AAAI 2025 Papers — Page 5
AAAI Conference on Artificial Intelligence · 3028 papers
Can LLMs Obfuscate Code? A Systematic Analysis of Large Language Models into Assembly Code Obfuscation
Seyedreza Mohseni (University of Maryland), Manas Gaur (University of Maryland)
AI Code AssistantTransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: This paper investigates whether large language models can generate obfuscated assembly code and proposes the METAMORPHASM benchmark and the MAD dataset for systematic evaluation.
Can LVLMs Obtain a Driver’s License? A Benchmark Towards Reliable AGI for Autonomous Driving
Yuhang Lu (ShanghaiTech University), Xinge Zhu (Shanghai Jiao Tong University)
Autonomous DrivingTransformerLarge Language ModelPrompt EngineeringVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: A large-scale visual language driving knowledge base and benchmark, IDKB, has been proposed and constructed to evaluate and enhance the driving knowledge mastery of 15 large visual language models.
Can Private Machine Learning Be Fair?
Joseph Rance (University of Cambridge), Filip Svoboda (University of Cambridge)
Federated LearningImageTabular
🎯 What it does: This paper proposes a new attack method that deliberately introduces unfairness through malicious clients during the federated learning process, and systematically evaluates the impact of existing robust aggregation methods on fairness.
Can Students Beyond the Teacher? Distilling Knowledge from Teacher’s Bias
Jianhua Zhang (Tianjin University of Technology), Shengyong Chen (Tianjin University of Technology)
ClassificationObject DetectionKnowledge DistillationImage
🎯 What it does: A knowledge distillation strategy based on bias elimination and correction is proposed, enabling the student model to surpass the teacher model.
Can Watermarking Large Language Models Prevent Copyrighted Text Generation and Hide Training Data?
Michael-Andrei Panaitescu-Liess (University of Maryland), Furong Huang (University of Maryland)
GenerationSafty and PrivacyLarge Language ModelText
🎯 What it does: This paper discusses embedding watermarks in large language models (LLMs) to suppress the generation of copyrighted text and evaluates its impact on copyright detection of training data.
Can We Get Rid of Handcrafted Feature Extractors? SparseViT: Nonsemantics-Centered, Parameter-Efficient Image Manipulation Localization Through Spare-Coding Transformer
Lei Su (Sichuan University), Ji-Zhe Zhou (Sichuan University)
Image TranslationAnomaly DetectionComputational EfficiencyTransformerImage
🎯 What it does: This paper proposes SparseViT, an image tampering localization network that achieves feature extraction without manual intervention through a sparse self-attention mechanism.
Capability Instruction Tuning
Yi-Kai Zhang (Nanjing University), Han-Jia Ye (Nanjing University)
Large Language ModelSupervised Fine-TuningTextMultimodality
🎯 What it does: This paper studies a method for instruction-level model routing achieved through capability instruction tuning and proposes the MODEL-SAT framework.
CAPrompt: Cyclic Prompt Aggregation for Pre-Trained Model Based Class Incremental Learning
Qiwei Li (Peking University), Jiahuan Zhou (Peking University)
ClassificationRecognitionTransformerPrompt EngineeringImage
🎯 What it does: A class-incremental learning method based on Circular Prompt Aggregation (CAPrompt) is proposed to eliminate the prompt inconsistency problem caused by task ID prediction.
Capture Global Feature Statistics for One-Shot Federated Learning
Zenghao Guan (Institute of Information Engineering Chinese Academy of Sciences), Xiaoyan Gu (Institute of Information Engineering Chinese Academy of Sciences)
ClassificationFederated LearningConvolutional Neural NetworkImage
🎯 What it does: Utilizing pre-trained models to extract features, clients upload statistical data for each type of feature, and the server aggregates to obtain global mean and covariance, directly constructing a Gaussian Naive Bayes classifier, thus achieving global and personalized federated learning with a single communication.
Capturing the Unseen: Vision-Free Facial Motion Capture Using Inertial Measurement Units
Youjia Wang (ShanghaiTech University), Jingyi Yu (ShanghaiTech University)
Pose EstimationTransformerDiffusion modelVideoMultimodalityTime Series
🎯 What it does: This paper presents CAPUS, a facial motion capture system that is completely based on inertial measurement units (IMUs) and can achieve high-precision expression reconstruction under conditions without visual signals.
CaRDiff: Video Salient Object Ranking Chain of Thought Reasoning for Saliency Prediction with Diffusion
Yunlong Tang (ByteDance), Chenliang Xu (University of Rochester)
Object DetectionSegmentationLarge Language ModelDiffusion modelVideoMultimodalityChain-of-Thought
🎯 What it does: Designed the CaRDiff framework, utilizing multimodal large language models and diffusion models to generate ranking maps through video titles and object ranking chain reasoning, thereby enhancing video saliency prediction;
CareBot: A Pioneering Full-Process Open-Source Medical Language Model
Lulu Zhao (Beijing Academy of Artificial Intelligence), Hua Zhou (Beijing University of Posts and Telecommunications)
Reinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextBiomedical Data
🎯 What it does: A bilingual medical large model, CareBot, has been developed, employing a two-stage continuous pre-training, supervised fine-tuning, and reinforcement learning based on human feedback, aimed at enhancing medical diagnosis, treatment planning, and medical education capabilities.
Cascaded Diffusion Models for Virtual Try-On: Improving Control and Resolution
Guangyuan Li (Zhejiang University), Binkai Ou (BoardWare Information System Limited)
GenerationSuper ResolutionDiffusion modelImage
🎯 What it does: This paper proposes a Cascade Diffusion Model (CDM-VTON) that generates low-resolution try-on images while preserving clothing details through a Multi-Conditional Diffusion Model (MC-DM), and then enhances them to 2K resolution using a Super-Resolution Diffusion Model (SR-DM).
CasFT: Future Trend Modeling for Information Popularity Prediction with Dynamic Cues-Driven Diffusion Models
Xin Jing (University of Macau), Dingqi Yang (University of Macau)
Diffusion modelTime SeriesSequentialOrdinary Differential Equation
🎯 What it does: A model named CasFT has been developed, which utilizes observed propagation cascades, neural ODE to predict growth rates, and diffusion models to generate future trends, combined with spatiotemporal features to achieve content popularity prediction.
CASUAL: Conditional Support Alignment for Domain Adaptation with Label Shift
Anh T Nguyen (University of Illinois Chicago), Toan Tran (VinAI Research)
Domain AdaptationImage
🎯 What it does: This paper proposes an unsupervised domain adaptation method under label shift conditions, called CASUAL, which learns more discriminative feature representations by aligning the conditional feature supports of the source and target domains.
Category Prompt Mamba Network for Nuclei Segmentation and Classification
Ye Zhang (Harbin Institute of Technology), Yongbing Zhang (Harbin Institute of Technology)
ClassificationSegmentationImageBiomedical Data
🎯 What it does: A category prompt network based on Mamba (CP-Mamba) is proposed, which enables direct nuclear cell segmentation and classification on complete large-size pathological images, significantly improving the recognition of minority classes through probability-guided sequential sorting.
Cauchy Diffusion: A Heavy-tailed Denoising Diffusion Probabilistic Model for Speech Synthesis
Qi Lian (Zhejiang University), Yueming Wang (Zhejiang University)
GenerationData SynthesisDiffusion modelAudio
🎯 What it does: This paper proposes Cauchy Diffusion, a heavy-tailed denoising diffusion model that introduces Cauchy noise into DDPM to enhance prosodic diversity in speech synthesis.
Causal Discovery by Interventions via Integer Programming
Abdelmonem Elrefaey (Arizona State University), Rong Pan (Arizona State University)
OptimizationTabular
🎯 What it does: This paper proposes a framework based on Integer Programming (IP) to design the minimum number of intervention experiment sets, ensuring that the causal structure can be fully identified under given hypotheses.
Causal Inference over Visual-Semantic-Aligned Graph for Image Classification
Lei Meng (Shandong University), Xiangxu Meng (Shandong University)
ClassificationRetrievalGraph Neural NetworkImage
🎯 What it does: Through fine-grained visual-semantic alignment and causal reasoning, an end-to-end graphical model is constructed that can locate regions in images corresponding to labels under borderless supervision, and utilizes pre-learned visual-semantic prototypes to filter out incorrect labels, ultimately improving image classification performance.
Causal Prompting: Debiasing Large Language Model Prompting Based on Front-Door Adjustment
Congzhi Zhang (Southeast University), Deyu Zhou (Southeast University)
Large Language ModelPrompt EngineeringContrastive LearningTextChain-of-Thought
🎯 What it does: Using causal debiasing through front-door adjustment at the prompt level, chain reasoning is employed as a mediating variable to alleviate model bias.
Causal-Inspired Multitask Learning for Video-Based Human Pose Estimation
Haipeng Chen (Jilin University), Zhenguang Liu (Zhejiang University)
Pose EstimationTransformerVideo
🎯 What it does: This paper proposes a causal inference-based multi-task learning framework CM-Pose for video human pose estimation. The framework is divided into two stages: the first stage enhances the network's causal spatiotemporal modeling ability by introducing two self-supervised auxiliary tasks, namely masking reconstruction and denoising; the second stage enhances coarse-grained features into fine-grained features through token causal importance selection and non-causal token clustering modules, improving the model's interpretability and robustness.
Causally Consistent Normalizing Flow
Qingyang Zhou (University of Waterloo), Meng Xu (University of Waterloo)
Flow-based ModelTabularFinance Related
🎯 What it does: A multi-layer causal consistent normalization flow model (CCNF) is proposed, achieving universal approximation under complex distributions.
CDE-Learning: Camera Deviation Elimination Learning for Unsupervised Person Re-identification
Jinjia Peng (Hebei University), Huibing Wang (Dalian Maritime University)
RecognitionRetrievalContrastive LearningImage
🎯 What it does: Proposes the CDE-Learning framework, achieving unsupervised person re-identification through camera bias elimination learning.
CDTR: Semantic Alignment for Video Moment Retrieval Using Concept Decomposition Transformer
Ran Ran (University of Electronic Science and Technology of China), Heng Tao Shen (Tongji University)
RetrievalTransformerContrastive LearningVideoTextMultimodality
🎯 What it does: This paper proposes a Concept Decomposition-based Transformer (CDTR) model to address the Video Moment Retrieval task. The core idea is to decompose video segments and text queries into semantic concept representations and use these concepts as pseudo-labels for fine-grained cross-modal and intra-video alignment.
CDW-CoT: Clustered Distance-Weighted Chain-of-Thoughts Reasoning
Yuanheng Fang (Harbin Institute of Technology), Dianhui Chu (Sichuan University)
TransformerLarge Language ModelPrompt EngineeringTextChain-of-Thought
🎯 What it does: The CDW-CoT method is proposed, which utilizes clustering to generate a specific prompt pool and dynamically constructs a prompt distribution through distance weighting to enhance the reasoning performance of large language models.
Certification of Speaker Recognition Models to Additive Perturbations
Dmitrii Korzh (AIRI), Ivan Oseledets (AIRI)
RecognitionAdversarial AttackContrastive LearningAudio
🎯 What it does: This paper proposes a robustness certification method for speaker recognition models based on Randomized Smoothing, which can provide a provable radius regarding the invariance of model predictions under additive perturbations constrained by the L₂ norm for a given audio sample.
Certified Causal Defense with Generalizable Robustness
Yiran Qiao (Case Western Reserve University), Jing Ma (Case Western Reserve University)
Domain AdaptationGaussian SplattingImage
🎯 What it does: This paper proposes a framework named GLEAN, aimed at achieving cross-domain generalizable and verifiable defenses from a causal perspective.
Certifying Bounds Propagation for Integer Multiplication Constraints
Matthew J. McIlree (University of Glasgow), Ciaran McCreesh (University of Glasgow)
Tabular
🎯 What it does: Log the proof of bounds-consistency propagation for integer multiplication constraints, providing pseudo-Boolean (PB) proof steps that can be checked in the VeriPB verifier.
CFDM: Contrastive Fusion and Disambiguation for Multi-View Partial-Label Learning
Qiuru Hai (Beijing University of Technology), Gengyu Lyu (Beijing Technology Co., Ltd.)
ClassificationAuto EncoderContrastive LearningImage
🎯 What it does: This paper proposes a Multi-View Partial Label Learning (MVPLL) framework called CFDM, which achieves accurate classification of samples containing candidate labels but only one true label in multiple views by comparing and fusing multi-view features and using multi-class contrast prototypes for disambiguation.
CG-TGAN: Conditional Generative Adversarial Networks with Graph Neural Networks for Tabular Data Synthesizing
Seungcheol Lee (Sungkyunkwan University), Moohong Min (Sungkyunkwan University)
Data SynthesisSafty and PrivacyGraph Neural NetworkGenerative Adversarial NetworkTabular
🎯 What it does: A conditional GAN based on graph neural networks (CG-TGAN) is proposed for generating high-quality, privacy-friendly tabular data.
Chain-of-Instructions: Compositional Instruction Tuning on Large Language Models
Shirley Anugrah Hayati (University of Minnesota), Dongyeop Kang (Amazon)
TransformerLarge Language ModelSupervised Fine-TuningTextChain-of-Thought
🎯 What it does: This paper introduces the concept of Chain-of-Instructions (CoI) and enhances the model's understanding and execution capabilities for multi-subtask combination instructions by fine-tuning using CoI instructions on large language models.
ChangeDiff: A Multi-Temporal Change Detection Data Generator with Flexible Text Prompts via Diffusion Model
Qi Zang (Xidian University), Zhun Zhong (Xidian University)
SegmentationGenerationData SynthesisDiffusion modelImage
🎯 What it does: This paper proposes ChangeDiff, a multi-temporal semantic change detection data generator that utilizes diffusion models to generate high-quality change images and layouts without the need for paired images or additional annotations.
Channel Merging: Preserving Specialization for Merged Experts
Mingyang Zhang (Zhejiang University), Bohan Zhuang (Zhejiang University)
Mixture of ExpertsTextBenchmark
🎯 What it does: Proposes the Channel Merging method, which clusters and merges the Delta parameters of expert models at the channel level, significantly reducing parameter conflicts while maintaining the expertise of each expert.
CharacterBench: Benchmarking Character Customization of Large Language Models
Jinfeng Zhou (Tsinghua University), Minlie Huang
OptimizationTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningTextBenchmark
🎯 What it does: This paper proposes and implements CHARACTERBENCH—a generative evaluation benchmark covering 3,956 characters and 22,859 bilingual dialogue samples, and develops a specialized judging model called CharacterJudge.
Characterising Simulation-Based Program Equilibria
Emery Cooper (Carnegie Mellon University), Vincent Conitzer (Carnegie Mellon University)
🎯 What it does: This study proposes a new class of simulation-based program equilibria (a generalization of ε-Grounded π Bot) and proves that under shared randomness, the complete folk theorem can be achieved, while under no shared randomness, a broader equilibrium can be realized, although it still does not cover Tennenholtz's folk theorem.
ChatterBox: Multimodal Referring and Grounding with Chain-of-Questions
Yunjie Tian (University of Chinese Academy of Sciences), Qixiang Ye (University of Chinese Academy of Sciences)
RecognitionObject DetectionTransformerSupervised Fine-TuningVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: A multi-modal continuous reference and localization task (MCQ) and the corresponding benchmark dataset CB-300K have been constructed, with the ChatterBox model proposed as a baseline, and evaluation metrics for chain-based question answering have been provided.
ChatTime: A Unified Multimodal Time Series Foundation Model Bridging Numerical and Textual Data
Chengsen Wang (Beijing University of Posts and Telecommunications), Jianxin Liao
TransformerLarge Language ModelSupervised Fine-TuningMultimodalityTime Series
🎯 What it does: ChatTime is proposed, a multimodal temporal foundation model that treats time series data as a foreign language and utilizes LLMs, supporting zero-shot prediction, context-assisted prediction, and temporal question answering.
Checking Consistency of CP-Theory Preferences in Polynomial Time
Erik Rauer (Iowa State University), Vasant Honavar (Pennsylvania State University)
🎯 What it does: This paper proposes a new sufficient condition—cardinality-based condition acyclicity (cc-acyclic)—for quickly detecting preference consistency in CP-theory and constructing a complete search tree (cs-tree) that satisfies preferences;
ChemVLM: Exploring the Power of Multimodal Large Language Models in Chemistry Area
Junxian Li (Shanghai Artificial Intelligence Laboratory), Dongzhan Zhou (Shanghai Artificial Intelligence Laboratory)
TransformerLarge Language ModelSupervised Fine-TuningVision Language ModelImageTextMultimodality
🎯 What it does: ChemVLM, a multimodal large language model for the field of chemistry, has been developed to simultaneously process chemical images and text information.
Cherry-Picking in Time Series Forecasting: How to Select Datasets to Make Your Model Shine
Luis Roque (Universidade do Porto), Luís Torgo
Time SeriesBenchmark
🎯 What it does: This paper proposes a framework to assess the performance bias caused by dataset selection (cherry-picking) in time series forecasting experiments, demonstrating that a small number of non-representative datasets can significantly exaggerate model performance.
Cirbo: A New Tool for Boolean Circuit Analysis and Synthesis
Daniil Averkov (St. Petersburg State University), Aleksey Vorobiev (Neapolis University Pafos)
OptimizationBenchmark
🎯 What it does: This paper presents the open-source tool Cirbo, designed for the analysis (such as satisfiability checking) and synthesis (such as minimum circuit solving) of Boolean circuits, achieving efficient circuit optimization and size reduction through various algorithms.
Citations and Trust in LLM Generated Responses
Yifan Ding (University of Notre Dame), Tim Weninger (University of Notre Dame)
TransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: This study evaluates the impact of citations (0, 1, 5 citations) on the credibility of responses from large language models through a designed RCT experiment, and observes the relationship between users' citation retrieval behavior and their self-reported trust levels.
CITI: Enhancing Tool Utilizing Ability in Large Language Models Without Sacrificing General Performance
Yupu Hao (Institute of Automation Chinese Academy of Sciences), Jun Zhao (Institute of Automation Chinese Academy of Sciences)
TransformerLarge Language ModelSupervised Fine-TuningMixture of ExpertsText
🎯 What it does: A framework named CITI has been developed, combining Mixture-of-LoRA and selective full parameter fine-tuning, significantly enhancing the tool invocation capability while maintaining the general performance of large language models.
CiTrus: Squeezing Extra Performance out of Low-data Bio-signal Transfer Learning
Eloy Geenjaar (Georgia Institute of Technology), Lie Lu (Dolby Laboratories)
ClassificationRecognitionDomain AdaptationConvolutional Neural NetworkTransformerAuto EncoderMultimodalityTime SeriesBiomedical DataElectrocardiogram
🎯 What it does: A convolution-Transformer hybrid architecture called CiTrus is proposed, combining frequency-aware masked autoencoding and multimodal pretraining to enhance the performance of low-sample transfer learning for biological signals.
CL-Attack: Textual Backdoor Attacks via Cross-Lingual Triggers
Jingyi Zheng (Hong Kong University of Science and Technology), Xinlei He (Hong Kong University of Science and Technology)
Adversarial AttackTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: A text backdoor attack method using cross-language structural triggers at the paragraph level is proposed, named CL-Attack.
Class and Attribute-Aware Logit Adjustment for Generalized Long-Tail Learning
Xiaoling Zhou (Tianjin University), Nan Yang (Tianjin University)
ClassificationMeta LearningConvolutional Neural NetworkImage
🎯 What it does: This paper proposes a new class attribute-aware logarithmic loss (CALA) that addresses both class imbalance and attribute imbalance issues, and presents two schemes for estimating the class conditional probability density ratio between training and testing (Heuristic-CALA and Meta-CALA).
Class Semantic Attribute Perception Guided Zero-Shot Learning
Qin Yue (Shanxi University), Liang Bai (Shanxi University)
ClassificationRecognitionTransformerContrastive LearningImage
🎯 What it does: A zero-shot learning method that simultaneously considers coarse-grained and fine-grained class semantic attribute awareness (CSAP-ZSL) is proposed, achieving precise alignment between image regions and semantic attributes through graph cut soft region partitioning and attribute prototype matching.
Clean-Label Graph Backdoor Attack in the Node Classification Task
Hui Xia (Ocean University of China), Luming Wang (Ocean University of China)
ClassificationAdversarial AttackGraph Neural NetworkGraph
🎯 What it does: A Clean-Label Graph Backdoor Attack (CGBA) is proposed, which does not alter labels in node classification tasks, selects target class nodes with high uncertainty and low degree under a low budget, and uses feature triggers to induce model misclassification.
CLEP: A Novel Contrastive Learning Method for Evolutionary Reentrancy Vulnerability Detection
Jie Chen (Southeast University), Victor S. Sheng (Texas Tech University)
Anomaly DetectionGraph Neural NetworkContrastive LearningGraph
🎯 What it does: The CLEP framework is proposed to detect reentrancy vulnerabilities by comparing learning through version evolution.
Click2Mask: Local Editing with Dynamic Mask Generation
Omer Regev (Hebrew University of Jerusalem), Dani Lischinski (Hebrew University of Jerusalem)
Image TranslationGenerationDiffusion modelImage
🎯 What it does: By dynamically evolving masks based on the Alpha-CLIP semantic gradient during the Blended Latent Diffusion (BLD) process, new content can be added at specified locations with just a single click by the user on the image.
CLIMB-ReID: A Hybrid CLIP-Mamba Framework for Person Re-Identification
Chenyang Yu (Dalian University of Technology), Huchuan Lu (Dalian University of Technology)
RecognitionRetrievalTransformerVision Language ModelContrastive LearningImageVideo
🎯 What it does: A new method called CLIMB-ReID is proposed, which integrates the CLIP visual-language model with the Mamba lightweight sequence modeling framework for person re-identification.
CLIP-CID: Efficient CLIP Distillation via Cluster-Instance Discrimination
Kaicheng Yang (DeepGlint), Jiankang Deng (Imperial College)
Computational EfficiencyKnowledge DistillationTransformerVision Language ModelContrastive LearningImageMultimodality
🎯 What it does: The CLIP-CID distillation method is proposed, which first removes 43.7% of redundant image-text pairs in LAION-400M through a one-time semantic balance filtering, and then combines cluster-instance discriminative distillation to efficiently transfer knowledge from a large-scale vision-language model to a smaller model.
CLIP-driven View-aware Prompt Learning for Unsupervised Vehicle Re-identification
Jiyang Xu (Nanchang University), Dong Wang (Nanchang University)
RecognitionRetrievalAutonomous DrivingGraph Neural NetworkPrompt EngineeringContrastive LearningImage
🎯 What it does: A view-aware prompt learning framework based on CLIP (ViewCoOp) and a cross-modal mutual graph matching method are proposed for unsupervised vehicle re-identification tasks.
CLIP-MSM: A Multi-Semantic Mapping Brain Representation for Human High-Level Visual Cortex
Guoyuan Yang (Beijing Institute of Technology), Xuesong Li (Beijing Institute of Technology)
Representation LearningTransformerVision Language ModelContrastive LearningImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: By combining the CLIP model with CLIP Dissection, a multi-semantic mapping framework called CLIP-MSM was constructed, capable of voxel-level predictions and multi-semantic interpretations of the human high-level visual cortex without prior assumptions.
CLIP-PCQA: Exploring Subjective-Aligned Vision-Language Modeling for Point Cloud Quality Assessment
Yating Liu (Shanghai Jiao Tong University), Yiling Xu (Shanghai Jiao Tong University)
ClassificationRetrievalTransformerPrompt EngineeringVision Language ModelContrastive LearningPoint Cloud
🎯 What it does: This paper proposes a no-reference point cloud quality assessment method based on CLIP, called CLIP-PCQA, which aligns visual features with multi-level quality descriptions using a language retrieval mapping and predicts opinion distribution (OSD) to provide a final quality score.
CLIP-RestoreX: Restore Image Structure and Perception in Exposure Correction
Xiang Huang (Sun Yat-sen University), Wei-Shi Zheng (Sun Yat-sen University)
RestorationTransformerDiffusion modelImage
🎯 What it does: This paper proposes an exposure correction method called CLIP-RestoreX based on the CLIP structure and perceptual priors. It utilizes the shallow structural features and deep perceptual features of CLIP to construct priors, enhances the damaged features through a frequency domain diffusion model (FFEDM), and finally injects the enhanced priors into the Restormer network to achieve exposure correction.
CLNX: Bridging Code and Natural Language for C/C++ Vulnerability-Contributing Commits Identification
Zeqing Qin (Huazhong University of Science and Technology), Lansheng Han (Huazhong University of Science and Technology)
RecognitionAI Code AssistantTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: By linearizing and naturalizing C/C++ source code and patch submissions before inputting them into LLM, efficient identification of vulnerability-related submissions is achieved.
Cluster Based Heterogeneous Federated Foundation Model Adaptation and Fine-Tuning
Xianda Wang (Chinese University of Hong Kong), Fangxin Wang (Chinese University of Hong Kong)
Federated LearningKnowledge DistillationNeural Architecture SearchImage
🎯 What it does: This paper proposes FedCKMS, a hierarchical clustering federated learning framework designed for resource-constrained edge devices, aimed at efficient fine-tuning across various heterogeneous base models.
Cluster-guided Contrastive Class-imbalanced Graph Classification
Wei Ju (Sichuan University), Ming Zhang (Peking University)
ClassificationRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: This paper proposes C3 GNN, a graph classification framework that combines adaptive subclass clustering, Mixup synthetic samples, and hierarchical supervised contrastive learning to address the class imbalance problem in graph data.
Clustering by Mining Density Distributions and Splitting Manifold Structure
Zhichang Xu (Southwest Jiaotong University), Hua Meng (Southwest Jiaotong University)
Supervised Fine-TuningTabular
🎯 What it does: A micro-cluster (pseudo-cluster) construction method based on local density distribution and manifold curvature is proposed, and the splitting rules are improved on this basis to accelerate and enhance the performance of spectral clustering.
CMT: A Memory Compression Method for Continual Knowledge Learning of Large Language Models
Dongfang Li (Harbin Institute of Technology), Min Zhang (Harbin Institute of Technology)
CompressionKnowledge DistillationTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Designed and implemented the CMT (Compression Memory Training) method, which achieves online knowledge adaptation and continuous learning by compressing new document information into a latent space memory bank while keeping the LLM parameters unchanged.
CNC: Cross-modal Normality Constraint for Unsupervised Multi-class Anomaly Detection
Xiaolei Wang (Xi'an Jiaotong-Liverpool University), Jimin Xiao (Beijing Jiaotong University)
Anomaly DetectionTransformerMixture of ExpertsVision Language ModelImageMultimodality
🎯 What it does: A reverse distillation framework based on cross-modal regularization is proposed, utilizing learnable prompts in the visual-language model to guide the decoder to only reconstruct normal features, thereby suppressing overgeneralization (OG) caused by multi-class training and enhancing anomaly detection and localization performance.
Co-Dream: Collaborative Dream Synthesis over Decentralized Models
Abhishek Singh (Massachusetts Institute of Technology), Ramesh Raskar (Amazon)
Federated LearningKnowledge DistillationImage
🎯 What it does: The Co-Dream framework is proposed, which achieves knowledge sharing in federated learning by collaboratively optimizing 'dreams' in the data space, avoiding the sharing of model parameters.
Co-Progression Knowledge Distillation with Knowledge Prototype for Industrial Anomaly Detection
Bokang Yang (Huazhong University of Science and Technology), Jie Ma (Huazhong University of Science and Technology)
Anomaly DetectionKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: A Co-Progression with Knowledge Prototype (CPKP) framework is proposed, achieving bidirectional learning between teacher and student models in unsupervised industrial defect detection, while maintaining the core knowledge of the teacher model through knowledge prototypes to enhance detection performance.
Cobra: Extending Mamba to Multi-Modal Large Language Model for Efficient Inference
Han Zhao (Westlake University), Donglin Wang (Westlake University)
Computational EfficiencyRepresentation LearningTransformerLarge Language ModelVision Language ModelImageMultimodality
🎯 What it does: A multimodal large language model named Cobra is proposed, which integrates the Mamba state space model with a visual encoder to achieve efficient inference.
CoCoCo: Improving Text-Guided Video Inpainting for Better Consistency, Controllability and Compatibility
Bojia Zi (Chinese University of Hong Kong), Lei Zhang (IntelliFusion Inc)
RestorationGenerationDiffusion modelVideoText
🎯 What it does: This paper proposes the CoCoCo framework, which implements text-guided video inpainting tasks.
Code-switching Mediated Sentence-level Semantic Learning
Shuai Zhang (Tsinghua University), Ruibo Fu (Chinese Academy of Sciences)
TransformerLarge Language ModelSupervised Fine-TuningTextAudio
🎯 What it does: A task-agnostic sentence-level semantic learning method is proposed, which achieves semantic sharing of mixed expressions in different languages through a semantic invariance constraint (semantic invariance loss), constructing a unified end-to-end ASR and AST model.
CoDe: Communication Delay-Tolerant Multi-Agent Collaboration via Dual Alignment of Intent and Timeliness
Shoucheng Song (Beijing Jiaotong University), Kai Lv (Beijing Jiaotong University)
Recurrent Neural NetworkReinforcement LearningSequentialBenchmark
🎯 What it does: The CoDe framework is proposed to address the asynchronous collaboration problem of multi-agent systems in the presence of communication delays.
CODE: Confident Ordinary Differential Editing
Bastien Van Delft, Alexandre Alahi (Ecole Polytechnique Federale de Lausanne)
RestorationGenerationDiffusion modelImageOrdinary Differential Equation
🎯 What it does: Using a pre-trained diffusion model for unsupervised repair and editing of out-of-distribution (OoD) images, generating images that retain input details while achieving high realism.
Codec Does Matter: Exploring the Semantic Shortcoming of Codec for Audio Language Model
Zhen Ye (Hong Kong University of Science and Technology), Wei Xue (Hong Kong University of Science and Technology)
GenerationCompressionTransformerLarge Language ModelAudio
🎯 What it does: This study investigates the semantic inadequacy of audio LLMs when using traditional compression codecs (such as EnCodec) and proposes X-Codec, which incorporates pre-trained semantic features before and after residual vector quantization (RVQ) and introduces a semantic reconstruction loss to enhance the semantic understanding and generation quality of audio LLMs.
CodecNeRF: Toward Fast Encoding and Decoding, Compact, and High-quality Novel-view Synthesis
Gyeongjin Kang (Sungkyunkwan University), Eunbyung Park (Sungkyunkwan University)
GenerationData SynthesisCompressionComputational EfficiencyNeural Radiance FieldAuto EncoderPoint CloudMesh
🎯 What it does: Introducing CodecNeRF - a neural encoder-decoder-fine-tuning pipeline that can generate NeRF representations in a single forward pass, achieving fast encoding, extremely low bit rates, and high-quality view synthesis through parameter-efficient fine-tuning.
CodeHalu: Investigating Code Hallucinations in LLMs via Execution-based Verification
Yuchen Tian (Hong Kong Baptist University), Dawn Song (University of California, Berkeley)
AI Code AssistantTransformerLarge Language ModelTextBenchmark
🎯 What it does: This paper defines the concept of 'code hallucination' in code generation, proposes a dynamic detection algorithm called CodeHalu based on execution verification, and constructs a large-scale evaluation benchmark named CodeHaluEval, followed by a systematic evaluation of 17 mainstream large language models.
CoDeR: Counterfactual Demand Reasoning for Sequential Recommendation
Shuai Tang (Harbin Institute of Technology), Xiaofeng Zhang (Harbin Institute of Technology)
Recommendation SystemGraph Neural NetworkTransformerReinforcement LearningSequential
🎯 What it does: Proposes the CoDeR framework to address demand drift identification and denoising in sequential recommendation.
CoDTS: Enhancing Sparsely Supervised Collaborative Perception with a Dual Teacher-Student Framework
Yushan Han (Beijing Jiaotong University), Yidong Li (Beijing Jiaotong University)
Object DetectionAutonomous DrivingPoint Cloud
🎯 What it does: An end-to-end collaborative perception dual teacher-student framework CoDTS is proposed for 3D object detection under sparse supervision, generating high-quality and sufficiently numerous pseudo-labels through adaptive complementary learning.
CoffeeBoost: Gradient Boosting Native Conformal Inference for Bayesian Optimization
Yuanhao Lai (Huawei Technologies), Yunfei Du (Huawei Technologies)
OptimizationReinforcement LearningTabularBenchmark
🎯 What it does: This paper proposes CoffeeBoost, a Bayesian optimization framework that combines gradient boosting trees (GBDT) with a distribution-free conformal inference for the automatic tuning of software system parameters, especially for databases.
CognitionCapturer: Decoding Visual Stimuli from Human EEG Signal with Multimodal Information
Kaifan Zhang (Xidian University), Xinbo Gao (Xidian University)
GenerationData SynthesisOptimizationTransformerDiffusion modelContrastive LearningMultimodality
🎯 What it does: The CognitionCapturer framework is proposed, which utilizes a multimodal expert encoder for cross-modal alignment of EEG and maps the EEG embedding into the CLIP space through a diffusion prior, thereby reconstructing high-fidelity visual stimuli without the need to fine-tune the generative model.
Cognitive Fluctuations Enhanced Attention Network for Knowledge Tracing
Mingliang Hou (Guangdong Institute of Smart Education), Weiqi Luo (TAL Education Group)
TabularTime SeriesSequential
🎯 What it does: The FlucKT framework is proposed, utilizing cognitive feature decomposition layers and kernelized bias attention to enhance the knowledge tracing model's ability to model short-term cognitive fluctuations and length generalization.
CognTKE: A Cognitive Temporal Knowledge Extrapolation Framework
Wei Chen (Beijing Jiaotong University), Huaiyu Wan (Beijing Jiaotong University)
Explainability and InterpretabilityKnowledge DistillationGraph Neural NetworkReinforcement LearningGraphTime Series
🎯 What it does: This study proposes the CognTKE framework, which infers temporal knowledge graphs based on dual-process theory, combining global shallow reasoning with local deep reasoning.
CogSQL: A Cognitive Framework for Enhancing Large Language Models in Text-to-SQL Translation
Hongwei Yuan (Zhejiang University), Huan Li (Zhejiang University)
TransformerLarge Language ModelPrompt EngineeringTextChain-of-Thought
🎯 What it does: The COGSQL framework is proposed to enhance the text-to-SQL translation capability of LLMs by simulating human cognitive processes.
Coherency Improved Explainable Recommendation via Large Language Model
Shijie Liu (East China Normal University), Wei Zhang (WeChat AI Tencent)
Recommendation SystemExplainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Utilize a large language model to first predict ratings, then map the ratings to word vectors, and finally generate interpretable recommendations consistent with the ratings using this vector along with user-item information.
CohEx: A Generalized Framework for Cohort Explanation
Fanyu Meng (University of California), Xin Chen (Georgia Institute of Technology)
Explainability and InterpretabilityImageTabular
🎯 What it does: This paper presents CohEx, a general framework that transforms existing local feature importance explanation methods into group explanations, enhancing the interpretability of group explanations through supervised clustering and iterative recalculation of importance.
Collaborative Evolution: Multi-Round Learning Between Large and Small Language Models for Emergent Fake News Detection
Ziyi Zhou (Beihang University), Chaozhuo Li (Beihang University)
ClassificationRetrievalTransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: A multi-round collaborative detection framework (MRCD) is proposed, allowing large language models (LLM) and small language models (SLM) to jointly detect emerging fake news through retrieval enhancement and complementary pseudo-labeling.
Collaborative Learning for 3D Hand-Object Reconstruction and Compositional Action Recognition from Egocentric RGB Videos Using Superquadrics
Tze Ho Elden Tse (University of Birmingham), Hyung Jin Chang (Dongguk University)
RecognitionPose EstimationTransformerVideo
🎯 What it does: A collaborative learning framework is proposed, combining 3D hand-object reconstruction with action recognition from perspective RGB videos.
Collaborative Semantic Consistency Alignment for Blended-Target Domain Adaptation
Yuwu Lu (South China Normal University), Haoyu Huang (South China Normal University)
ClassificationDomain AdaptationGraph Neural NetworkImage
🎯 What it does: The Collaborative Semantic Consistency Alignment (CSCA) method is proposed, which combines slice Wasserstein distance distribution alignment, graph neural network cross-domain semantic consistency alignment, and dual consistency regularization to address the mixed target domain adaptation (BTDA) problem.
Collaborative Similarity Fusion and Consistency Recovery for Incomplete Multi-view Clustering
Bingbing Jiang (Hangzhou Normal University), Weiguo Sheng (Nantong University)
OptimizationGraph Neural NetworkTabular
🎯 What it does: This paper proposes a method for Synergistic Similarity Fusion and Consistency Recovery (SFCR) to address the problem of incomplete multi-view clustering.
CollageNoter: Real-Time and Adaptive Collage Layout Design for Screenshot-Based E-Note-Taking
Qiuyun Zhang (Northwestern Polytechnical University), Zhiwen Yu (Northwestern Polytechnical University)
GenerationOptimizationTransformerImageMultimodality
🎯 What it does: A real-time adaptive collage layout system named CollageNoter is proposed for screenshot-based electronic notes.
Color Transfer with Modulated Flows
Maria Larchenko (Skolkovo Institute of Science and Technology), Vladimir Vladimirovich Palyulin (Skolkovo Institute of Science and Technology)
Image TranslationRestorationFlow-based ModelRectified FlowImageOrdinary Differential Equation
🎯 What it does: The paper proposes a color transfer method based on reversible regularized flows (ModFlows), which maps the color distribution of the target image to that of the reference image in the RGB color space.
COLUMBUS: Evaluating COgnitive Lateral Understanding Through Multiple-Choice reBUSes
Koen Kraaijveld (Vrije Universiteit Amsterdam), Filip Ilievski (Vrije Universiteit Amsterdam)
TransformerLarge Language ModelVision Language ModelTextMultimodalityBenchmark
🎯 What it does: A COLUMBUS visual lateral thinking benchmark is proposed, which evaluates the performance of VLM in creative reasoning by automatically generating multiple-choice puzzles through automated rule generation.
Column-Oriented Datalog on the GPU
Yihao Sun (Syracuse University), Kristopher Micinski (Washington State University)
GraphTabularBenchmark
🎯 What it does: FVLOG has been implemented, a columnar Datalog engine based on NVIDIA H100 GPU, supporting columnar storage, dual indexing (sorting + hashing), and parallel relational algebra operations, completing the full reasoning process from rules to fixed points.
Combating Multimodal LLM Hallucination via Bottom-Up Holistic Reasoning
Shengqiong Wu (National University of Singapore), Tat-Seng Chua (National University of Singapore)
TransformerLarge Language ModelVision Language ModelTextMultimodalityChain-of-Thought
🎯 What it does: An untrained bottom-up to top-down reasoning framework is proposed, systematically reducing hallucinations in multimodal LLMs through structured scene graphs, input text conflict detection and correction, and common-sense level validation.
Combating Semantic Contamination in Learning with Label Noise
Wenxiao Fan (Beijing Institute of Technology), Kan Li (Beijing Institute of Technology)
ClassificationData-Centric LearningContrastive LearningImage
🎯 What it does: This study addresses the issue of semantic pollution under label noise and proposes a collaborative cross-learning method to reconstruct labels and enhance robustness.
Combining Priors with Experience: Confidence Calibration Based on Binomial Process Modeling
Jinzong Dong (Central South University), Haoyang Yu (Central South University)
ClassificationConvolutional Neural NetworkImage
🎯 What it does: A method is proposed to model using a binomial process and employ maximum likelihood estimation to fuse prior distributions with empirical data, resulting in a continuous confidence calibration curve.
COMM: Concentrated Margin Maximization for Robust Document-Level Relation Extraction
Zhichao Duan (Tsinghua University), Jianyong Wang (Beijing Institute of Technology)
OptimizationGraph Neural NetworkLarge Language ModelText
🎯 What it does: A COMM framework is proposed, achieving robust optimization for document-level relation extraction through a two-stage method of Instance-Aware Reasoning Enhancement (IARA) and Centralized Margin Maximization (CMM).
COMMIT: Certifying Robustness of Multi-Sensor Fusion Systems Against Semantic Attacks
Zijian Huang (University of Michigan), Bo Li (California Institute of Technology)
Object DetectionAutonomous DrivingMultimodality
🎯 What it does: Proposes the COMMIT framework for robustness certification of multi-sensor fusion systems under semantic transformations such as rotation and translation;
Commitment to Sparse Strategies in Two-Player Games
Salam Afiouni (Columbia University), Christian Kroer (Columbia University)
OptimizationReinforcement LearningGraph
🎯 What it does: This paper studies the design and solution of sparse strategies (k-sparse commitment) in two-player games, proposing a general solution framework based on Mixed Integer Linear Programming (MILP) and extending it to zero-sum games, general and Stackelberg equilibria, structured sparse constraints, and large action spaces; extensive experiments validate that this method can achieve over 90% of the original Nash value while maintaining efficiency, outperforming traditional k-uniform sparse strategies.
Common Sense Bias Modeling for Classification Tasks
Miao Zhang (New York University), Gaurav Bharaj (Reality Defender Inc)
ClassificationImageText
🎯 What it does: Utilizing textual descriptions (image captions) to mine common-sense related features in the dataset, constructing a feature set and calculating their pairwise correlations, thereby discovering and eliminating implicit biases in visual models.
Community-Aware Variational Autoencoder for Continuous Dynamic Networks
Junwei Cheng (South China Normal University), Yong Tang (South China Normal University)
Graph Neural NetworkAuto EncoderGraph
🎯 What it does: A variational autoencoder for continuous dynamic networks, CT-VAE, and its scalable variant, CT-CAVAE, are proposed for community detection.
Community-Centric Graph Unlearning
Yi Li (Guangxi Normal University), Debo Cheng (China University of Mining and Technology)
Computational EfficiencyData-Centric LearningGraph Neural NetworkGraph
🎯 What it does: This paper proposes a graph de-learning framework GSMU based on community structure mapping and implements a community-centered graph eraser CGE, which can achieve node-level de-learning quickly and efficiently while maintaining model performance.
Complete Symmetry Breaking for Finite Models
Marek Dančo (Czech Technical University in Prague), João Jorge Araújo (Ben-Gurion University of the Negev)
Tabular
🎯 What it does: A SAT-based technique is proposed for computing compact and complete symmetry breaking in finite model finding, focusing on structures with a single binary operation (such as magic structures).
Complex-Cycle-Consistent Diffusion Model for Monaural Speech Enhancement
Yi Li (Lancaster University), Plamen P Angelov (University of Oxford)
RestorationGenerationRecurrent Neural NetworkDiffusion modelAudio
🎯 What it does: A monaural speech enhancement method based on diffusion models, SEDM, is proposed, which uses independent networks for speech amplitude and phase spectra, and replaces Gaussian noise with real-world noise during the diffusion process. Additionally, a noise-aware reverse process and a Complex Cycle Consistency (CCC) mechanism are designed to fully utilize the intrinsic relationship between amplitude and phase.