AAAI 2026 Papers — Page 7
AAAI Conference on Artificial Intelligence · 4149 papers
CoLM: Collaborative Large Models via a Client-Server Paradigm
Siqi Huang (Institute of Artificial Intelligence, China Telecom), Hongyuan Zhang (Institute of Artificial Intelligence, China Telecom)
Federated LearningComputational EfficiencyLarge Language ModelMixture of ExpertsVision Language ModelTextMultimodality
🎯 What it does: Propose the CoLM framework to achieve collaborative inference based on a client-server architecture for large models, supporting multi-party collaboration between language models and audio-visual language models.
Colonel Blotto with Battlefield Games
Salam Afiouni (Columbia University), Christian Kroer (Columbia University)
OptimizationReinforcement Learning
🎯 What it does: Proposed and analyzed a two-level Colonel Blotto game model, where players first allocate forces to battlefields and then engage in zero-sum subgames parameterized by force allocation on each battlefield; investigated discrete/continuous forces, sum/minimization aggregation, and bilateral/unilateral strategy settings, providing existence and computational analysis along with corresponding algorithms.
CoMA-SLAM: Collaborative Multi-Agent Gaussian SLAM with Geometric Consistency
Lin Chen (Northwestern Polytechnical University), Guangming Wang (Northwestern Polytechnical University)
Pose EstimationDepth EstimationOptimizationComputational EfficiencyGaussian SplattingSimultaneous Localization and MappingImagePoint CloudBenchmark
🎯 What it does: Proposes CoMA-SLAM, a distributed multi-agent Gaussian SLAM framework, achieving high-precision tracking, rendering, and globally consistent scene reconstruction.
CoMA: Compositional Human Motion Generation with Multi-modal Agents
Shanlin Sun (University of California, Irvine), Xiaohui Xie (Columbia University)
GenerationTransformerLarge Language ModelAgentic AIPrompt EngineeringVision Language ModelAuto EncoderVideoTextSequential
🎯 What it does: CoMA proposes a framework based on multimodal agents to generate high-quality, editable, and self-correcting 3D human motion from text.
Combining LLM Semantic Reasoning with GNN Structural Modeling for Multi-View Multi-Label Feature Selection
Zhiqi Chen (Jilin University), Wanfu Gao (Jilin University)
Graph Neural NetworkLarge Language ModelGraph
🎯 What it does: Propose a multi-perspective multi-label feature selection method that combines the semantic reasoning of large language models (LLMs) with the structural modeling of graph neural networks (GNNs)
CometNet: Contextual Motif-guided Long-term Time Series Forecasting
Weixu Wang (Tianjin University), Tie Qiu (Tianjin University)
TransformerMixture of ExpertsTime SeriesFinance Related
🎯 What it does: Proposes CometNet, a long-term time series prediction framework guided by global context motifs.
ComLQ: Benchmarking Complex Logical Queries in Information Retrieval
Ganlin Xu (Fudan University), Deqing Yang (United Automotive Electronic Systems)
Data SynthesisRetrievalTransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: This paper proposes a complex logic query benchmark dataset called ComLQ for information retrieval, and automatically generates queries conforming to specific logical structures using LLM and subgraph-guided prompts;
CommitMoE: Efficient Fallback-Free MoE Inference with Offloading Under GPU Memory Constraints
Han Li (University of Science and Technology of China), Guangzhong Sun (University of Science and Technology of China)
Computational EfficiencyMixture of ExpertsText
🎯 What it does: Developed a Mixture of Experts inference framework named CommitMoE, which eliminates fallback mechanisms and improves inference speed in GPU memory-constrained environments.
Commonality in Few: Few-Shot Multimodal Anomaly Detection via Hypergraph-Enhanced Memory
Yuxuan Lin (Fudan University), Wenqiang Zhang (Fudan University)
Anomaly DetectionGraph Neural NetworkTransformerMultimodalityPoint Cloud
🎯 What it does: Proposed a hypergraph-based few-shot multi-modal industrial anomaly detection framework named CIF.
Communication-Efficient Heterogeneous Federated Learning with Sparse Prototypes in Resource-Constrained Environments
Gyuejeong Lee (SAKAK Inc), Daeyoung Choi (Korea Cyber University)
CompressionFederated LearningImageText
🎯 What it does: Proposed the TinyProto framework for achieving communication-efficient prototype exchange in heterogeneous federated learning;
Communication-efficient Multi-Agent Reinforcement Learning with Spatiotemporal Information Hub
Ling Ding (Wuhan University), Wei Yu (Wuhan University)
TransformerReinforcement LearningBenchmark
🎯 What it does: This paper proposes a Multi-Agent Communication Hub (MATCH), achieving efficient and robust communication through a central information hub within the CTDE framework.
ComoRAG: A Cognitive-Inspired Memory-Organized RAG for Stateful Long Narrative Reasoning
Juyuan Wang (South China University of Technology), Liyan Xu (Tencent)
RetrievalRepresentation LearningLarge Language ModelAgentic AITextRetrieval-Augmented Generation
🎯 What it does: Introduces ComoRAG, a cognitive-inspired memory-organized retrieval-augmented generation framework for long-form narrative reasoning, capable of dynamically constructing and updating the global context across multiple rounds of retrieval;
Compensate to Not Deviate: On Subsidised Equilibria
Vittorio Bilò (University of Salento), Luca Moscardelli (University of Chieti-Pescara)
Optimization
🎯 What it does: Proposed and studied the new deterministic stable solution concept called Subsidized Equilibrium, which uses subsidies to eliminate players' regret and achieve stable strategy configurations;
Compensating Distribution Drifts in Continual Learning with Pre-trained Vision Transformers
Xuan Rao (Beijing Normal University), Cesare Alippi (Politecnico di Milano)
Domain AdaptationKnowledge DistillationRepresentation LearningTransformerSupervised Fine-TuningImageBenchmark
🎯 What it does: Achieve sample-free incremental learning on pre-trained vision Transformers, proposing a compensation method to address distribution drift caused by gradual fine-tuning.
CompEvent: Complex-valued Event-RGB Fusion for Low-light Video Enhancement and Deblurring
Mingchen Zhong (University of Science and Technology of China), Baocai Yin (University of Science and Technology of China)
RestorationConvolutional Neural NetworkRecurrent Neural NetworkVideoMultimodality
🎯 What it does: Designed and implemented an end-to-end complex-valued event-RGB fusion framework named CompEvent for low-light video deblurring and enhancement.
Complex Instruction Following with Diverse Style Policies in Football Games
Chenglu Sun (Tencent), Chen Chen (Tencent)
TransformerLarge Language ModelReinforcement LearningText
🎯 What it does: In the complex multi-agent football game (Google Research Football 5v5), the Language-Controlled Diverse Style Policies (LCDSP) framework is proposed, combining Diverse Style Training (DST) and Style Interpreter (SI) to enable a single reinforcement learning policy to be controlled by high-level abstract language instructions and rapidly generate diverse behavioral styles.
Complex Mathematical Expression Recognition: Benchmark, Large-Scale Dataset and Strong Baseline
Weikang Bai (Fudan University), Zhineng Chen (Fudan University)
RecognitionConvolutional Neural NetworkTransformerLarge Language ModelVision Language ModelImageTextBenchmark
🎯 What it does: Studies complex mathematical expression recognition, proposes the CMER-Bench benchmark, the CMER-3M/MER-17M datasets, and introduces the CMERNet model.
Composition-Incremental Learning for Compositional Generalization
Zhen Li (Beijing Institute Of Technology), Yunde Jia (Shenzhen Msu-Bit University)
Knowledge DistillationRepresentation LearningTransformerVision Language ModelAuto EncoderImageTextMultimodalityBenchmark
🎯 What it does: Proposes a Composition-Incremental Learning (CompIL) setting, enabling models to progressively learn new combinations over time while keeping the primitive set fixed, thereby continuously enhancing compositional generalization ability; simultaneously designs a pseudo-replay framework that generates visual representations of past tasks using a visual synthesizer and maintains consistent primitive representations through language primitive distillation.
Compositional Attribute Imbalance in Vision Datasets
Yanbiao Ma (Renmin University of China), Andi Zhang (University of Manchester)
Data-Centric LearningConvolutional Neural NetworkTransformerImageBenchmark
🎯 What it does: This paper studies the problem of image attribute imbalance and proposes using CLIP to construct an attribute dictionary and quantify the combination attribute scarcity (CAS).
Comprehensive Urban Region Representation Learning via Multi-View Joint Learning and Contrastive Learning
Yingde Lin (Jilin University), Pengyang Wang (Dalian Maritime University)
Representation LearningGraph Neural NetworkContrastive LearningGraphTabular
🎯 What it does: Studied a multi-view joint learning and structure-aware contrastive learning framework (MVJC) for urban area embeddings, enhancing crime prediction and land use clustering performance.
Compress-then-Rank: Faster and Better Listwise Reranking with Large Language Models via Ranking-Aware Passage Compression
Zhewei Zhi (Dalian University of Technology), Yongliang Ding (Meituan)
RetrievalCompressionTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes the Compress-then-Rank (C2R) framework, which compresses retrieved long documents into a series of short vectors and then uses large language models (LLMs) for list-wise re-ranking, significantly reducing input length and inference latency.
Compression Artifacts Removal for VVC with Frequency Domain Mixture of Experts Network
Qijun Wang (Anhui University), Jun Wang (Anhui University)
RestorationConvolutional Neural NetworkTransformerMixture of ExpertsImage
🎯 What it does: Propose a hybrid expert network named ARMoE that combines multi-frequency domain transformations with a sparse activation strategy to remove various artifacts from H.266/VVC compressed images.
CompTrack: Information Bottleneck-Guided Low-Rank Dynamic Token Compression for Point Cloud Tracking
Sifan Zhou (Southeast University), Shuo Wang (Mininglamp Technology)
Object TrackingAutonomous DrivingTransformerPoint Cloud
🎯 What it does: Designed the CompTrack framework, using spatial foreground prediction and information bottleneck guided low-rank dynamic token compression to address spatial redundancy and information redundancy in LiDAR point cloud tracking.
Computer Vision Modeling of the Development of Geometric and Numerical Concepts in Humans
Zekun Wang (Georgia Institute of Technology), Sashank Varma (Georgia Institute of Technology)
ClassificationConvolutional Neural NetworkImage
🎯 What it does: Conduct epoch-by-epoch inspections on the ResNet-50 model to evaluate its sensitivity to geometric/topological concepts (odd-one-out tasks) and numerical concepts (distance, size, proportion effects) during image classification training, and compare the changes in sensitivity with human developmental curves.
Computing Approximately Proportional Allocations of Indivisible Goods: Beyond Additive and Monotone Valuations
Martin Jupakkal Andersen (Aarhus University), Alexander Søltoft (Aarhus University)
Optimization
🎯 What it does: Investigated the relationship between PROP1 (relative proportional fairness) and EF1 (relative envy-freeness), as well as Pareto optimality, in non-additive and potentially non-monotonic fair division models, and provided corresponding existence, algorithmic, and compatibility conclusions.
Computing Equilibrium Nominations in Presidential Elections
Piotr Faliszewski (AGH University of Kraków), Paolo Turrini (University of Warwick)
🎯 What it does: This paper studies the issue of strategic candidate nominations by parties in elections under plurality voting, and analyzes the existence and computational complexity of pure Nash equilibria within the Party-Aligned Single-Peaked preference domain.
Computing Probabilistic Explanations for ML Models: Fixed-Parameter Algorithms
Sebastian Ordyniak (University of Leeds), Stefan Szeider (TU Wien)
Explainability and InterpretabilityComputational Efficiency
🎯 What it does: Studied the probabilistic explanation problem for interpretable models such as decision trees, decision lists, and decision sets, and provided a unified fixed-parameter tractable (FPT) algorithm framework
Computing Syntax Tree-based Minimal Unsatisfiable Cores of LTLf Formulas
Valeria Fionda (University of Calabria), Francesco Ricca (University of Calabria)
TextBenchmark
🎯 What it does: This paper proposes a method based on Answer Set Programming (ASP) for computing minimal unsatisfiable cores (MUC) and minimal correction sets (MCS) at the syntax tree level in finite-trace linear temporal logic (LTLf), supporting arbitrary formulas rather than only conjunctions.
Concept-RuleNet: Grounded Multi-Agent Neurosymbolic Reasoning in Vision Language Models
Sanchit Sinha (University of Virginia), Aidong Zhang (University of Virginia)
ClassificationExplainability and InterpretabilityTransformerLarge Language ModelVision Language ModelImageBiomedical Data
🎯 What it does: Proposed a multi-agent neuro-symbolic framework called Concept-RuleNet, which achieves interpretable reasoning for vision-language models and enhances classification performance through visual concept extraction, symbolic generation, and rule reasoning.
Concepts from Representations: Post-hoc Concept Bottleneck Models via Sparse Decomposition of Visual Representations
Shizhan Gong (Chinese University of Hong Kong), Qi Dou (Chinese University of Hong Kong)
ClassificationExplainability and InterpretabilityRepresentation LearningLarge Language ModelAuto EncoderImageMultimodalityBenchmark
🎯 What it does: Propose the posterior concept bottleneck model PCBM-ReD, which automatically decomposes representations from pre-trained visual models into sparse interpretable concepts and uses a linear layer to predict categories.
CondDiff-AMO: Integrating Conditional Diffusion Mechanism for Unified Amodal Mask Generation
CaiJie Zhao (University of Macau), Bob Zhang (University of Macau)
SegmentationTransformerDiffusion modelImage
🎯 What it does: Propose CondDiff-AMO, which treats the task of generating complete masks for blindly occluded objects as a step-by-step denoising problem using diffusion models, providing a single-stage, unsupervised panoramic mask prediction framework.
Condensed Data Expansion Using Model Inversion for Knowledge Distillation
Kuluhan Binici (SAP), Tulika Mitra (National University Of Singapore)
Data SynthesisKnowledge DistillationImage
🎯 What it does: Expand sparse datasets using model inversion techniques and apply them to knowledge distillation to improve the accuracy of student models.
Conditional Diffusion Model for Multi-Agent Dynamic Task Decomposition
Yanda Zhu (Nanjing University), Chunlin Chen (Nanjing University)
Convolutional Neural NetworkRecurrent Neural NetworkTransformerReinforcement LearningDiffusion modelBenchmark
🎯 What it does: Proposed the CD³T framework, which utilizes conditional diffusion models to learn action representations and achieve dynamic subtask decomposition in multi-agent tasks, thereby improving collaboration efficiency.
Conditional Distribution Learning for Graph Classification
Jie Chen (Sichuan University), Xi Peng (Sichuan University)
ClassificationRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Proposed a semi-supervised graph classification method CDL, which learns graph representations by aligning conditional distributions between weakly and strongly augmented views.
Conditional Information Bottleneck for Multimodal Fusion: Overcoming Shortcut Learning in Sarcasm Detection
Yihua Wang (Beijing Normal University), Zhichun Wang (IEIT SYSTEMS Co., Ltd.)
ClassificationRepresentation LearningTransformerMultimodality
🎯 What it does: Proposed a multimodal fusion framework MCIB based on conditional information bottleneck, achieving multimodal sarcasm detection on the MUStARD++ R dataset with shortcut learning signals removed.
Conditional Probabilistic Bipolar Argumentation Framework: Explanations, Complexity and Approximation
Gianvincenzo Alfano (University of Calabria), Irina Trubitsyna (University of Calabria)
Explainability and InterpretabilityComputational Efficiency
🎯 What it does: Propose a conditional probability bipolar argumentation framework (CPBAF) that integrates conditional probability with attack/support, supporting cyclic structures;
Conditional Prompt Learning via Degradation Perception for Underwater Image Enhancement
Mingze Yao (Dalian Maritime University), Huibing Wang (Dalian Maritime University)
RestorationConvolutional Neural NetworkTransformerPrompt EngineeringVision Language ModelContrastive LearningImage
🎯 What it does: Proposed a Conditional Prompt Learning via Degradation Perception (CPLDP) model, which enhances visual quality by achieving perception and enhancement of different underwater degradations through conditional prompt learning, dynamic learning network, and adaptive loss function.
ConfGuard: A Simple and Effective Backdoor Detection for Large Language Models
Zihan Wang (University of Electronic Science and Technology of China), Guowen Xu (University of Electronic Science and Technology of China)
Anomaly DetectionTransformerLarge Language ModelText
🎯 What it does: Proposed ConfGuard, a lightweight LLM backdoor detection method based on a sliding window of token confidence;
Confidence Estimation for Text-to-SQL in Large Language Models
Sepideh Entezari Maleki (University of Alberta), Davood Rafiei (University of Alberta)
Explainability and InterpretabilityAI Code AssistantTransformerLarge Language ModelTextTabularBenchmark
🎯 What it does: This paper proposes and systematically evaluates multiple confidence estimation methods for assessing the reliability of generated SQL queries in text-to-SQL tasks by large language models (LLMs) without obtaining golden answers;
Confidence-Guided Stepwise Model Routing for Cost-Efficient Reasoning
Sangmook Lee (Seoul National University), Kyomin Jung (Seoul National University)
Computational EfficiencyKnowledge DistillationLarge Language ModelTextBenchmark
🎯 What it does: Proposed a model confidence-based stepwise routing framework called STEER, which dynamically switches between small and large models during inference to reduce computational costs.
Conformable Convolution for Topologically Constrained Learning of Complex Anatomical Structures
Yousef Yeganeh (Technical University of Munich), Azade Farshad (ELLIS Institute Finland)
SegmentationConvolutional Neural NetworkBiomedical Data
🎯 What it does: Proposes Conformable Convolution and Topological Posterior Generator (TPG), explicitly embedding topological priors from persistent homology into convolutional networks to enhance topological connectivity in medical image segmentation.
Conformal Constrained Policy Optimization for Cost-Effective LLM Agents
Wenwen Si (University of Pennsylvania), Osbert Bastani (University of Pennsylvania)
Large Language ModelReinforcement LearningAgentic AIText
🎯 What it does: Propose a Constrained Conformal Policy Optimization (CCPO) framework based on online synthetic prediction to achieve collaborative decision-making for cost minimization and reliability guarantees in multi-model LLM agents
Conformal Prediction for Multi-Source Detection on a Network
Xingchao Jian (Technical University of Munich), Felix Krahmer (Technical University of Munich)
Anomaly DetectionGraph Neural NetworkGraph
🎯 What it does: Propose a network multi-source detection framework based on homomorphic prediction, which can estimate the set of infection sources without relying on diffusion model assumptions.
Conformal Prediction Meets Long-tail Classification
Shuqi Liu (Nanyang Technological University), Luke Ong (Nanyang Technological University)
ClassificationImage
🎯 What it does: Propose Tail‑Aware Conformal Prediction (TACP) and its soft version sTACP to balance prediction set coverage under long-tailed label distributions.
ConInstruct: Evaluating Large Language Models on Conflict Detection and Resolution in Instructions
Xingwei He, Siu-Ming Yiu (University of Hong Kong)
TransformerLarge Language ModelTextBenchmark
🎯 What it does: Construct the ConInstruct dataset and evaluate the conflict detection and resolution capabilities of LLMs under instructions with conflicting constraints.
Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning
Liqin Luo (Peking University), Yonghong Tian (Peking University)
RetrievalTransformerLarge Language ModelAgentic AIVision Language ModelImageTextMultimodalityChain-of-Thought
🎯 What it does: This paper proposes a training-agnostic visual alignment framework named GroundingAgent, which leverages open-vocabulary detectors, cross-modal large language models, and regular large language models for structured agent reasoning to achieve zero-shot visual localization.
Connectivity-Guided Sparsification of 2-FWL GNNs: Preserving Full Expressivity with Improved Efficiency
Rongqin Chen (University of Macau), Leong Hou U (University of Macau)
Computational EfficiencyRepresentation LearningDrug DiscoveryGraph Neural NetworkGraph
🎯 What it does: Proposed a sparsification framework called Co-Sparsify, guided by graph connectivity, to enhance the efficiency of 2-FWL graph neural networks while maintaining full expressive power.
Consensus Learning with Multi-Party Perturbation Triggers for Secure Model Access
Yizhun Zhang (Southeast University), Xuan Chen (Southeast University)
Safty and PrivacyConvolutional Neural NetworkImage
🎯 What it does: Proposed a Consensus Learning paradigm based on multi-party joint authorization, utilizing a collaborative perturbation trigger mechanism to achieve secure model access control.
Consensus-Aligned Neuron Efficient Fine-Tuning Large Language Models for Multi-Domain Machine Translation
Shuting Jiang (Kunming University of Science and Technology), Zhengtao Yu (Kunming University of Science and Technology)
Domain AdaptationComputational EfficiencyRepresentation LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Proposes the CANEFT framework, achieving efficient fine-tuning for multi-domain machine translation by identifying and updating only consistent aligned neurons;
Consensus-Driven Multi-Agent Cognitive Reasoning for Enhancing the Emotional Intelligence of Large Language Models
Geng Tu (Harbin Institute of Technology), Ruifeng Xu (Harbin Institute of Technology)
Explainability and InterpretabilityTransformerLarge Language ModelAgentic AITextBenchmarkChain-of-Thought
🎯 What it does: Propose the MACRo framework, which employs multi-agent generation of structured emotional chains CogCoT (Context, Cue, Thought, Action, Emotion), and ensures reasoning coherence through coordination and consensus games.
Consistency-based Abductive Reasoning over Perceptual Errors of Multiple Pre-trained Models in Novel Environments
Mario Leiva (Universidad Nacional del Sur), Gerardo Simari (Syracuse University)
Domain AdaptationAnomaly DetectionImage
🎯 What it does: This paper proposes a method based on consistency-based refutation reasoning to correct perception errors of multiple pre-trained models in new environments.
ConsistTalk: Intensity Controllable Temporally Consistent Talking Head Generation with Diffusion Noise Search
Zhenjie Liu (University of Science and Technology of China), Shangfei Wang (University of Science and Technology of China)
GenerationKnowledge DistillationConvolutional Neural NetworkDiffusion modelOptical FlowVideoMultimodalityAudio
🎯 What it does: Propose the ConsistTalk framework for audio-driven talking head generation, addressing issues such as flickering, identity drift, and visual-audio synchronization inconsistencies.
Constrained and Robust Policy Synthesis with Satisfiability-Modulo-Probabilistic-Model-Checking
Linus Heck (Radboud University), Sebastian Junges (Radboud University)
OptimizationReinforcement LearningBenchmark
🎯 What it does: We propose the SMPMC framework, which tightly integrates a SAT solver with a probabilistic model checker for automatically synthesizing optimal strategies in constrained and robust Markov decision processes.
Constrained Best Arm Identification with Tests for Feasibility
Ting Cai (University of Wisconsin-Madison), Kirthevasan Kandasamy (University of Wisconsin-Madison)
OptimizationDrug DiscoveryBiomedical Data
🎯 What it does: Proposed a new framework for Best Arm Identification (BAI) that allows testing feasibility constraints independently, and designed an adaptive algorithm that simultaneously decides whether to pull performance distribution or constraint distribution.
Constrained Molecule Generation Modelled Using the Grammar Constraint
David Saikali (Polytechnique Montréal), Gilles Pesant (Polytechnique Montréal)
GenerationDrug DiscoveryTextBiomedical Data
🎯 What it does: Propose a molecular generation model based on context-free grammar (CFG) constraints using SMILES, incorporating multiple chemical and physicochemical property constraints.
Constrained Online Convex Optimization with Memory and Predictions
Mohammed Abdullah (Université Paris-Saclay), Tijani Chahed (Delft University of Technology)
Optimization
🎯 What it does: This paper investigates the scenario of incorporating a memory window (the past m decisions) and unreliable predictions within the constrained online convex optimization (COCO) framework, proposing two categories of algorithms—one suitable for scenarios without predictions, and another that employs an 'optimistic learning' approach when short-term predictions are available, providing corresponding sublinear loss bounds and theoretical upper bounds on cumulative constraint violation (CCV);
Constrained Particle Seeking: Solving Diffusion Inverse Problems with Just Forward Passes
Hongkun Dou (Beihang University), Yue Deng (Beihang University)
RestorationOptimizationDiffusion modelImagePhysics Related
🎯 What it does: Propose a gradient-free particle optimization method called CPS, which utilizes information from all candidate particles combined with high-density constraints to solve inverse problems in diffusion models without requiring gradients.
Constraint Optimization of MicroPlate Designs
Ramiz Gindullin (Uppsala University), María Andreína Francisco Rodríguez (Uppsala University)
OptimizationBiomedical Data
🎯 What it does: Propose a constrained optimization model (COMPD) to generate microplate layouts, aiming to reduce plate effects and improve experimental data quality.
Constraint-Guided Clustering for Identifying in-Vehicle Electronic Control Units from Voltage Data
Bogdan Groza (Politehnica University of Timisoara), Lucian Popa (Politehnica University of Timisoara)
Autonomous DrivingTime Series
🎯 What it does: Studied unsupervised clustering based on voltage features in vehicle internal electronic control units (ECUs) and proposed a method that utilizes domain constraints to guide the selection of the number of clusters.
Constraints-Guided Diffusion Reasoner for Neuro-Symbolic Learning
Xuan Zhang (Fudan University), Yuan Qi (Fudan University)
OptimizationReinforcement LearningDiffusion model
🎯 What it does: Propose a diffusion model named DDReasoner, serving as a symbolic reasoner. It first acquires basic reasoning capabilities through supervised learning with masked DDPM, then treats the denoising process as a Markov Decision Process (MDP), and fine-tunes the model using reinforcement learning (PPO-CLIP) to ensure outputs meet hard logical constraints.
Constructing Superior Representations Beyond the Original Documents via a Contrastive Gaussian Fusion Network for Clustering
Ao Shen (Guizhou University), Ruina Bai (Guizhou University)
Representation LearningAuto EncoderContrastive LearningText
🎯 What it does: Proposes the Contrastive Gaussian Fusion Network (CGFN), which constructs superior document representations for unsupervised text clustering by fusing neighbor information with the Gaussian distribution of original text in the latent space, combined with contrastive learning.
ConSurv: Multimodal Continual Learning for Survival Analysis
Dianzhi Yu (Chinese University of Hong Kong), Irwin King (Hong Kong University of Science and Technology)
TransformerMixture of ExpertsMultimodalityBiomedical DataBenchmark
🎯 What it does: This paper proposes ConSurv, a framework for multi-modal continual learning (WSI and genomic) for cancer survival prediction.
Content Diversity-guided Ambiguity Mitigation for Open-Set Noisy Label Learning
Zhihao Zhou (Qingdao University of Science and Technology), Xueying Li (Qingdao University of Science and Technology)
ClassificationData-Centric LearningTransformerLarge Language ModelVision Language ModelImageMultimodality
🎯 What it does: Aiming at IDN and OOD samples in open-noise label learning, the CDgAM framework is proposed, dynamically refining class boundaries through multimodal text descriptions to achieve more precise sample purification.
Content-aware Information Compression and Selection for Whole Slide Image Analysis
Tingting Zheng (Harbin Institute of Technology), Yi Xiao (Zhengzhou University)
ClassificationComputational EfficiencyRepresentation LearningGraph Neural NetworkTransformerBiomedical Data
🎯 What it does: Proposed a content-aware information compression and selection framework (CICS) for efficient and accurate whole slide image (WSI) multi-instance learning;
Context-aware Dynamic Contrastive Learning Network and E-Bike Rider Benchmark for Person Search
Hongchao Li (Anhui Normal University), YongLong Luo (Anhui Normal University)
RetrievalConvolutional Neural NetworkContrastive LearningImageVideoBenchmark
🎯 What it does: This paper proposes the E-Bike Rider Search (EBRS) dataset specifically for urban electric bike riders and designs a Context-aware Dynamic Contrastive Learning (CDCL) framework combining context-aware mixed convolution and dynamic contrastive learning, aiming to address occlusion, pose variation, and scale differences in person search tasks.
Context-aware Graph Meta-learning
Ningbo Huang (State Key Laboratory of Mathematical Engineering and Advanced Computing), Yi Xia (State Key Laboratory of Mathematical Engineering and Advanced Computing)
ClassificationRepresentation LearningMeta LearningGraph Neural NetworkTransformerLarge Language ModelPrompt EngineeringTextGraphBenchmark
🎯 What it does: Proposes CAGML, a context-aware meta-learning model that unifies cross-domain and cross-granularity graph few-shot tasks through structural-aware sequence modeling, achieving in-graph parameter learning without fine-tuning via pre-trained Structure-Aware Transformer (SAT).
Context-Sensitive Abstractions for Reinforcement Learning with Parameterized Actions
Rashmeet Kaur Nayyar (Arizona State University), Siddharth Srivastava (Arizona State University)
Reinforcement Learning
🎯 What it does: This paper proposes the PEARL algorithm, which can adaptively learn context-aware abstractions for states and parameterized actions in reinforcement learning, and uses TD(λ) for policy learning in the abstract space.
ContextFlow: Training-Free Video Object Editing via Adaptive Context Enrichment
Yiyang Chen (Peking University), Jack Ma (Hong Kong University of Science and Technology)
GenerationTransformerDiffusion modelRectified FlowVideoBenchmark
🎯 What it does: Propose the ContextFlow framework to achieve training-free video object editing (insertion, replacement, deletion)
ContextGraph: Lifelog Intelligence Framework for Contextual Subgraph Evolution
Anil Sharma (Samsung R&D Institute India-Bangalore), Barath Raj Kandur Raja (Samsung R&D Institute India-Bangalore)
Explainability and InterpretabilityRepresentation LearningGraph Neural NetworkAuto EncoderMultimodalityGraphTime Series
🎯 What it does: Designed and implemented the ContextGraph framework, modeling smartphone multimodal lifecycle logs as temporal knowledge graphs, and computing daily context embeddings (DCE) through dual VAEs, followed by leveraging the Lens module to extract and track anchor-based subgraph evolution signatures, providing dynamic insights into user lifestyle patterns.
Continual Out-of-Distribution Detection with Analytic Neural Collapse
Saleh Momeni (University of Illinois Chicago), Bing Liu (University of Illinois Chicago)
ClassificationAnomaly DetectionTransformerContrastive LearningImage
🎯 What it does: Studied out-of-distribution (OOD) detection in the continual learning (CIL) scenario, proposing a training-free Analytic Neural Collapse (AnaNC) method that achieves continuous OOD detection and classification by constructing Neural Collapse geometry through analytical projection on frozen pre-trained models.
Continuous Degradation Modeling via Latent Flow Matching for Real-World Super-Resolution
Hyeonjae Kim (Hanyang University), Tae Hyun Kim (Hanyang University)
Data SynthesisSuper ResolutionFlow-based ModelAuto EncoderImage
🎯 What it does: Propose a continuous degradation modeling framework called DegFlow that leverages potential flow matching, enabling the synthesis of realistic low-resolution images of arbitrary scales from a single high-resolution image, thereby generating large-scale realistic scene SR training data.
Continuous Vision-Language-Action Co-Learning with Semantic-Physical Alignment for Behavioral Cloning
Xiuxiu Qi (Nankai University), Hongpeng Wang (Nankai University)
Robotic IntelligenceTransformerVision-Language-Action ModelAuto EncoderMultimodalityOrdinary Differential Equation
🎯 What it does: Developed a behavior cloning framework called CCoL, leveraging continuous multimodal collaborative learning and semantic-physical alignment to achieve smooth continuous control for language-driven robotic manipulation.
Continuum Dropout for Neural Differential Equations
Jonghun Lee (Ulsan National Institute of Science and Technology), Dong-Young Lim (Ulsan National Institute of Science and Technology)
ClassificationExplainability and InterpretabilityImageTime SeriesBiomedical DataElectronic Health RecordsStochastic Differential EquationOrdinary Differential EquationAudio
🎯 What it does: Proposed a continuous-time dropout mechanism called Continuum Dropout, and applied it to Neural Differential Equations (NDE) to enhance model generalization capability and uncertainty quantification.
Contribution-aware Token Compression for Efficient Video Understanding via Reinforcement Learning
Yinchao Ma (Taobao & Tmall Group of Alibaba), Bo Zheng (Taobao & Tmall Group of Alibaba)
CompressionComputational EfficiencyTransformerReinforcement LearningVision Language ModelVideo
🎯 What it does: Propose CaCoVID, a reinforcement learning-based video large language model compression method, which reduces redundancy by training a small policy network to actively select visual tokens that contribute the most to the answer.
Control Illusion: The Failure of Instruction Hierarchies in Large Language Models
Yilin Geng (The University of Melbourne), Lea Frermann (The Hebrew University of Jerusalem)
TransformerLarge Language ModelText
🎯 What it does: Evaluate the hierarchical priority of large language models when facing conflicting instructions, propose a systematic assessment framework based on constraint prioritization and new metrics, and systematically test the performance of six mainstream LLMs;
ControlFuse: Instruction-guided Multi-Granularity Controllable Image Fusion
Libo Zhao (Jilin University), Zeyu Wang (Dalian Minzu University)
RestorationGraph Neural NetworkTransformerVision Language ModelContrastive LearningMultimodality
🎯 What it does: Proposes ControlFuse, an infrared-visible image fusion method controllable at global, semantic, and instance levels through user instructions.
Controllable Financial Market Generation with Diffusion Guided Meta Agent
Yu-Hao Huang (Nanjing University), Jiang Bian (Microsoft Research Asia)
GenerationData SynthesisDiffusion modelTime SeriesFinance Related
🎯 What it does: Designed Diffusion Guided Meta Agent (DigMA), generating controllable order flow in financial markets scenarios through conditional diffusion models and financial economics priors;
Convergence of Fast Policy Iteration in Markov Games and Robust MDPs
Keith Badger (University of New Hampshire), Marek Petrik (University of New Hampshire)
OptimizationComputational EfficiencyTabularBenchmark
🎯 What it does: Study the convergence of fast policy iteration algorithms for Markov games and robust MDPs, prove that the original Filar-Tolwinski (FT) algorithm may not converge, and propose a new Residual Conditioned Policy Iteration (RCPI) algorithm that ensures convergence while maintaining speed.
Convergent Semantics for Weighted Bipolar Argumentation
Zongshun Wang (Sun Yat-sen University), Yuping Shen (Sun Yat-sen University)
Graph
🎯 What it does: Proposed three bidirectional asymptotic semantics (QBS, CBS, HBS) that converge on weighted bipolar argumentation graphs (wBAG), addressing the divergence issue of traditional semantics in cyclic graphs.
Conversational Learning Diagnosis via Reasoning Multi-Turn Interactive Learning
Fangzhou Yao (University of Science and Technology of China), Qi Liu (University of Science and Technology of China)
Explainability and InterpretabilityTransformerLarge Language ModelPrompt EngineeringTextSequentialChain-of-Thought
🎯 What it does: Proposed the ParLD framework, which uses behavior preview, state analysis, and performance reasoning in multi-turn dialogues to track students' cognitive states in real-time and support teaching decisions.
Convex Clustering Redefined: Robust Learning with the Median of Means Estimator
Koustav Chowdhury (Indian Statistical Institute), Saptarshi Chakraborty (Indian Statistical Institute)
OptimizationTabularBiomedical DataBenchmark
🎯 What it does: Proposed a robust clustering framework COMET that combines convex clustering with Median-of-Means (MoM) estimation, along with an Adam-based optimization algorithm.
ConvMix: A Mixed-Criteria Data Augmentation Framework for Conversational Dense Retrieval
Fengran Mo (University of Montreal), Jian-Yun Nie (University of Montreal)
RetrievalData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes the ConvMix framework, which utilizes large language models to perform bidirectional relevance judgment augmentation on conversational retrieval data, and combines quality control with near-distribution supervision to significantly enrich the training set.
Cook and Clean Together: Teaching Embodied Agents for Parallel Task Execution
Dingkang Liang (Huazhong University of Science and Technology), Xiang Bai (Huazhong University of Science and Technology)
OptimizationRobotic IntelligenceTransformerLarge Language ModelVision-Language-Action ModelPoint Cloud
🎯 What it does: Proposes the 3D Oriented Task Scheduling (ORS3D) problem based on operations research knowledge, requiring agents to simultaneously generate efficient parallel task plans and locate target objects in real 3D scenarios.
Cooperative Graph Transformer with Structural Consensus for Multi-View Learning
Zhiyuan Lai (Fuzhou University), Shiping Wang (Fuzhou University)
Representation LearningGraph Neural NetworkTransformerGraph
🎯 What it does: Propose CoGFormer by jointly using denoised structural consensus graph convolution and structure-guided attention mechanisms to achieve fusion of local and global representations in multi-view learning.
CoordAR: One-Reference 6D Pose Estimation of Novel Objects via Autoregressive Coordinate Map Generation
Dexin Zuo (Shanghai Jiao Tong University), Danping Zou (Shanghai Jiao Tong University)
Pose EstimationTransformerAuto EncoderImage
🎯 What it does: Propose a 6D pose estimation framework under a single reference view, CoordAR, which predicts 3D-3D correspondences between the reference view and the query view by generating discrete coordinate tokens through autoregression, and calculates the object's pose based on this.
Coordinated Humanoid Robot Locomotion with Symmetry Equivariant Reinforcement Learning Policy
Buqing Nie (Shanghai Jiao Tong University), Yue Gao (Shanghai Jiao Tong University)
Robotic IntelligenceReinforcement Learning
🎯 What it does: Proposed the Symmetry Equivariant Policy (SE-Policy), achieving bilateral symmetric locomotion control in the DRL framework through strict symmetry equivariance (actor) and invariance (critic);
Copyright Infringement Detection in Text-to-Image Diffusion Models via Differential Privacy
Xiafeng Man, Jingjing Chen (Fudan University)
Anomaly DetectionSafty and PrivacySupervised Fine-TuningVision Language ModelDiffusion modelImage
🎯 What it does: Propose a copyright infringement detection framework called D-Plus-Minus (DPM) based on differential privacy theory, which evaluates the model's conditional sensitivity by fine-tuning two branches for learning and forgetting the target concept, thereby determining whether infringement exists.
CoRA: A Collaborative Robust Architecture with Hybrid Fusion for Efficient Perception
Gong Chen (Tianjin University), Xiaohui Xie (Tianjin University)
Object DetectionPose EstimationAutonomous DrivingConvolutional Neural NetworkTransformerMultimodality
🎯 What it does: This paper proposes CoRA, a collaborative perception framework with low communication overhead and strong robustness, which employs a dual-branch architecture to simultaneously achieve high-performance feature-level fusion and object-level correction for pose errors.
CoRe-Fed: Bridging Collaborative and Representation Fairness via Federated Embedding Distillation
Noorain Mukhtiar (Macquarie University), Quan Z. Sheng (Macquarie University)
Federated LearningKnowledge DistillationRepresentation LearningContrastive LearningImage
🎯 What it does: Proposes the CoRe-Fed framework by simultaneously addressing representation fairness and collaborative fairness in federated learning;
CoRE-Learning with Look-Ahead and Immediate Resource Allocation
Jing Wang (Nanjing University), Zhi-Hua Zhou (Nanjing University)
OptimizationComputational EfficiencyImage
🎯 What it does: This study addresses the scheduling problem in multi-task learning under limited computational resources and proposes the LaiRA algorithm.
CorrectAD: A Self-Correcting Agentic System to Improve End-to-end Planning in Autonomous Driving
Enhui Ma (Zhejiang University), Kaicheng Yu (Autolab, Westlake University)
Autonomous DrivingTransformerSupervised Fine-TuningAgentic AIVision Language ModelDiffusion modelVideoMultimodality
🎯 What it does: Proposed a fully automated self-correcting agent system called CorrectAD, which leverages PM-Agent to analyze driving failure cases and generate multimodal requirements, then uses DriveSora to synthesize high-fidelity multi-perspective, 3D layout-consistent videos, thereby enhancing the robustness of the end-to-end planning model without altering the original model.
Correcting False Alarms from Unseen: Adapting Graph Anomaly Detectors at Test Time
Junjun Pan (Griffith University), Shirui Pan (Griffith University)
Domain AdaptationAnomaly DetectionGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Propose a test-time adaptation framework called TUNE to address distribution drift caused by the emergence of unknown normal patterns in graph anomaly detection.
Correcting Quantization-Induced Gradient Mismatch in Neural Image Compression
Changhao Peng (Peking University), Wei Gao (Peking University)
CompressionAuto EncoderImage
🎯 What it does: This paper addresses the gradient mismatch problem caused by quantization in neural image compression, proposing two improvements: ① using STE quantization to ensure consistency between likelihood calculations during training and inference, thereby correcting the gradient direction of standard deviation parameters; ② manually setting the gradient of the mean parameter in the |y-µ|<0.5 interval (MSE-guided or gradient transfer), achieving precise learning of entropy model parameters. This method does not alter the original network structure and only requires freezing the main trunk and performing lightweight fine-tuning on a pre-trained model.
CorrectNav: Self-Correction Flywheel Empowers Vision-Language-Action Navigation Model
Zhuoyuan Yu (CFCS, School of Computer Science, Peking University), Hao Dong (PKU-Agibot Lab)
Autonomous DrivingRobotic IntelligenceTransformerLarge Language ModelVision-Language-Action ModelImageVideoTextMultimodality
🎯 What it does: This paper proposes the Self-correction Flywheel post-training paradigm and trains the CorrectNav model based on this paradigm, achieving vision-language action navigation using only monocular RGB and language instructions.
Correspondence Coverage Matters for Multi-Modal Dataset Distillation
Zhuohang Dang (Xi'an Jiaotong University), Ivor Tsang (Agency for Science, Technology and Research)
Knowledge DistillationMultimodality
🎯 What it does: Propose the ProCo framework, which initializes and regularizes multimodal data distillation data using retrieval consistency clustering, enhancing correspondence coverage and diversity.
CoS: Towards Optimal Event Scheduling via Chain-of-Scheduling
Yiming Zhao (Sun Yat-sen University), Jian Yin (Sun Yat-sen University)
OptimizationKnowledge DistillationTransformerLarge Language ModelSupervised Fine-TuningTextChain-of-Thought
🎯 What it does: This paper proposes the Chain-of-Scheduling (CoS) framework, which leverages the ordered reasoning of large language models across three stages—exploration, verification, and integration—to automatically generate event schedules that comply with time and location constraints;
COSMOS: Coherent SuperGaussian Modeling with Spatial Priors for Sparse-View 3D Splatting
Chaeyoung Jeong (Sungkyunkwan University), Kwangsu Kim (Sungkyunkwan University)
GenerationTransformerGaussian SplattingImage
🎯 What it does: Proposes COSMOS, which introduces spatial priors for 3D Gaussian Splatting by leveraging supergaussian grouping based on local geometric and appearance features, enhancing generalization performance in sparse-view 3D reconstruction.
Cost-Effective Communication: An Auction-based Method for Language Agent Interaction
Yijia Fan (Sun Yat-sen University), Keze Wang (Snap Inc)
OptimizationComputational EfficiencyLarge Language ModelReinforcement LearningAgentic AITextBenchmark
🎯 What it does: Introduce DALA, an auction-based communication mechanism in multi-agent LLM systems, which significantly reduces communication costs and improves task performance by treating communication as a scarce resource and allowing agents to bid based on value density to determine speaking rights;
Cost-Free Neutrality for the River Method
Michelle Döring, Stefan Neubert (Hasso Plattner Institute University of Potsdam)
OptimizationComputational EfficiencyTabular
🎯 What it does: To address the Parallel-Universe Tiebreaking (PUT) problem under the River voting rule, this paper proposes a Multi-Path Fusion (FUN) algorithm that simultaneously simulates the River process across multiple tie-breaking orders, thereby computing all PUT winners in polynomial time.
Cost-Sensitive Conformal Training with Provably Controllable Learning Bounds
Xuesong Jia (Washington State University), Yan Yan (Washington State University)
ClassificationConvolutional Neural NetworkTransformerImageText
🎯 What it does: This paper proposes a label ranking weight-based cost-sensitive conformal training algorithm (RWCE), which directly optimizes the weighted cross-entropy of true label rankings to approximate and upper bound the prediction set size;