Classification
Linh Le (University of Queensland), Gianluca Demartini (University of Queensland)
ClassificationTransformerLarge Language ModelImageText
🎯 What it does: A new active learning algorithm called Co-learning Active Learning (CoLAL) is proposed, which identifies the most valuable unlabeled samples by integrating the predictions of the target model and the peer model, thereby enhancing the learning efficiency of text and image classification.
Collaborative Consortium of Foundation Models for Open-World Few-Shot Learning
Shuai Shao (Zhejiang Lab), Bin Liu (Zhejiang Lab)
ClassificationTransformerLarge Language ModelContrastive LearningImageMultimodality
🎯 What it does: This paper proposes an open-world few-shot learning framework CO3 based on a collaborative foundational model, which can achieve high-accuracy classification in scenarios with severe label noise.
Collaborative Synthesis of Patient Records through Multi-Visit Health State Inference
Hongda Sun (Renmin University of China), Rui Yan (Renmin University of China)
GenerationData SynthesisSafty and PrivacyTransformerLarge Language ModelTextBiomedical DataElectronic Health Records
🎯 What it does: A multi-visit medical history health status inference model (MSIC) is proposed to collaboratively generate electronic health records (EHR) and corresponding medical reports.
Collaborative Tooth Motion Diffusion Model in Digital Orthodontics
Yeying Fan (Shandong University), Yuanfeng Zhou (Shandong University)
GenerationGraph Neural NetworkDiffusion modelAuto EncoderPoint Cloud
🎯 What it does: This paper proposes a collaborative tooth movement generation framework based on diffusion models, capable of learning and sampling the entire intermediate motion process under the conditions of given initial and target tooth arrangements, achieving full process automation.
Collaborative Weakly Supervised Video Correlation Learning for Procedure-Aware Instructional Video Analysis
Tianyao He (Shanghai Jiao Tong University), Weiyao Lin (Shanghai Jiao Tong University)
RecognitionContrastive LearningVideo
🎯 What it does: A Collaborative Program Alignment (CPA) framework is proposed to achieve program-aware related learning of instructional videos without step annotations.
ColNeRF: Collaboration for Generalizable Sparse Input Neural Radiance Field
Zhangkai Ni (Tongji University), Sam Kwong (City University of Hong Kong)
GenerationData SynthesisTransformerNeural Radiance FieldPoint Cloud
🎯 What it does: A new transferable sparse input neural radiance field model, ColNeRF, is proposed, which achieves high-quality novel view rendering under sparse viewpoints by utilizing collaborative fusion between input views and self-supervised constraints at the output layer.
Color Event Enhanced Single-Exposure HDR Imaging
Mengyao Cui (University of Hong Kong), Xuelong Li (Northwestern Polytechnical University)
RestorationData SynthesisTransformerImage
🎯 What it does: A framework for HDR reconstruction using a combination of color event cameras and single-exposure LDR images is proposed, along with the construction of corresponding synthetic and real datasets.
Colored Noise in PPO: Improved Exploration and Performance through Correlated Action Sampling
Jakob Hollenstein (University of Innsbruck), Justus Piater (University of Innsbruck)
Reinforcement LearningSequential
🎯 What it does: This paper introduces correlated (colored) noise into PPO, and experiments verify that it can significantly enhance exploration and performance;
Colorizing Monochromatic Radiance Fields
Yean Cheng (Peking University), Boxin Shi (Peking University)
Image TranslationGenerationTransformerNeural Radiance FieldImage
🎯 What it does: This paper proposes ColorNeRF, a framework for reconstructing a three-dimensional neural radiance field with achievable color consistency and saturation from monochromatic multi-view images.
Colour Passing Revisited: Lifted Model Construction with Commutative Factors
Malte Luttermann (German Research Center for Artificial Intelligence), Marcel Gehrke (Institute of Information Systems)
OptimizationGraph
🎯 What it does: An improved color propagation algorithm (ACP) is proposed, which can convert propositional factor graphs into parameterized factor graphs (PFG) containing logical variables, thus achieving a dimensional representation independent of inference algorithms.
COMBAT: Alternated Training for Effective Clean-Label Backdoor Attacks
Tran Huynh (VinAI Research), Anh Tran (VinAI Research)
GenerationAdversarial AttackConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: A framework for clean-label backdoor attacks based on alternating training, called COMBAT, is proposed, which utilizes a generator to produce low-frequency, blurred triggers and optimizes them together with a surrogate model.
Combating Data Imbalances in Federated Semi-supervised Learning with Dual Regulators
Sikai Bai (Hong Kong University of Science and Technology), Junyu Gao (Northwestern Polytechnical University)
Federated LearningMeta LearningImage
🎯 What it does: A federated semi-supervised learning framework named FedDure is proposed, which dynamically updates the model on the client side through dual regulators to address external and internal data imbalance issues.
COMBHelper: A Neural Approach to Reduce Search Space for Graph Combinatorial Problems
Hao Tian (Tongji University), Wei Ye (University of Illinois Chicago)
OptimizationComputational EfficiencyKnowledge DistillationGraph Neural NetworkGraph
🎯 What it does: This paper proposes a framework called COMBHELPER, which uses Graph Neural Networks (GNN) to predict which nodes are likely to appear in the solutions of combinatorial optimization (CO), thereby pruning the search space and running traditional CO algorithms (such as linear programming, greedy algorithms, and local search) on the pruned subgraphs, significantly improving solving efficiency.
Combinatorial CNN-Transformer Learning with Manifold Constraints for Semi-supervised Medical Image Segmentation
Huimin Huang (Zhejiang University), Yefeng Zheng (Guangxi Medical University)
SegmentationConvolutional Neural NetworkTransformerImageBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: The study proposes a dual-student semi-supervised medical image segmentation framework M-CnT based on convolutional networks and Transformers, introducing class-level consistency and knowledge transfer constraints between students in the manifold space to enhance segmentation performance.
Combinatorial Stochastic-Greedy Bandit
Fares Fourati (King Abdullah University of Science and Technology), Vaneet Aggarwal (Purdue University)
Recommendation SystemOptimizationGraph
🎯 What it does: This paper proposes a full-bandwidth feedback algorithm for the Combinatorial Multi-Armed Bandit (CMAB) problem—Stochastic-Greedy Bandit (SGB). It explores by randomly sampling the unselected arms with an optimized ratio during each round of greedy iteration, and then utilizes the constructed super arms.
Combining Multiple Supervision for Robust Zero-Shot Dense Retrieval
Yan Fang (Tsinghua University), Zhao Cao (Tsinghua University)
RetrievalTransformerContrastive LearningText
🎯 What it does: Proposes the RMSC (Robust Multi-Supervision Combining) strategy, which uses soft tokens to explicitly distinguish between source domain/target domain and the matching signals of human annotations/weak supervision during training, thereby achieving zero-shot dense retrieval.
COMMA: Co-articulated Multi-Modal Learning
Lianyu Hu (Tianjin University), Wei Feng (University of Macau)
ClassificationDomain AdaptationKnowledge DistillationTransformerPrompt EngineeringImageMultimodality
🎯 What it does: The COMMA method is proposed, which enhances CLIP's performance on zero-shot generalization tasks by generating prompts that are interrelated between the visual and language branches and constraining the feature gap between learned prompts and pre-trained prompts.
Commonsense for Zero-Shot Natural Language Video Localization
Meghana Holla (Virginia Tech), Ismini Lourentzou (University of Illinois at Urbana Champaign)
Graph Neural NetworkVideoText
🎯 What it does: Developed the CORONET framework, which enhances pseudo-queries and video representations through consensus knowledge in zero-shot natural language video localization tasks, achieving cross-modal alignment.
Communication Efficient Distributed Newton Method over Unreliable Networks
Ming Wen (Fudan University), Yuedong Xu (Fudan University)
OptimizationTabular
🎯 What it does: Designed and proposed the RED-New distributed Newton method, achieving communication-efficient and robust second-order optimization in the packet loss environment of unreliable networks.
Communication-Efficient Collaborative Regret Minimization in Multi-Armed Bandits
Nikolai Karpov (Indiana University), Qin Zhang (Indiana University)
OptimizationReinforcement Learning
🎯 What it does: In the collaborative learning model of multi-agent multi-armed bandits, the trade-off relationship between the number of communication rounds and cumulative risk is studied, providing lower and upper bounds when the number of communication rounds is limited.
Compact HD Map Construction via Douglas-Peucker Point Transformer
Ruixin Liu (Xi'an Jiaotong University), Zejian Yuan (Xi'an Jiaotong University)
CompressionAutonomous DrivingTransformerPoint Cloud
🎯 What it does: This paper proposes DPFormer, an end-to-end Transformer framework that utilizes Douglas-Peucker (DP) points for high-precision and compact HD map construction.
Comparing the Robustness of Modern No-Reference Image- and Video-Quality Metrics to Adversarial Attacks
Anastasia Antsiferova (Moscow State University Institute for Artificial Intelligence), Dmitriy Vatolin (Moscow State University Institute for Artificial Intelligence)
Adversarial AttackImageVideoBenchmark
🎯 What it does: This paper establishes a benchmark to evaluate the robustness of 15 no-reference image/video quality assessment (NR IQA/VQA) metrics under various adversarial attacks (FGSM, I-FGSM, MI-FGSM, UAP, etc.).
Competition among Pairwise Lottery Contests
Xiaotie Deng (Peking University), Qi Qi (Renmin University of China)
Optimization
🎯 What it does: This paper studies how designers allocate awards, select participants, and set biases in pairwise lottery competitions (PLC) among multiple participants. It analyzes the balance of effort distribution among participants and proves the existence of subgame perfect equilibrium (SPE) and weak designer equilibrium (WDE) within a two-stage game framework, while also providing an efficient algorithm for computing approximate equilibria.
Complementary Knowledge Distillation for Robust and Privacy-Preserving Model Serving in Vertical Federated Learning
Dashan Gao (Southern University of Science and Technology), Qiang Yang (Webank AI Lab)
Recommendation SystemFederated LearningSafty and PrivacyKnowledge DistillationTabularBiomedical Data
🎯 What it does: In the model inference phase of vertical federated learning, a Complementary Knowledge Distillation (CKD) framework is proposed, which utilizes Complementary Label Coding (CLC) to only transmit the label information that the active party has not learned, thereby enhancing model robustness and reducing the risk of label leakage.
Complete Neural Networks for Complete Euclidean Graphs
Snir Hordan (Technion Israel Institute of Technology), Nadav Dym (Technion Israel Institute of Technology)
Graph Neural NetworkPoint Cloud
🎯 What it does: A method for determining the integrity of three-dimensional point clouds is proposed, proving that 3-WL and 2-EWL can completely distinguish any non-equivalent point clouds, and based on this, a differentiable Euclidean graph neural network is designed.
Completing Priceable Committees: Utilitarian and Representation Guarantees for Proportional Multiwinner Voting
Markus Brill (University of Warwick), Jannik Peters (TU Berlin)
🎯 What it does: This paper studies how to achieve optimal utility and representation guarantees in approval-based multi-winner elections by completing 'affordable committees' while satisfying strong proportionality axioms (such as EJR+).
Complexity of Credulous and Skeptical Acceptance in Epistemic Argumentation Framework
Gianvincenzo Alfano (University of Calabria), Irina Trubitsyna (University of Calabria)
🎯 What it does: This paper studies the computational complexity issues of credible and dubious acceptance in the extended argumentation framework with knowledge constraints (Epistemic Argumentation Framework, EAF) and its sub-framework (Labelled Constrained AF, LCAF).
Complexity of Neural Network Training and ETR: Extensions with Effectively Continuous Functions
Teemu Hankala (University of Helsinki), Jonni Virtema (University of Sheffield)
🎯 What it does: This study investigates the computational complexity of training neural networks and relates it to the extended real existence theory, proving that training problems caused by different activation functions belong to the corresponding ∃R class.
Component Fourier Neural Operator for Singularly Perturbed Differential Equations
Ye Li (Tsinghua University), Zhongyi Huang (Tsinghua University)
Time SeriesStochastic Differential EquationOrdinary Differential Equation
🎯 What it does: This paper proposes the Component Fourier Neural Operator (ComFNO), a neural operator learning method specifically designed to solve singular partial differential equations (SPDEs) with thin layer structures.
Composing Biases by Using CP to Decompose Minimal Functional Dependencies for Acquiring Complex Formulae
Ramiz Gindullin (IMT Atlantique), Claude-Guy Quimper (Universite Laval)
Tabular
🎯 What it does: A decomposition method based on constraint programming is proposed, which gradually breaks down minimal functional dependencies and combines various learning biases to learn complex formulas.
Composite Active Learning: Towards Multi-Domain Active Learning with Theoretical Guarantees
Guang-Yuan Hao (Hong Kong University of Science and Technology), Hao Wang (Rutgers University)
ClassificationDomain AdaptationImage
🎯 What it does: A multi-domain active learning framework called Composite Active Learning (CAL) is proposed, which constructs proxy domains by first estimating inter-domain similarity, then allocates labeling budgets in each domain, and combines instance-level query strategies to enhance overall classification performance.
Composite Sketch+Text Queries for Retrieving Objects with Elusive Names and Complex Interactions
Prajwal Gatti (Indian Institute of Technology Jodhpur), Anand Mishra (Microsoft)
RetrievalTransformerContrastive LearningImageTextMultimodality
🎯 What it does: This study investigates a new retrieval mode, which combines hand-drawn sketches with short text queries to retrieve target objects from a natural scene image database.
Compositional Generalization for Multi-Label Text Classification: A Data-Augmentation Approach
Yuyang Chai (Wuhan University), Chong Teng (Wuhan University)
ClassificationGenerationTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This study investigates the combinatorial generalization (CG) problem in multi-label text classification, proposing a specialized training/testing split and evaluation metrics. It introduces two generative models based on label representation disentanglement (LS-PT and LD-VAE) for data augmentation to enhance the model's ability to recognize rare combinatorial labels.
Compositional Inversion for Stable Diffusion Models
Xulu Zhang (Hong Kong Polytechnic University), Qing Li (Hong Kong Polytechnic University)
GenerationData SynthesisTransformerDiffusion modelImageText
🎯 What it does: This paper proposes a Compositional Inversion method that improves the text inversion of the Stable Diffusion model using semantic anchoring and spatial regularization, addressing the overfitting and concept dominance issues caused by the original inversion.
Compositional Text-to-Image Synthesis with Attention Map Control of Diffusion Models
Ruichen Wang (OPPO Research Institute), Xiaodong Lin (Rutgers University)
GenerationData SynthesisTransformerDiffusion modelImageText
🎯 What it does: The BoxNet module is proposed to predict the bounding boxes of each entity during the sampling process of Stable Diffusion, and based on these boxes, unique mask controls are applied to cross-attention and self-attention, achieving multi-entity and attribute semantic consistent synthesis from text to image.
Compound Text-Guided Prompt Tuning via Image-Adaptive Cues
Hao Tan (Institute of Automation Chinese Academy of Sciences), Xiangyu Zhang (MEGVII Technology)
ClassificationOptimizationComputational EfficiencyTransformerSupervised Fine-TuningPrompt EngineeringImage
🎯 What it does: This paper proposes the Compound Text-Guided Prompt Tuning (TGP-T) method, which utilizes category-level and content-level text supervision to optimize prompts, significantly reducing GPU memory usage and improving few-shot visual classification performance.
Comprehensive View Embedding Learning for Single-Cell Multimodal Integration
Zhenchao Tang (Sun Yat-sen University), Calvin Yu-Chian Chen (Peking University)
Representation LearningData-Centric LearningGraph Neural NetworkTransformerContrastive LearningMultimodalityBiomedical Data
🎯 What it does: An unsupervised single-cell multimodal integration method called CoVEL is proposed, which can learn unified embeddings from three perspectives (regulatory relationships, cell fine-grained features, and intercellular similarity) to address the information loss problem caused by feature space mismatches between different modalities.
Comprehensive Visual Grounding for Video Description
Wenhui Jiang (Peking University), Yang Liu (Sany Heavy Industry Company Limited)
RecognitionObject DetectionObject TrackingTransformerVideoText
🎯 What it does: This paper proposes a comprehensive visual localization network that combines spatial-temporal entity localization and action localization to enhance the accuracy of video subtitles.
Compressing Image-to-Image Translation GANs Using Local Density Structures on Their Learned Manifold
Alireza Ganjdanesh (University of Maryland), Heng Huang (University of Pittsburgh)
Image TranslationCompressionAuto EncoderGenerative Adversarial NetworkImage
🎯 What it does: A compression method is proposed for image-to-image GANs (Pix2Pix and CycleGAN), with the core idea being to maintain a similar local density structure on the learned data manifold of the pruned model compared to the original model.
Computing Nash Equilibria in Potential Games with Private Uncoupled Constraints
Nikolas Patris (University of California), Ioannis Panageas (University of California)
OptimizationGraph
🎯 What it does: This paper proposes a distributed gradient descent algorithm IGDλ based on the regularized Lagrangian function, aimed at solving the ε-approximate Nash equilibrium of potential games with private convex constraints.
Computing the Why-Provenance for Datalog Queries via SAT Solvers
Marco Calautti (University of Milan), Markus Schneider (University of Edinburgh)
🎯 What it does: This paper proposes an efficient method for calculating the Why-Provenance of Datalog queries using SAT solvers, and implements on-demand (incremental) generation of explanation sets.
ConcaveQ: Non-monotonic Value Function Factorization via Concave Representations in Deep Multi-Agent Reinforcement Learning
Huiqun Li (Shandong University), Tian Lan (The George Washington University)
Reinforcement Learning
🎯 What it does: Proposes ConcaveQ, which utilizes non-monotonic concave mixture networks to achieve multi-agent value function decomposition;
Concept-Guided Prompt Learning for Generalization in Vision-Language Models
Yi Zhang (Harbin Institute of Technology), Zhihai He (Carnegie Mellon University)
ClassificationDomain AdaptationRepresentation LearningTransformerPrompt EngineeringVision Language ModelContrastive LearningImageMultimodality
🎯 What it does: This paper proposes the Concept-Guided Prompt Learning (CPL) scheme, which constructs a visual concept cache and projects multi-layer visual features into the text space to leverage the pre-trained knowledge of CLIP and generate fine-grained concept prompts, thereby enhancing the model's cross-domain generalization performance.
ConceptBed: Evaluating Concept Learning Abilities of Text-to-Image Diffusion Models
Maitreya Patel (Arizona State University), Yezhou Yang (University of Maryland Baltimore County)
GenerationData SynthesisConvolutional Neural NetworkDiffusion modelImageBenchmark
🎯 What it does: The CONCEPTBED dataset and the Concept Confidence Deviation (CCD) evaluation metric are proposed to quantify the ability of text-to-image diffusion models in learning and combining new visual concepts.
Conditional Backdoor Attack via JPEG Compression
Qiuyu Duan (Harbin Institute of Technology), Leo Yu Zhang (Griffith University)
CompressionAdversarial AttackImage
🎯 What it does: A conditional backdoor attack is proposed, using JPEG compression as the trigger condition, allowing the model to trigger the attack only when the input is JPEG compressed.
Conditional Variational Autoencoder for Sign Language Translation with Cross-Modal Alignment
Rui Zhao (Xiamen University), Yidong Chen (Xiamen University)
RecognitionGenerationAuto EncoderVideoText
🎯 What it does: A framework for sign language translation without gloss based on Conditional Variational Autoencoders (CV-SLT) is proposed, which directly aligns sign language videos with cross-modal representations of text.
ConditionVideo: Training-Free Condition-Guided Video Generation
Bo Peng (Shanghai Jiao Tong University), Yu Qiao (Shanghai Artificial Intelligence Laboratory)
GenerationData SynthesisPose EstimationDiffusion modelVideo
🎯 What it does: A training-free ConditionVideo method is proposed, utilizing existing text-to-image diffusion models and conditional guidance to generate high-quality, temporally consistent videos.
Conformal Autoregressive Generation: Beam Search with Coverage Guarantees
Nicolas Deutschmann (IBM Research), María Rodríguez Martínez (IBM Research)
GenerationTransformerSupervised Fine-TuningSequential
🎯 What it does: Two types of conformal prediction-based beam search extensions are proposed to provide coverage guarantees during sequence generation, addressing the lack of theoretical basis in traditional beam search.
Conformal Crystal Graph Transformer with Robust Encoding of Periodic Invariance
Yingheng Wang (Cornell University), Carla P. Gomes (Cornell University)
TransformerGraphPhysics Related
🎯 What it does: A new crystal graph construction method and an angle-preserving graph Transformer model, CrystalFormer, are proposed for more accurate prediction of crystal material properties.
Confucius: Iterative Tool Learning from Introspection Feedback by Easy-to-Difficult Curriculum
Shen Gao (Shandong University), Zhaochun Ren (Leiden University)
TransformerLarge Language ModelPrompt EngineeringTextChain-of-Thought
🎯 What it does: The Confucius framework is proposed, which combines multi-stage learning with self-reflective iterative self-instruction to train LLMs to efficiently use various tools in real-world scenarios.
Confusing Pair Correction Based on Category Prototype for Domain Adaptation under Noisy Environments
Churan Zhi (Beijing Jiaotong University), Shuhui Wang (Institute of Computing Technology, Chinese Academy of Sciences)
Domain AdaptationImage
🎯 What it does: Proposes an unsupervised domain adaptation method in noisy environments, utilizing category prototypes to identify and correct mislabeling of similar class pairs.
ConSequence: Synthesizing Logically Constrained Sequences for Electronic Health Record Generation
Brandon Theodorou (University of Illinois at Urbana-Champaign), Jimeng Sun (GE Healthcare)
GenerationData SynthesisTime SeriesSequentialBiomedical DataElectronic Health Records
🎯 What it does: The ConSequence model is constructed to ensure logical consistency through rule constraints when generating electronic health records, avoiding the generation of invalid or clinically non-compliant samples.
CONSIDER: Commonalities and Specialties Driven Multilingual Code Retrieval Framework
Rui Li (University of Science and Technology of China), Shijin Wang (Hefei Normal University)
RetrievalTransformerContrastive LearningText
🎯 What it does: Proposes the CONSIDER framework, which enhances multilingual code retrieval performance by modeling linguistic commonality and specificity.
Considering Nonstationary within Multivariate Time Series with Variational Hierarchical Transformer for Forecasting
Muyao Wang (Xidian University), Bo Chen (Xidian University)
TransformerAuto EncoderTime Series
🎯 What it does: The HTV-Trans model is designed, combining the Hierarchical Time Series Variational Generation Module (HTPGM) with Transformer to address the non-stationarity and non-deterministic issues of multivariate time series.
Consistency-GAN: Training GANs with Consistency Model
Yunpeng Wang (Nanchang University), Hong Rao (Nanchang University)
GenerationData SynthesisGenerative Adversarial NetworkImageStochastic Differential EquationOrdinary Differential Equation
🎯 What it does: A Consistency-GAN is proposed, which combines a consistency mapping module with GAN to inject instance noise in a few steps while ensuring high quality, rich diversity, and fast sampling.
Consistency-Guided Temperature Scaling Using Style and Content Information for Out-of-Domain Calibration
Wonjeong Choi (Korea Advanced Institute of Science and Technology), Jaekyun Moon (Korea Advanced Institute of Science and Technology)
Domain AdaptationConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: A post-hoc calibration method based on temperature scaling is proposed—Consistency Guided Temperature Scaling (CTS), which utilizes the style and content information of source domain samples to enhance the model's calibration performance in unknown domains.
ConsistentEE: A Consistent and Hardness-Guided Early Exiting Method for Accelerating Language Models Inference
Ziqian Zeng (South China University of Technology), Cen Chen (Nanjing University of Aeronautics and Astronautics)
Computational EfficiencyTransformerReinforcement LearningText
🎯 What it does: This paper proposes ConsistentEE, an early exit method based on reinforcement learning, which requires only one internal classifier to correctly predict instances during the training phase, thereby achieving consistency between training and inference.
ConsistNER: Towards Instructive NER Demonstrations for LLMs with the Consistency of Ontology and Context
Chenxiao Wu (Southeast University), Wanyi Chen (Beijing Institute of Computer Technology and Application)
RecognitionTransformerLarge Language ModelTextBiomedical Data
🎯 What it does: A three-stage framework called ConsistNER is proposed, which utilizes LLM to retrieve highly relevant demonstration samples under low-resource conditions by maintaining entity ontology and contextual consistency, thereby improving NER performance.
Constrained Bayesian Optimization under Partial Observations: Balanced Improvements and Provable Convergence
Shengbo Wang (University of Electronic Science and Technology of China), Ke Li (University of Exeter)
OptimizationReinforcement LearningTabular
🎯 What it does: A Bayesian optimization framework for partially observable constraints (CBOB) is proposed, which efficiently solves POCOPs by balancing exploration and exploitation and utilizing HLGP (Heterogeneous Likelihood Gaussian Process).
Constraint Latent Space Matters: An Anti-anomalous Waveform Transformation Solution from Photoplethysmography to Arterial Blood Pressure
Cheng Bian (OPPO Health Lab), Zijing Zeng (OPPO Health Lab)
GenerationAnomaly DetectionTransformerAuto EncoderTime SeriesBiomedical DataElectronic Health Records
🎯 What it does: This paper proposes the Latent Space Constraint Transformer (LSCT), which addresses the potential space drift caused by abnormal signals in the process of converting PPG to ABP by quantifying the latent space, CAM, and MSEK, achieving more accurate blood pressure waveform predictions.
ContactGen: Contact-Guided Interactive 3D Human Generation for Partners
Dongjun Gu (Ulsan National Institute of Science and Technology), Kyungdon Joo (Ulsan National Institute of Science and Technology)
GenerationPose EstimationTransformerDiffusion modelMesh
🎯 What it does: A 3D human interaction generation method based on a guided diffusion model (ContactGen) is proposed, which can generate diverse human poses that physically interact with a partner under the conditions of a given partner model and interaction labels.
Content Filtering with Inattentive Information Consumers
Ian Ball (Microsoft), Aleksandrs Slivkins (Microsoft)
🎯 What it does: A game model between content filtering and information consumers is established, considering the rational attention cost of consumers and the quality improvement of filters, analyzing the impact of different revenue alignment situations and attacker strategies on equilibrium.
Context Enhanced Transformer for Single Image Object Detection in Video Data
Seungjun An (Korea University), Seungryong Kim (SK Telecom)
Object DetectionTransformerVideo
🎯 What it does: A single-frame video object detection framework CETR is proposed, which incorporates a context memory module into the Transformer encoder, score-based multi-threshold sampling, and a memory-guided Transformer decoder to achieve single-frame video detection.
Context-Aware Iteration Policy Network for Efficient Optical Flow Estimation
Ri Cheng (Fudan University), Bo Yan (Fudan University)
Image TranslationOptimizationComputational EfficiencyConvolutional Neural NetworkRecurrent Neural NetworkOptical FlowImageVideo
🎯 What it does: A context-aware iterative strategy network has been designed and implemented, allowing optical flow estimation to dynamically determine the number of iterations based on the improvement of each sample, significantly reducing FLOPs.
Context-I2W: Mapping Images to Context-Dependent Words for Accurate Zero-Shot Composed Image Retrieval
Yuanmin Tang (Institute of Information Engineering, Chinese Academy of Sciences), Qi Wu (Australia Institute of Machine Learning, University of Adelaide)
RetrievalVision Language ModelContrastive LearningImage
🎯 What it does: Proposes the Context-I2W network, which maps reference images to pseudo-words related to descriptions, achieving zero-shot composite image retrieval.
Contextual Pandora’s Box
Alexia Atsidakou (University of Texas), Christos Tzamos (University of Wisconsin)
Optimization
🎯 What it does: A Contextual Pandora's Box model is proposed, along with a no-regret algorithm in the online learning framework.
Contextual Pre-planning on Reward Machine Abstractions for Enhanced Transfer in Deep Reinforcement Learning
Guy Azran (Technion), Sarah Keren (Technion)
Reinforcement Learning
🎯 What it does: A Contextual PRE-Planning (C-PREP) framework is proposed, which utilizes a Reward Machine (RM) to generate task-specific abstractions for contextual MDPs and employs optimal abstraction transfer labels and potential reward shaping in deep reinforcement learning to achieve zero/few-shot transfer.
Continual Relation Extraction via Sequential Multi-Task Learning
Thanh-Thien Le (VinAI Research), Thien Huu Nguyen (University of Oregon)
OptimizationKnowledge DistillationTransformerContrastive LearningTextSequential
🎯 What it does: This paper proposes a CREST framework that utilizes frozen BERT, GMM generative replay, and adaptive gradient multi-objective optimization to achieve continuous relation extraction while mitigating catastrophic forgetting.
Continual Vision-Language Retrieval via Dynamic Knowledge Rectification
Zhenyu Cui (Wangxuan Institute of Computer Technology Peking University), Jiahuan Zhou (ByteDance Inc)
RetrievalKnowledge DistillationContrastive LearningImageTextMultimodality
🎯 What it does: A framework named Dynamic Knowledge Rectification (DKR) is proposed and implemented to dynamically filter and correct erroneous affinity knowledge generated by old models in continuous visual-text retrieval tasks, alleviating catastrophic forgetting and maintaining the learning of new knowledge.
Continuous Piecewise-Affine Based Motion Model for Image Animation
Hexiang Wang (Shanghai Jiao Tong University), Lizhuang Ma (East China Normal University)
Image TranslationGenerationTransformerImageVideo
🎯 What it does: This paper proposes an unsupervised image animation method that utilizes a Continuous Piecewise Affine (CPAB) transformation model to map source images to the motion of driving videos, and enhances animation quality by extracting keypoint semantic information through SAM and structural information through DINO ViT.
Continuous Rotation Group Equivariant Network Inspired by Neural Population Coding
Zhiqiang Chen (Beijing Academy of Artificial Intelligence), Shan Yu (Institute of Automation)
ClassificationRecognitionConvolutional Neural NetworkGaussian SplattingImage
🎯 What it does: This paper proposes a Continuous Rotational Group Equivariant Convolutional Network (CRGEN) based on neural population coding of Bell-shaped tuning curves, achieving continuous equivariance under discrete convolution through sparse learnable weights and Gaussian tuning to generate convolution kernels.
Continuous Treatment Effect Estimation Using Gradient Interpolation and Kernel Smoothing
Lokesh Nagalapatti (Indian Institute of Technology Bombay), Sunita Sarawagi (Indian Institute of Technology Bombay)
Biomedical Data
🎯 What it does: A method for generating unsupervised control samples based on gradient interpolation and Gaussian process kernel smoothing is proposed for the individualized estimation of continuous treatment effects.
Continuous-Time Graph Representation with Sequential Survival Process
Abdulkadir Çelikkanat, Morten Mørup (Technical University of Denmark)
Representation LearningGraph Neural NetworkGraphStochastic Differential Equation
🎯 What it does: A new continuous-time graph representation method called GRAS 2P is proposed, which explicitly models the persistence and disappearance of network links using sequential survival processes and survival functions, resulting in time-varying node embeddings.
ContraNovo: A Contrastive Learning Approach to Enhance De Novo Peptide Sequencing
Zhi Jin (Shanghai Artificial Intelligence Laboratory), Siqi Sun (Research Institute of Intelligent Complex Systems Fudan University)
Protein Structure PredictionTransformerContrastive LearningBiomedical Data
🎯 What it does: A de novo peptide sequence prediction algorithm called ContraNovo based on contrastive learning is proposed, which utilizes quality information to enhance prediction accuracy during the decoding process.
Contrastive Balancing Representation Learning for Heterogeneous Dose-Response Curves Estimation
Minqin Zhu (Zhejiang University), Kun Kuang (Didi Chuxing)
Representation LearningContrastive LearningTabular
🎯 What it does: A deep network named CRNet is proposed for estimating individual heterogeneous dose-response curves (HDRC).
Contrastive Continual Learning with Importance Sampling and Prototype-Instance Relation Distillation
Jiyong Li (Sun Yat-sen University), Shangsong Liang (Sun Yat-sen University)
ClassificationKnowledge DistillationContrastive LearningImage
🎯 What it does: A contrastive continual learning framework based on importance sampling (CCLIS) is proposed, which recovers the data distribution of previous tasks by replaying important samples from a buffer and maintains knowledge through prototype-instance relationship distillation.
Contrastive Tuning: A Little Help to Make Masked Autoencoders Forget
Johannes Lehner (Johannes Kepler University), Sepp Hochreiter (Johannes Kepler University)
ClassificationRepresentation LearningTransformerAuto EncoderContrastive LearningImage
🎯 What it does: Fine-tune the pre-trained Masked Autoencoder (MAE) through sequential contrastive learning to form well-clustered semantic representations;
Contributing Dimension Structure of Deep Feature for Coreset Selection
Zhijing Wan (Wuhan University), Shin'ichi Satoh (National Institute of Informatics)
ClassificationOptimizationConvolutional Neural NetworkSupervised Fine-TuningImage
🎯 What it does: A diversity measurement and constraint mechanism based on Contribution Dimension Structure (CDS) is proposed to improve the coreset selection process.
Controllable 3D Face Generation with Conditional Style Code Diffusion
Xiaolong Shen (Zhejiang University), Zongxin Yang (Alibaba Group)
GenerationData SynthesisDiffusion modelGenerative Adversarial NetworkImage
🎯 What it does: A controllable 3D face generation framework TEx-Face based on a three-stage approach is proposed, which completes 3D GAN inversion, conditional style code diffusion, and 3D face decoding.
Controllable Mind Visual Diffusion Model
Bohan Zeng (Beihang University), Baochang Zhang (Beihang University)
GenerationDiffusion modelImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: Based on functional magnetic resonance imaging (fMRI) signals, a Controlled Visual Diffusion Model (CMVDM) is constructed to generate images that are highly similar to the original visual stimuli.
Controller-Guided Partial Label Consistency Regularization with Unlabeled Data
Qian-Wei Wang (Tsinghua University), Shu-Tao Xia (Tsinghua University)
ClassificationContrastive LearningImage
🎯 What it does: A Partial Label Learning framework (ConCont) is proposed that utilizes unlabeled data for label consistency regularization.
Convolutional Channel-Wise Competitive Learning for the Forward-Forward Algorithm
Andreas Papachristodoulou (University of Cyprus), Theocharis Theocharides (University of Cyprus)
ClassificationConvolutional Neural NetworkImage
🎯 What it does: An improvement of the Forward-Forward (FF) algorithm based on convolutional channel competitive learning is proposed, using CFSE blocks and CwC loss to achieve hierarchical training without negative samples.
ConVQG: Contrastive Visual Question Generation with Multimodal Guidance
Li Mi (École Polytechnique Fédérale de Lausanne), Devis Tuia (École Polytechnique Fédérale de Lausanne)
GenerationTransformerVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: This paper proposes a visual question generation method based on dual-modal contrastive learning, named ConVQG, which can generate questions that are highly relevant to image content and rich in knowledge while satisfying textual constraints (answers, knowledge triples, or titles).
Cooper: Coordinating Specialized Agents towards a Complex Dialogue Goal
Yi Cheng (Hong Kong Polytechnic University), Yefeng Zheng (Tencent)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: The COOPER framework is proposed, which coordinates specialized multi-agents, each responsible for different aspects of complex dialogue goals, to jointly advance the dialogue process to achieve the goals.
Cooperative Knowledge Distillation: A Learner Agnostic Approach
Michael Livanos (University of California), Stephen Wong (University of California)
Knowledge DistillationImageTabular
🎯 What it does: A Cooperative Knowledge Distillation framework is proposed, allowing multiple models to act as both teachers and students, achieving targeted knowledge sharing through the generation of 'essential' counterfactual instances.
CoPL: Contextual Prompt Learning for Vision-Language Understanding
Koustava Goswami (Adobe Research), Balaji Vasan Srinivasan (Adobe Research)
ClassificationDomain AdaptationTransformerPrompt EngineeringVision Language ModelImageMultimodality
🎯 What it does: A Contextual Prompt Learning (CoPL) method is designed to dynamically adjust soft prompts using local image features within the CLIP framework, enhancing performance in cross-domain, few-shot, and zero-shot classification tasks.
CORECODE: A Common Sense Annotated Dialogue Dataset with Benchmark Tasks for Chinese Large Language Models
Dan Shi (Tianjin University), Deyi Xiong (Zhejiang Lab)
TransformerLarge Language ModelSupervised Fine-TuningTextBenchmark
🎯 What it does: A large-scale Chinese two-person dialogue commonsense knowledge annotation dataset, CORECODE, has been constructed, and six benchmark tasks have been designed to evaluate the commonsense reasoning and conflict detection capabilities of large language models.
Coreference Graph Guidance for Mind-Map Generation
Zhuowei Zhang (Nankai University), Zhen Zhang (Nankai University)
GenerationGraph Neural NetworkContrastive LearningTextGraph
🎯 What it does: This paper proposes CMGN (Coreference-guided Mind-Map Generation Network) to generate mind maps for documents: it first constructs a coreference graph and encodes sentence relationships through a graph neural network, then utilizes a graph contrastive learning module to enhance the representation of the graph structure, ultimately generating a sentence governance relationship graph.
CoreRec: A Counterfactual Correlation Inference for Next Set Recommendation
Kexin Li (Zhejiang University), Jun Xiao (Zhejiang University)
Recommendation SystemGraph Neural NetworkTabularTime Series
🎯 What it does: This paper proposes the CoreRec framework, which utilizes counterfactual causal reasoning to identify and suppress spurious correlations across sets (cross-basket) in purchase sequences, thereby improving the accuracy of next basket predictions.
Correlation Matching Transformation Transformers for UHD Image Restoration
Cong Wang (Hong Kong Polytechnic University), Jun Liu (National University of Singapore)
RestorationTransformerImage
🎯 What it does: A universal Transformer named UHDformer is proposed for ultra-high-definition (UHD) image denoising, dehazing, and deblurring, achieving effective transformation of high-resolution features to low-resolution space through two modules, DualCMT and ACM, thereby enhancing recovery quality.
Cost Minimization for Equilibrium Transition
Haoqiang Huang (Hong Kong University of Science and Technology), Jie Zhang (Peking University)
Optimization
🎯 What it does: The paper studies the problem of using minimal rewards (subsidies) to encourage players to transition from an initial Nash equilibrium to a better equilibrium in multi-player games, and provides various solution strategies and complexity analyses.
CoSTA: End-to-End Comprehensive Space-Time Entanglement for Spatio-Temporal Video Grounding
Yaoyuan Liang (Tsinghua University), Shao-Lun Huang
RecognitionObject DetectionConvolutional Neural NetworkTransformerVideoText
🎯 What it does: This paper proposes an end-to-end spatiotemporal video localization framework called CoSTA, which integrates spatial and temporal information, utilizes Transformers for multimodal alignment, and predicts event time boundaries and target object boxes simultaneously through 'entangled queries', addressing the issues of inaccurate time predictions and object inconsistencies across frames in traditional methods.
Count What You Want: Exemplar Identification and Few-Shot Counting of Human Actions in the Wild
Yifeng Huang (Stony Brook University), Minh Hoai (Stony Brook University)
RecognitionConvolutional Neural NetworkSupervised Fine-TuningVideoMultimodality
🎯 What it does: This paper proposes a few-shot human action counting framework based on recordable voice commands, utilizing the user reading 'one, two, three' to extract action samples and perform counting.
Counterfactual-Enhanced Information Bottleneck for Aspect-Based Sentiment Analysis
Mingshan Chang (Shenzhen Institutes of Advanced Technology), Ruifeng Xu (Harbin Institute of Technology)
ClassificationRecommendation SystemTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: A Contrastive Experiment Enhanced Information Bottleneck (CEIB) framework is proposed to reduce spurious correlations caused by surface features in Aspect-Based Sentiment Analysis (ABSA) and enhance model robustness.
Coupled Confusion Correction: Learning from Crowds with Sparse Annotations
Hansong Zhang (Institute of Information Engineering, Chinese Academy of Sciences), Shiming Ge (Shanghai University)
Data-Centric LearningMeta LearningImage
🎯 What it does: The study investigates how to learn in a sparsely annotated crowdsourcing environment, proposing the CCC model that corrects the confusion matrix through a dual model collaboration.
Coupling Graph Neural Networks with Fractional Order Continuous Dynamics: A Robustness Study
Qiyu Kang (Nanyang Technological University), Wee Peng Tay (Nanyang Technological University)
OptimizationAdversarial AttackGraph Neural NetworkGraphOrdinary Differential Equation
🎯 What it does: This paper proposes a Fractional Order Differential Equation-based Graph Neural Network (FROND) and systematically evaluates its robustness under graph structure and feature perturbations.