IJCAI 2025 Papers — Page 2
International Joint Conference on Artificial Intelligence · 1014 papers
Asset Pricing with Contrastive Adversarial Variational Bayes
Ruirui Liu (King's College London), Johannes Ruf (London School of Economics and Political Science)
Auto EncoderGenerative Adversarial NetworkContrastive LearningTabularFinance Related
🎯 What it does: Propose a Contrastive Adversarial Variational Bayesian (CAVB) network for nonlinear conditional asset pricing models, combining factor networks (AVB) with beta networks (multi-head contrastive learning) to learn latent factors and quantile-related factor loadings based on asset characteristics;
Asymptotic Analysis of Weighted Fair Division
Pasin Manurangsi (Google Research), Tomohiko Yokoyama (University of Tokyo)
Optimization
🎯 What it does: This paper studies the fair allocation of weighted divisible items from an asymptotic perspective, proving that a weighted envy-free (WEF) allocation exists when the weight ratio is bounded and the number of items satisfies m = Ω(n log n / log log n); it also provides the threshold for weighted proportionality (WPROP) as m = n/(1-µ), where µ is the mean of the value distribution; for the two-person case, it proves that WEF/WPROP also occurs with high probability when m = ω(√r) (r being the weight ratio of the two persons).
Asymptotic Fair Division: Chores Are Easier Than Goods
Pasin Manurangsi (Google Research), Warut Suksompong (National University of Singapore)
Optimization
🎯 What it does: Under random distributions, the existence and probability thresholds of envy-free and proportional fairness were studied when allocating indivisible chores to n agents, along with a polynomial-time constructive algorithm.
Asynchronous Credit Assignment for Multi-Agent Reinforcement Learning
Yongheng Liang (Sun Yat-sen University), Hao Cai (Shantou University)
Reinforcement LearningBenchmark
🎯 What it does: Propose an asynchronous credit assignment framework that combines virtual synchronous agents (VSP) with multiplicative value decomposition (MVD) to address the credit assignment challenge in multi-agent asynchronous decision-making.
Attention-based Conditional Random Field for Financial Fraud Detection
Xiaoguang Wang (Xi'an Jiaotong University), Tao Qin (Xi'an Jiaotong University)
Anomaly DetectionRecurrent Neural NetworkTabularTime SeriesFinance Related
🎯 What it does: The paper proposes an attention-based conditional random fields recurrent neural network (ACRF-RNN) for detecting financial statement fraud.
AttentionDrag: Exploiting Latent Correlation Knowledge in Pre-trained Diffusion Models for Image Editing
Biao Yang (Fudan University), Hualei Liu (Alibaba Group)
GenerationDiffusion modelImageBenchmark
🎯 What it does: Developed a single-step point-based image editing method called AttentionDrag based on the self-attention latent correlations of pre-trained diffusion models, which can automatically generate masks and maintain image semantic consistency through semantic interpolation;
Attractor-based Closed List Search: Sparsifying the Closed List for Efficient Memory-Constrained Planning
Alvin Zou (Carnegie Mellon University), Maxim Likhachev (Carnegie Mellon University)
OptimizationComputational Efficiency
🎯 What it does: Proposed the Attractor-based Closed List Search (ACLS) framework, which sparsifies the closed list by retaining a small number of key states (called attractors), achieving significant memory compression without significantly increasing runtime, and provided a lazy version called LazyACLS to further reduce computational overhead.
Attribute Association Driven Multi-Task Learning for Session-based Recommendation
Xinyao Wang (Tianjin University), Di Jin (Tianjin University)
Recommendation SystemRecurrent Neural NetworkGraph Neural NetworkGraphSequential
🎯 What it does: This paper proposes an attribute association-driven multi-task learning framework (A-D-MTL) to enhance session recommendation systems.
Automated Detection of Pre-training Text in Black-box LLMs
Ruihan Hu (Beijing University of Posts and Telecommunications), Xi Zhang (Beijing University of Posts and Telecommunications)
Anomaly DetectionTransformerLarge Language ModelText
🎯 What it does: Proposes the VeilProbe framework to automatically detect pre-trained text in black-box LLMs without any human intervention;
Automated Strategy Invention for Confluence of Term Rewrite Systems
Liao Zhang (University of Innsbruck), Cezary Kaliszyk (University of Melbourne)
Data SynthesisOptimizationHyperparameter SearchData-Centric LearningGraphBenchmark
🎯 What it does: Leverage AI to automatically generate strategies for the CSI termination analyzer, and perform random TRS data augmentation on the original ARI-COPS dataset to create a large-scale hybrid dataset.
Automated Superscalar Processor Design by Learning Data Dependencies
Shuyao Cheng (State Key Lab Of Processors Institute Of Computing Technology Chinese Academy Of Sciences), Yunji Chen (State Key Lab Of Processors Institute Of Computing Technology Chinese Academy Of Sciences)
OptimizationTextSequential
🎯 What it does: Automatically designed a RISC-V based superscalar processor QiMeng-CPU-v2 by learning data dependencies between instructions.
Avoiding Undesired Future with Sequential Decisions
Lue Tao (Nanjing University), Zhi-Hua Zhou (Nanjing University)
OptimizationTabularTime Series
🎯 What it does: This paper proposes a multi-stage rehearsal method that dynamically integrates real-time observational information into a structural rehearsal model through retrospective inference, thereby generating a series of decisions to avoid adverse outcomes.
Backdoor Attack on Vertical Federated Graph Neural Network Learning
Jirui Yang (Fudan University), Ruijun Deng (Fudan University)
Federated LearningAdversarial AttackGraph Neural NetworkGraph
🎯 What it does: This paper studies backdoor attacks in vertical federated graph neural networks (VFGNN), proposing a method that can implement attacks with a small number of target class nodes in scenarios where labels are not accessible.
Balance-Aware Sequence Sampling Makes Multi-Modal Learning Better
Zhi-Hao Guan (Nanjing University of Science and Technology), Yang Yang (Nanjing University of Science and Technology)
Representation LearningData-Centric LearningImageTextMultimodalityAudio
🎯 What it does: This paper proposes an improved multi-modal learning method based on training sequences—Balance-Aware Sequence Sampling (BSS)—which mitigates performance degradation caused by modal imbalance by evaluating sample balance and iteratively training from balanced to unbalanced orders.
Balancing Imbalance: Data-Scarce Urban Flow Prediction via Spatio-Temporal Balanced Transfer Learning
Xinyan Hao (Beijing Jiaotong University), Youfang Lin (Beijing Jiaotong University)
Domain AdaptationAuto EncoderContrastive LearningTime Series
🎯 What it does: Proposes the STBaT framework for cross-city traffic flow prediction based on regional imbalance perception, correcting source city distribution bias and enabling knowledge transfer through the Region Imbalance Acquisition Module and Spatio-Temporal Balanced Learning Module, achieving precise prediction in data-scarce cities.
Balancing Invariant and Specific Knowledge for Domain Generalization with Online Knowledge Distillation
Di Zhao (University Of Auckland), Yun Sing Koh (University Of Auckland)
ClassificationDomain AdaptationKnowledge DistillationVision Language ModelImageBenchmark
🎯 What it does: This paper proposes the BOLD framework, achieving online knowledge distillation that simultaneously distills domain-invariant and domain-specific knowledge.
Balancing User-Item Structure and Interaction with Large Language Models and Optimal Transport for Multimedia Recommendation
Haodong Li (China University of Petroleum (East China)), Xiaokang Zhou (Kansai University)
Recommendation SystemGraph Neural NetworkTransformerLarge Language ModelMultimodality
🎯 What it does: Propose the BLAST framework, which leverages large language models to generate user/item-side information, constructs user-user and item-item graphs, and combines Optimal Transport data augmentation with negative sample constraints to achieve structural and interactive balance in multi-modal recommendation.
Balans: Multi-Armed Bandits-based Adaptive Large Neighborhood Search for Mixed-Integer Programming Problems
Junyang Cai (University of Southern California), Bistra Dilkina (University of Southern California)
OptimizationReinforcement LearningTabular
🎯 What it does: Proposed an online meta-solver called BALANS, integrating Mixed Integer Programming (MIP), Adaptive Large Neighborhood Search (ALNS), and Multi-Armed Bandit (MAB), to dynamically select destroy-repair operators for solving MIP problems without relying on offline training.
BankTweak: Adversarial Attack Against Multi-Object Trackers by Manipulating Feature Banks
Woojin Shin (University of Seoul), Hyeongboo Baek (University of Seoul)
Object TrackingAdversarial AttackConvolutional Neural NetworkTransformerVideo
🎯 What it does: Propose a new adversarial attack called BankTweak for multi-object trackers, which induces persistent ID switches during the association phase by manipulating the feature bank without altering target locations, maintaining stealthiness.
Base-Detail Feature Learning Framework for Visible-Infrared Person Re-Identification
Zhihao Gong (Harbin Institute of Technology (Shenzhen)), Yong Xu (Harbin Institute of Technology (Shenzhen))
RecognitionRetrievalKnowledge DistillationConvolutional Neural NetworkTransformerImageMultimodality
🎯 What it does: This paper proposes a framework based on base and detail feature learning (BDLF) for visible and infrared (VI) person re-identification, aiming to simultaneously extract modality-shared base features and modality-specific detail features, thereby compensating for the neglect of detail information in traditional methods.
BEVTrack: A Simple and Strong Baseline for 3D Single Object Tracking in Bird's-Eye View
Yuxiang Yang (Hangzhou Dianzi University), Jing Zhang (Wuhan University)
Object TrackingAutonomous DrivingComputational EfficiencyConvolutional Neural NetworkPoint Cloud
🎯 What it does: This paper proposes BEVTrack, a 3D single-target tracking method based on bird's-eye view (BEV) that uses only a single regression loss, significantly simplifying the complex frameworks of traditional multi-task, multi-loss approaches;
Beyond Feature Mapping GAP: Integrating Real HDRTV Priors for Superior SDRTV-to-HDRTV Conversion
Gang He (Xidian University), Xianyun Wu (Xidian University)
Image TranslationSuper ResolutionAuto EncoderImageVideo
🎯 What it does: Propose a high-quality conversion method from SDR to HDR based on HDR video priors.
Beyond Fixed Length: Bucket Pre-training is All You Need
Qing Yang (Harbin Institute of Technology), Ting Liu (Harbin Institute of Technology)
Computational EfficiencyData-Centric LearningTransformerLarge Language ModelTextBenchmark
🎯 What it does: Propose a multi-bucket data organization method called BucketLLM to replace the traditional fixed-length data slicing approach in LLM pre-training, and evaluate data organization quality using quantitative metrics.
Beyond Individual and Point: Next POI Recommendation via Region-aware Dynamic Hypergraph with Dual-level Modeling
Xixi Li (China University of Mining and Technology), Wen-liang Du
Recommendation SystemGraph Neural NetworkTransformerGraphSequential
🎯 What it does: Propose a Region-aware Dynamic Hypergraph with Dual-level Modeling (ReHDM) framework to address the next POI recommendation problem, combining quadruple-key region encoding, dual-level reconstruction modeling at the POI and trajectory levels, and jointly capturing high-order cross-user associations and dynamic spatiotemporal patterns through hypergraph convolution and Transformer.
Beyond Low-rankness: Guaranteed Matrix Recovery via Modified Nuclear Norm
Jiangjun Peng (Northwestern Polytechnical University), Deyu Meng (Xi'an Jiaotong University)
OptimizationImageVideoBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: This paper proposes an improved nuclear norm (MNN) framework that captures both the global low-rank structure and local smoothness information of a matrix by computing the nuclear norm after transformation, and provides theoretical recoverability guarantees for its applications in robust PCA and matrix completion.
Beyond Statistical Analysis: Multimodal Framework for Time Series Forecasting with LLM-Driven Temporal Pattern
Jiahong Xiong (State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications), Jingyu Wang (State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications)
TransformerLarge Language ModelPrompt EngineeringMultimodalityTime SeriesBenchmark
🎯 What it does: Integrate LLM with traditional numerical models, leveraging LLM to generate future time pattern texts, which are then converted into numerical features through memory networks to assist in time series prediction.
Beyond Symmetry in Repeated Games with Restarts
Henry Fleischmann (Carnegie Mellon University), Ratip Emin Berker (Carnegie Mellon University)
🎯 What it does: Propose and analyze stable sequences in infinite repeated games with restart functions that support asymmetric games and asymmetric strategies, providing necessary and sufficient conditions for stability.
Beyond the Known: Decision Making with Counterfactual Reasoning Decision Transformer
Minh Hoang Nguyen (Deakin University), Hung Le (Deakin University)
OptimizationTransformerReinforcement LearningSequentialBenchmark
🎯 What it does: Propose Counterfactual Reasoning Decision Transformer (CRDT), which enhances Decision Transformer with counterfactual reasoning capabilities by generating and leveraging counterfactual experiences to improve offline RL decision-making;
Beyond the Map: Learning to Navigate Unseen Urban Dynamics Using Diffusion-Guided Deep Reinforcement Learning
Monu Nagar (Indian Institute of Technology Jodhpur), Debasis Das (Indian Institute of Technology Jodhpur)
Autonomous DrivingReinforcement LearningDiffusion modelImageBenchmark
🎯 What it does: Propose the Diffusion-Guided Deep Reinforcement Learning framework, combining diffusion models with Soft Actor-Critic to achieve navigation learning in unknown urban environments.
Beyond Winning Strategies: Admissible and Admissible Winning Strategies for Quantitative Reachability Games
Karan Muvvala (University of Colorado at Boulder), Morteza Lahijanian (University of Colorado at Boulder)
Reinforcement Learning
🎯 What it does: This paper investigates admissible strategies and admissible winning strategies in quantified reachability games, proposing necessary and sufficient conditions for these strategies, and providing synthesis algorithms with polynomial time (when the budget is fixed) or pseudopolynomial time. Subsequently, the feasibility and superiority of the approach are demonstrated in grid world and robot manipulation domains.
Bi-DiffCD: Bidirectional Diffusion Guided Collaborative Change Detection for Arbitrary-Modal Remote Sensing Images
Jingyu Zhao (Xidian University), Wenqian Dong (Xidian University)
Diffusion modelMultimodality
🎯 What it does: Propose a Bi-Directional Diffusion-Guided Collaborative Change Detection Model (Bi-DiffCD), which achieves modal alignment for remote sensing images of arbitrary modalities through conditional diffusion, and realizes more accurate change detection by leveraging multi-level difference features.
Bidirectional Search while Ensuring Meet-In-The-Middle via Effective and Efficient-to-Compute Termination Conditions
Yi Wang (University of New Hampshire), Oren Salzman (Technion-Israel Institute of Technology)
OptimizationBenchmark
🎯 What it does: Proposed a new bidirectional heuristic search algorithm called MEET, which combines more compact termination conditions and dynamically updated heuristics to achieve faster search while satisfying the 'midpoint meeting' property (MMP).
BILE: An Effective Behavior-based Latent Exploration Scheme for Deep Reinforcement Learning
Yiming Wang (University of Macau), Leong Hou U (University of Macau)
Robotic IntelligenceReinforcement LearningBenchmark
🎯 What it does: Proposed a behavior-indexed latent exploration scheme called BILE to improve state space exploration efficiency in deep reinforcement learning.
Bimodal Depth-First Search for Scalable GAC for AllDifferent
Sulian Le Bozec-Chiffoleau (IMT Atlantique), Xavier Lorca (IMT Mines Albi)
OptimizationComputational EfficiencyGraph
🎯 What it does: Proposed a Bimodal DFS/BFS that can efficiently execute on both sparse and dense graphs, embedded into Regin's GAC algorithm to achieve efficient strong arc consistency inference for ALLDIFFERENT constraints.
Binary Event-Driven Spiking Transformer
Honglin Cao (University of Electronic Science and Technology of China), Haizhou Li (Chinese University of Hong Kong)
ClassificationComputational EfficiencyKnowledge DistillationSpiking Neural NetworkTransformerImage
🎯 What it does: Proposed the Binary Event-Driven Spiking Transformer (BESTformer), achieving efficient integration of Transformer and Spiking Neural Network;
BinMetric: A Comprehensive Binary Code Analysis Benchmark for Large Language Models
Xiuwei Shang (University of Science and Technology of China), Nenghai Yu (University of Science and Technology of China)
AI Code AssistantTransformerLarge Language ModelTextBenchmark
🎯 What it does: The BinMetric benchmark provides a set of 1,000 questions covering six binary analysis tasks (call site reconstruction, decompilation, signature recovery, code summarization, algorithm classification, assembly generation), aiming to unify the evaluation of LLMs in binary analysis.
Block Circulant Adapter for Large Language Models
Xinyu Ding (Sun Yat-sen University), Zhongfeng Wang (Sun Yat-sen University)
Hyperparameter SearchTransformerLarge Language ModelText
🎯 What it does: Propose Block Circulant Adapter (BCA), utilizing block circulant matrices and 1D FFT for parameter-efficient fine-tuning;
BMIP: Bi-directional Modality Interaction Prompt Learning for VLM
Song-Lin Lv (Nanjing University), Lan-Zhe Guo (Nanjing University)
ClassificationDomain AdaptationTransformerSupervised Fine-TuningPrompt EngineeringVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Proposes a bidirectional modal interaction prompt learning method (BMIP) based on CLIP, achieving dynamic weight adjustment between visual and language modalities through deep language prompts, deep visual prompts, and attention-driven learnable aggregation functions, thereby enhancing the performance of VLM on downstream tasks.
Boost Embodied AI Models with Robust Compression Boundary
Chong Yu (Fudan University), Zhongxue Gan (Fudan University)
CompressionAutonomous DrivingRobotic IntelligenceContrastive LearningImagePoint Cloud
🎯 What it does: Propose the BRCB algorithm to achieve robust compression of embodied AI models, fine-grained splitting model weights into anti-disturbance stable and sensitive branches, and adding a gating layer during compression training to enhance the robustness and efficiency of the compressed model compared to the original model.
Boosting Few-Shot Open-Set Object Detection via Prompt Learning and Robust Decision Boundary
Zhaowei Wu (Beihang University), Zhong Zhou (Beihang University)
Object DetectionAnomaly DetectionTransformerPrompt EngineeringVision Language ModelImage
🎯 What it does: Propose a few-shot open-set object detection framework based on prompt learning, which utilizes text information to help identify known and unknown targets.
Boosting Visual Knowledge-Intensive Training for LVLMs Through Causality-Driven Visual Object Completion
Qingguo Hu (Xiamen University), Jinsong Su (Xiamen University)
TransformerLarge Language ModelVision Language ModelMultimodalityChain-of-Thought
🎯 What it does: Propose a self-improving framework based on Causal-Driven Visual Object Completion (CVC), enabling Large Vision-Language Models (LVLMs) to enhance visual perception and reasoning capabilities by generating and filtering reasonable reasoning paths.
Boosting Zero-shot Stereo Matching Using Large-Scale Mixed Images Sources in the Real World
Yuran Wang (Beijing Institute of Technology), Ying Fu (Beijing Institute of Technology)
Data SynthesisDepth EstimationAutonomous DrivingConvolutional Neural NetworkTransformerDiffusion modelContrastive LearningImage
🎯 What it does: Propose the BooSTer framework, which generates dense stereo matching data from single-view images by combining monocular depth estimation with diffusion models; supervise sparse real data using pseudo monocular depth labels and a dynamic scale/translation-invariant loss; integrate a large-scale vision foundation model (DINOv2) with traditional CNNs into a hybrid encoder to enhance feature expression and generalization capabilities.
Brain-Inspired Stepwise Patch Merging for Vision Transformers
Yonghao Yu (University of Chinese Academy of Sciences), Yi Zeng (University of Chinese Academy of Sciences)
ClassificationObject DetectionSegmentationTransformerImage
🎯 What it does: This paper proposes a hierarchical downsampling module based on visual Transformer called Stepwise Patch Merging (SPM), which improves feature representation by first performing multi-scale aggregation (MSA) and then guided local enhancement (GLE).
Breaking the Self-Evaluation Barrier: Reinforced Neuro-Symbolic Planning with Large Language Models
Jie-Jing Shao (Nanjing University), Lan-Zhe Guo (Nanjing University)
OptimizationReinforcement Learning from Human FeedbackLarge Language ModelReinforcement LearningTextBenchmark
🎯 What it does: Proposed a reinforced neuro-symbolic planning framework (RNSP) that integrates large language models with symbolic verifiers, enhancing constraint satisfaction and planning accuracy through symbolic verification feedback.
BridgeVoC: Neural Vocoder with Schrödinger Bridge
Tong Lei (Key Laboratory of Modern Acoustics, Nanjing University), Andong Li (Key Laboratory of Noise and Vibration Research, Institute of Acoustics Chinese Academy of Sciences)
GenerationData SynthesisScore-based ModelGenerative Adversarial NetworkStochastic Differential EquationAudio
🎯 What it does: Proposed a time-frequency domain neural vocoder called BridgeVoC based on the Schr"odinger bridge, treating the vocoder as a recovery process from a degraded linear spectrum to a complete spectrum;
Bridging Generative and Discriminative Learning: Few-Shot Relation Extraction via Two-Stage Knowledge-Guided Pre-training
Quanjiang Guo (University of Electronic Science and Technology of China), Weidong Xiao (National University of Defense Technology)
Representation LearningTransformerLarge Language ModelPrompt EngineeringContrastive LearningText
🎯 What it does: Propose a two-stage knowledge-guided framework (TKRE), which generates explanation-driven knowledge and structured synthetic data via large language models (LLMs), and integrates them with traditional relation extraction models to address data scarcity and poor generalization in few-shot relation extraction (FSRE).
Bridging Local and Global Knowledge via Transformer in Board Games
Yan-Ru Ju (Academia Sinica), Ti-Rong Wu (Academia Sinica)
OptimizationConvolutional Neural NetworkTransformerReinforcement LearningSequential
🎯 What it does: Designed the ResTNet network, which interleaves residual blocks with Transformer blocks to bridge local and global knowledge, and trained within the AlphaZero framework to improve win rates in board games such as Go and Hex.
BRIGHT-VO: Brightness-Guided Hybrid Transformer for Visual Odometry with Multi-modality Refinement Module
Dongzhihan Wang (Shanghai University), Liang Xu (Shanghai University)
Pose EstimationAutonomous DrivingOptimizationTransformerSimultaneous Localization and MappingImageMultimodality
🎯 What it does: Propose BrightVO, a Transformer-based visual odometry framework that integrates IMU data in the backend through pose graph optimization to achieve iterative refinement of poses, specifically designed to enhance the accuracy and robustness of visual odometry in low-light environments.
BTPG: A Platform and Benchmark for Behavior Tree Planning in Everyday Service Robots
Xinglin Chen (National University of Defense Technology), Ji Wang (National University of Defense Technology)
Robotic IntelligenceLarge Language ModelBenchmark
🎯 What it does: Designed and implemented Behavior Tree Planning Gym (BTPG), providing a unified predicate logic + STRIPS behavior tree planning platform, four real/simulated environments (RoboWaiter, VirtualHome, OmniGibson, RobotHow), as well as automatically generated task datasets and benchmark evaluation frameworks.
CABIN: Debiasing Vision-Language Models Using Backdoor Adjustments
Bo Pang (University of Auckland), Yun Sing Koh (University of Auckland)
Explainability and InterpretabilityData-Centric LearningVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Utilize the backdoor adjustment method from causal inference to neutralize the image embeddings in the sensitive attribute subspace of vision-language models (VLM), achieving unbiased prediction.
CADP: Towards Better Centralized Learning for Decentralized Execution in MARL
Yihe Zhou (Zhejiang University), Mingli Song (Zhejiang University)
TransformerReinforcement LearningBenchmark
🎯 What it does: Proposes the Centralized Advising and Decentralized Pruning (CADP) framework, which enhances training efficiency in CTDE through centralized advice exchange and achieves communication-free independent policies during execution via self-pruning.
Can Large Models Teach Student Models to Solve Mathematical Problems Like Human Beings? A Reasoning Distillation Method via Multi-LoRA Interaction
Xinhe Li (Southeast University), Peng Wang (Southeast University)
Knowledge DistillationTransformerLarge Language ModelTextChain-of-Thought
🎯 What it does: Propose LoRID, a math reasoning distillation framework based on multi-LoRA interaction, which enhances the reasoning capabilities of small language models by explicitly augmenting knowledge and improving the interaction between System 1/System 2.
Can Retelling Have Adequate Information for Reasoning? An Enhancement Method for Imperfect Video Understanding with Large Language Model
Mingxin Li (Northwestern Polytechnical University), Chao Gao (Northwestern Polytechnical University)
TransformerLarge Language ModelVideoChain-of-Thought
🎯 What it does: Research on self-enhancing reasoning methods based on entities and relationships to complete incomplete video descriptions and answer video question-answering tasks.
Can We Translate Code Better with LLMs and Call Graph Analysis?
Yang Luo (Peking University)
AI Code AssistantTransformerLarge Language ModelPrompt EngineeringGraph
🎯 What it does: Propose TransGraph, a code translation method based on call graphs and bridge debuggers that significantly improves compilation and runtime correctness rates for large-scale projects.
Can We Verify Step by Step for Incorrect Answer Detection?
Xin Xu (Hong Kong University of Science and Technology), Yang Wang (Hong Kong University of Science and Technology)
Anomaly DetectionTransformerTextBenchmarkChain-of-Thought
🎯 What it does: This paper constructs the R2PE benchmark to systematically assess the relationship between the reasoning chain and the correctness of the final answer in chain-of-thought (CoT) generated by large language models, and proposes the Process Discernibility Score (PDS) framework to determine the authenticity of answers.
CAN-ST: Clustering Adaptive Normalization for Spatio-temporal OOD Learning
Min Yang (Shandong University), Yongshun Gong (Shandong University)
Domain AdaptationGraph Neural NetworkTime Series
🎯 What it does: Propose a general cluster-adaptive normalization framework named CAN-ST for addressing prediction problems under spatiotemporal distribution drift.
Cap-and-Penalize: Competitive Mechanisms for Multi-Phase Regularized Online Allocation
Seyedehkimia Alaviyar (University of Alberta), Xiaoqi Tan (University of Alberta)
OptimizationOrdinary Differential Equation
🎯 What it does: Proposed the Cap-and-Penalize (CNP) multi-stage non-separable regularized online allocation model, and provided the optimal posted-price mechanism;
Capturing Individuality and Commonality Between Anchor Graphs for Multi-View Clustering
Zhoumin Lu (Northwestern Polytechnical University), Rong Wang (Northwestern Polytechnical University)
OptimizationComputational EfficiencyRepresentation LearningMultimodality
🎯 What it does: Proposed a method called CICAG that simultaneously captures the individuality and commonality of multi-view anchor graphs, aiming to enhance multi-view clustering performance.
CASA: CNN Autoencoder-based Score Attention for Efficient Multivariate Long-term Time-series Forecasting
Minhyuk Lee (Seoul National University), MyungJoo Kang (Seoul National University)
Computational EfficiencyConvolutional Neural NetworkTransformerScore-based ModelAuto EncoderTime SeriesBenchmark
🎯 What it does: This paper proposes a CNN autoencoder attention module named CASA, which replaces the self-attention mechanism in Transformers to achieve more efficient and accurate multivariate long-term time series prediction.
Categorical Attention: Fine-grained Language-guided Noise Filtering Network for Occluded Person Re-Identification
Minghui Chen (Chinese Academy of Sciences), Zheng Lin (Chinese Academy of Sciences)
RecognitionConvolutional Neural NetworkTransformerVision Language ModelContrastive LearningMultimodality
🎯 What it does: This paper proposes FLaN-Net, a fine-grained language-guided noise filtering network for handling occluded person re-identification problems.
Category-aware EEG Image Generation Based on Wavelet Transform and Contrast Semantic Loss
Enshang Zhang (Kyoto University), Takashi Hanakawa (Kyoto University)
GenerationConvolutional Neural NetworkTransformerDiffusion modelContrastive LearningBiomedical Data
🎯 What it does: Employ a Transformer encoder combined with Discrete Wavelet Transform (DWT) and a gated attention mechanism to map EEG signals to visual stimulus images.
Causal Learning Meet Covariates: Empowering Lightweight and Effective Nationwide Air Quality Forecasting
Jiaming Ma (University of Science and Technology of China), Yang Wang (University of Science and Technology of China)
Computational EfficiencyData-Centric LearningTransformerTime Series
🎯 What it does: Proposed a nationwide air quality prediction framework called CauAir and released a large-scale dataset named LargeAQ covering 1341 monitoring stations with 8 years of hourly data.
Causal View of Time Series Imputation: Some Identification Results on Missing Mechanism
Ruichu Cai (Guangdong University of Technology), Zhifeng Hao (Shantou University)
RestorationFlow-based ModelTime Series
🎯 What it does: Studied the problem of missing value imputation in time series, proposing a DMM framework based on causal mechanisms that can model both MAR and MNAR missing mechanisms.
Causal-aware Large Language Models: Enhancing Decision-Making Through Learning, Adapting and Acting
Wei Chen, Ruichu Cai (Guangdong University of Technology)
Explainability and InterpretabilityTransformerLarge Language ModelReinforcement LearningPrompt EngineeringText
🎯 What it does: Proposes a Causal-aware LLMs framework that integrates structural causal models (SCM) with large language models (LLM), achieving adaptive causal knowledge acquisition and decision-making through a learning-adaptation-action loop.
Causality-Inspired Disentanglement for Fair Graph Neural Networks
Guixian Zhang (China University of Mining and Technology), Yanmei Zhang (China University of Mining and Technology)
ClassificationExplainability and InterpretabilityGraph Neural NetworkGenerative Adversarial NetworkGraphTabularBenchmark
🎯 What it does: Proposes a causality-inspired separable framework CDFG for disentangling causal factors and sensitive factors from graph neural networks to achieve fair node classification.
Cause-Effect Driven Optimization for Robust Medical Visual Question Answering with Language Biases
Huanjia Zhu (Beijing Institute of Technology), Bingzhi Chen (Beijing Institute of Technology)
OptimizationVision Language ModelContrastive LearningMultimodalityBiomedical Data
🎯 What it does: Proposes a causality-based optimization framework CEDO to eliminate language bias in medical visual question answering
CD^2: Constrained Dataset Distillation for Few-Shot Class-Incremental Learning
Kexin Bao (Institute Of Information Engineering Chinese Academy Of Sciences), Shiming Ge (Institute Of Information Engineering Chinese Academy Of Sciences)
ClassificationData SynthesisKnowledge DistillationRepresentation LearningData-Centric LearningConvolutional Neural NetworkImage
🎯 What it does: Proposes the Constrained Dataset Distillation (CD2) framework, combining dataset distillation with distillation constraints to build a memory that retains only critical knowledge while suppressing catastrophic forgetting, thereby enhancing FSCIL performance.
CFDONEval: A Comprehensive Evaluation of Operator-Learning Neural Network Models for Computational Fluid Dynamics
Menghan Liu (Sun Yat-sen University), Qingsong Zou (Sun Yat-sen University)
Graph Neural NetworkTransformerAuto EncoderMeshBenchmarkPhysics Related
🎯 What it does: Proposed the CFDONEval evaluation framework, systematically assessing 12 operator learning neural network models across 22 datasets, covering 7 classical CFD benchmark problems, and exploring five major challenges: multi-scale, convection-dominated, long-term prediction, multiphase flow, and complex geometry with non-uniform grids.
CFII-Net: Explicit Class Embeddings and Feature Maps Through Iterative Interaction for Boosting Medical Image Segmentation
Xinyu Zhu (East China Normal University), Yin Wen (East China Normal University)
SegmentationConvolutional Neural NetworkTransformerImageBiomedical Data
🎯 What it does: Propose CFII-Net, which improves medical image segmentation performance through an explicit category embedding generator and iterative interaction module.
CGI: Identifying Conditional Generative Models with Example Images
Zhi Zhou (Nanjing University), Lan-Zhe Guo (Nanjing University)
GenerationTransformerPrompt EngineeringVision Language ModelImageTextMultimodalityBenchmark
🎯 What it does: Propose Conditional Generative Model Identification (CGI) and design Prompt-based Model Identification (PMI) method, which quickly locates the most suitable conditional generative model through user example images.
ChronoFact: Timeline-based Temporal Fact Verification
Anab Maulana Barik (National University of Singapore), Mong Li Lee (National University of Singapore)
ClassificationRepresentation LearningRecurrent Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningTime SeriesSequential
🎯 What it does: Proposes the ChronoFact framework, which automatically verifies the authenticity of temporal statements containing multiple events using timeline reasoning.
Circuit-Aware d-DNNF Compilation
Vincent Derkinderen (KU Leuven), Jean-Marie Lagniez (CRIL)
OptimizationComputational Efficiency
🎯 What it does: They improved the d4 compiler, preserving the original Boolean circuit and implementing dynamic elimination of don't-care variables (irrelevant gates).
CLLMRec: Contrastive Learning with LLMs-based View Augmentation for Sequential Recommendation
Fan Lu (Nanjing University of Information Science and Technology), Wanchun Dou (Nanjing University)
Recommendation SystemTransformerLarge Language ModelContrastive LearningSequential
🎯 What it does: Proposes a framework named CLLMRec, which applies large language models (LLM) for view enhancement and combines contrastive learning to improve the accuracy of sequence recommendation.
CMFS: CLIP-Guided Modality Interaction for Mitigating Noise in Multi-Modal Image Fusion and Segmentation
Guilin Su (Harbin Institute of Technology), Zhenyu He (Harbin Institute of Technology)
SegmentationConvolutional Neural NetworkMixture of ExpertsVision Language ModelMultimodality
🎯 What it does: Propose a unified multi-task framework that accomplishes both infrared-visible image fusion and multi-modal semantic segmentation, leveraging CLIP-guided modal interaction to significantly suppress noise.
Co-Learning of Strategy and Structure Achieves Full Cooperation in Complex Networks with Dynamical Linking
Xiaoqing Fan (University of Warwick), Paolo Turrini (University of Warwick)
Reinforcement LearningGraph
🎯 What it does: Enabling reinforcement learning agents to achieve full cooperation through co-learning strategies and dynamic links on complex networks.
Coalition Obstruction Temporal Logic: A New Obstruction Logic to Reason About Demon Coalitions
Davide Catta (Université Sorbonne Paris Nord), James Ortiz (Télécom Paris)
Graph
🎯 What it does: Proposed Coalition Obstruction Temporal Logic (COTL), a temporal logic that supports defenders (called demons) in dynamic game models to block attack paths by temporarily disabling edges, and provided its semantics, model checking algorithm, and case studies.
Code-BT: A Code-Driven Approach to Behavior Tree Generation for Robot Tasks Planning with Large Language Models
Siyang Zhang (National University of Defense Technology), Jinjing Sun (Intelligent Game and Decision Lab)
Robotic IntelligenceTransformerLarge Language ModelTextBenchmark
🎯 What it does: This paper proposes the Code-BT framework, which leverages the code generation capabilities of large language models to automatically construct behavior trees for robot task planning.
CoderAgent: Simulating Student Behavior for Personalized Programming Learning with Large Language Models
Yi Zhan (University of Science and Technology of China), Zhenya Huang (University of Science and Technology of China)
AI Code AssistantLarge Language ModelAgentic AITextSequentialChain-of-Thought
🎯 What it does: This paper proposes a CoderAgent based on a large language model to fine-grainedly simulate the student programming process and generate interpretable learning trajectories.
COGRASP: Co-Occurrence Graph Based Stock Price Forecasting
Zhengze Li (University of Gttingen), Xiaoming Fu (University of Gttingen)
Recurrent Neural NetworkGraph Neural NetworkTextGraphTime SeriesFinance Related
🎯 What it does: Construct a social media co-occurrence relationship graph and combine it with a multi-scale attention LSTM model to predict stock prices.
CoLA-Former: Graph Transformer Using Communal Linear Attention for Lightweight Sequential Recommendation
Zhongying Zhao (Shandong University of Science and Technology), Qingtian Zeng (Shandong University of Science and Technology)
Recommendation SystemGraph Neural NetworkTransformerGraphSequential
🎯 What it does: Propose CoLA-Former, a lightweight graph Transformer model that combines shared linear attention and low-rank approximation to improve the accuracy and efficiency of sequence recommendation.
Collaborative Multi-LoRA Experts with Achievement-based Multi-Tasks Loss for Unified Multimodal Information Extraction
Li Yuan (South China University of Technology), Tao Wang (King's College London)
TransformerMixture of ExpertsVision Language ModelMultimodalityBenchmark
🎯 What it does: Propose a collaborative multi-LoRA expert model (C-LoRAE), utilizing a two-layer structure combining general experts and task-specific experts, along with mutual information maximization, a gating router, and achievement-based multi-task loss, to enable knowledge sharing and mitigate gradient conflicts across multi-modal information extraction tasks (MNER, MRE, MEE).
Combining Deep Reinforcement Learning and Search with Generative Models for Game-Theoretic Opponent Modeling
Zun Li (Google DeepMind), Michael P. Wellman (University of Michigan)
Reinforcement LearningWorld ModelTextTabular
🎯 What it does: Proposed a general multi-agent training framework that combines AlphaZero-style MCTS with deep generative models to form the GenBR (Generative Best Response) algorithm, which builds opponent models and achieves best responses in large-scale imperfect information games; subsequently, GenBR is embedded into the PSRO loop to iteratively generate opponent strategy families, and further optimized by a meta-solver based on Nash bargaining; the method's performance was validated in Colored Trails and Deal-or-No-Deal negotiation games, with comparisons to human-human gameplay.
Combining MORL with Restraining Bolts to Learn Normative Behaviour
Emery A. Neufeld (TU Wien), Radu Florin Tulcan (TU Wien)
Reinforcement LearningBenchmark
🎯 What it does: Convert the learning problem of NRBs into multi-objective reinforcement learning, propose Ordered Normative Restraining Bolts (ONRB), and automatically determine penalty weights through preprocessing to achieve hierarchical norm compliance.
Coming Out of the Dark: Human Pose Estimation in Low-light Conditions
Yong Su (Tianjin Normal University), Jieyang Li
RestorationPose EstimationTransformerImageBenchmark
🎯 What it does: This paper first constructs a human pose estimation dataset under low-light conditions, named LLIP, and proposes a reference-free low-light pose estimation framework, MHFCL. The framework consists of a Retinex-inspired high-frequency recovery flow and a Vision Transformer-based pose estimation flow. It integrates the recovered high-frequency information into pose prediction through multi-granularity feature consistency learning.
Community-Aware Graph Transformer for Brain Disorder Identification
Shengbing Pei (Anhui University), Jihong Guan (Tongji University)
ClassificationGraph Neural NetworkTransformerGraphBiomedical Data
🎯 What it does: Propose Community-Aware Graph Transformer (CAGT), which identifies brain disorders by integrating the community structure and global topological information of brain functional networks.
CompLex: Music Theory Lexicon Constructed by Autonomous Agents for Automatic Music Generation
Zhejing Hu (Hong Kong Polytechnic University), Bruce X. B. Yu
GenerationTransformerLarge Language ModelAgentic AIPrompt EngineeringAudio
🎯 What it does: Constructed an automated music theory dictionary called CompLex and proposed a multi-agent collaborative method named LexConstructor to improve the quality of text-to-music generation.
Computational Complexity of Planning for Recursive Primitive Task Networks: Selective Action Nullification with State Preservation
Yifan Zhang (Australian National University), Pascal Bercher (Australian National University)
Review/Survey Paper
🎯 What it does: Analyze the computational complexity of recursive primitive task networks (r-primitive) HTN planning, revealing the complete complexity spectrum from PSPACE to Ackermann levels under different recursive and goal constraints;
Concentrate on Weakness: Mining Hard Prototypes for Few-Shot Medical Image Segmentation
Jianchao Jiang (Nanjing University of Science and Technology), Haofeng Zhang (Nanjing University of Science and Technology)
SegmentationContrastive LearningBiomedical DataMagnetic Resonance ImagingComputed Tomography
🎯 What it does: Proposed a Few-Shot medical image segmentation method called CoW based on 'mining weak features', which locates edge weak features through a support self-predict module, subsequently generates hard prototypes, and achieves more accurate segmentation results by leveraging multi-similarity graph fusion;
Concurrent Planning and Execution Using Dispatch-Dependent Values
Andrew Coles (King's College London), Wheeler Ruml (University of New Hampshire)
OptimizationBenchmark
🎯 What it does: Propose a concurrent planning and execution algorithm based on dual open lists, which simplifies the CoPE problem using a dispatch dependency value function, and implements it within the OPTIC framework.
Conditional Causal Representation Learning for Heterogeneous Single-cell RNA Data Integration and Prediction
Jiayi Dong (Fudan University), Fei Wang (Fudan University)
Representation LearningData-Centric LearningAuto EncoderBiomedical Data
🎯 What it does: Propose the scConCRL framework, which integrates and predicts heterogeneous single-cell RNA-seq data using conditional causal representation learning.
Conditional Denoising Meets Polynomial Modeling: A Flexible Decoupled Framework for Time Series Forecasting
Jintao Zhang (University of Science and Technology of China), Daoyu Wang (University of Science and Technology of China)
Convolutional Neural NetworkDiffusion modelTime Series
🎯 What it does: This paper proposes a Conditional Denoising Polynomial Modeling (CDPM) framework that can be end-to-end trained, decomposing time series into trend and seasonal components, where the seasonal fluctuations are captured by conditional denoising diffusion models and the trend is modeled using a polynomial trend module.
Conditional Independent Test in the Presence of Measurement Error with Causal Structure Learning
Hongbin Zhang, Zhengming Chen (Shantou University)
Graph
🎯 What it does: This paper proposes a method for learning the causal structure of hidden variables under a linear non-Gaussian measurement error model (LiNGMME) with measurement errors, by utilizing the rank constraint of high-order cumulant matrices for conditional independence testing, and embedding this test into the PC algorithm to obtain the PC-ME algorithm.
Conditional Information Bottleneck-Based Multivariate Time Series Forecasting
Xinhui Li (Yunnan University), Yuehua Li (Yunnan University)
OptimizationRepresentation LearningTransformerContrastive LearningTime Series
🎯 What it does: This paper proposes a multivariate time series prediction framework based on Conditional Information Bottleneck (CIB) and Conditional Mutual Information (CMI), which can simultaneously capture cross-correlations between variables and temporal coherence within each variable.
Connecting Giants: Synergistic Knowledge Transfer of Large Multimodal Models for Few-Shot Learning
Hao Tang (Centre for Smart Health, Hong Kong Polytechnic University), Jing Qin (Centre for Smart Health, Hong Kong Polytechnic University)
Knowledge DistillationMeta LearningPrompt EngineeringVision Language ModelAuto EncoderMultimodalityChain-of-Thought
🎯 What it does: Propose a SYNTRANS framework that leverages large-scale multimodal models (CLIP, LLM, VLM) to transfer and fuse knowledge for few-shot learning (FSL), significantly enhancing FSL performance.
Consensus-Guided Incomplete Multi-view Clustering via Cross-view Affinities Learning
Qian Liu, Yang Wang (Hebei University)
Representation LearningMultimodality
🎯 What it does: Proposed the CAL method, based on high-order interactions and consensus-guided cross-view affinity learning, to achieve incomplete multi-view clustering.
Consistency-Aware Padding for Incomplete Multi-Modal Alignment Clustering Based on Self-Repellent Greedy Anchor Search
Shubin Ma (Dalian University of Technology), Bo Xu (Dalian University of Technology)
Representation LearningContrastive LearningMultimodalityBenchmark
🎯 What it does: Propose a framework named CAPIMAC for co-clustering incomplete and misaligned multimodal data.
Constrained Preferential Bayesian Optimization and Its Application in Banner Ad Design
Koki Iwai (Hakuhodo DY Holdings Inc), Masataka Goto (National Institute of Advanced Industrial Science and Technology)
OptimizationImage
🎯 What it does: This paper proposes Constraint-Prioritized Bayesian Optimization (CPBO) and applies it to banner ad design, integrating CTR constraints to achieve interactive optimization between designers and algorithms;
Constrained Sequential Inference in Machine Learning Using Constraint Programming
Virasone Manibod (GIRO), Gilles Pesant (Polytechnique Montreal)
GenerationComputational EfficiencyTransformerLarge Language ModelSequential
🎯 What it does: Designed and implemented the GeAI-BLAnC framework, which combines the probability distribution output by pre-trained sequence generation models (such as GPT-2, CMT) with the marginal probabilities calculated by constraint programming (CP), applying long-range structural constraints to generated sequences during inference.
Constrained Serial Dictatorships Can Be Fair
Sylvain Bouveret (University of Grenoble Alpes), Guillaume Méroué (University of Côte d'Azur)
Optimization
🎯 What it does: This paper studies how to balance fairness and efficiency under the constraint of strategyproofness through Constrained Serial Dictatorship (CSD), and provides algorithms for finding the optimal CSD in various scenarios.
Constructive Conflict-Driven Multi-Agent Reinforcement Learning for Strategic Diversity
Yuxiang Mai (University of Chinese Academy of Sciences), Kaiqi Huang (University of Chinese Academy of Sciences)
Reinforcement Learning
🎯 What it does: Propose a multi-agent reinforcement learning method called CoDiCon based on competitive intrinsic rewards, which utilizes a ranking mechanism to stimulate competition and collaboration, thereby enhancing team diversity and performance.