AAAI 2026 Papers with AI Summaries
AAAI Conference on Artificial Intelligence · 4149 papers
→ AAAI 2026 papers with code (2140)
Each paper below shows an AI-generated one-line summary. Get the full 6-part summary (innovation, method, data, results, limitations) and search all 4149 AAAI 2026 papers by keyword, author or institution —
free trial on arXivSub.
“As Eastern Powers, I Will Veto.”: An Investigation of Nation-Level Bias of Large Language Models in International Relations
Jonghyeon Choi (Yonsei University), Beakcheol Jang (Yonsei University)
TransformerLarge Language ModelTextRetrieval-Augmented Generation
🎯 What it does: This paper systematically evaluates country-level bias in large language models within the field of international relations, constructing a real-world dataset using United Nations Security Council resolutions and voting records, and designing three experiments (direct question answering, association testing, and voting simulation) to quantify bias.
2-ASP(Q) Solving Based on CEGAR
Andrea Cuteri (University of Calabria), Francesco Ricca (University of Calabria)
OptimizationBenchmark
🎯 What it does: Propose a 2-ASP(Q) solving method based on CEGAR and implement the CASPER system.
2D Gaussians Spatial Transport for Point-supervised Density Regression
Miao Shang (Harbin Institute of Technology), Xiaopeng Hong (Harbin Institute of Technology)
Object DetectionPose EstimationConvolutional Neural NetworkTransformerGaussian SplattingImagePoint Cloud
🎯 What it does: This paper proposes the Gaussian Spatial Transport (GST) framework for point annotation density regression tasks. By first performing 2D Gaussian splatting on the input image to estimate the conditional probability between pixels and annotated points, a precomputable transport kernel is constructed; subsequently, during network training, the predicted density map is projected into the annotation space with a single matrix multiplication to compute the error, thereby avoiding the high computational cost of traditional OT iterative solutions for transport plans.
2D-CrossScan Mamba: Enhancing State Space Models with Spatially Consistent Multi-Path 2D Information Propagation
Longlong Yu (Hangzhou Dianzi University), Yuchen Guo (Hangzhou Dianzi University)
ClassificationObject DetectionSegmentationImage
🎯 What it does: Proposed a 2D-compatible scanning framework called 2D-CrossScan, improving state space models such as Mamba to enable spatially consistent multi-path information propagation in 2D images.
360Explorer: Exploring 4D Controllable World in Panoramic Videos
Xinhua Cheng (Peking University Shenzhen Graduate School), Li Yuan (Peking University Shenzhen Graduate School)
GenerationDepth EstimationDiffusion modelVideoPoint Cloud
🎯 What it does: Propose 360Explorer, a 4D controllable video generation framework based on dynamic point clouds and panoramic perspectives, enabling exploration and manipulation within panoramic videos according to user-defined 3D camera trajectories and object motions.
3D-ANC: Adaptive Neural Collapse for Robust 3D Point Cloud Recognition
Yuanmin Huang (Fudan University), Min Yang (Fudan University)
ClassificationRecognitionRepresentation LearningAdversarial AttackPoint CloudBenchmark
🎯 What it does: Propose the 3D-ANC method, which employs an ETF structure-based classification head in point cloud classification models and enhances adversarial robustness through an adaptive training framework.
3D-DRES: Detailed 3D Referring Expression Segmentation
Qi Chen (Xiamen University), Liujuan Cao (Xiamen University)
SegmentationConvolutional Neural NetworkTransformerVision Language ModelTextPoint Cloud
🎯 What it does: This paper proposes the detailed 3D referring expression segmentation (3D-DRES) task, which requires the model to map each noun phrase in a sentence to its corresponding 3D instance and generate a segmentation mask.
3DAlign-DAER: Dynamic Attention Policy and Efficient Retrieval Strategy for Fine-grained 3D-Text Alignment at Scale
Yijia Fan (Sun Yat-sen University), Keze Wang (Sun Yat-sen University)
ClassificationRetrievalRepresentation LearningTransformerContrastive LearningTextMultimodalityPoint Cloud
🎯 What it does: Proposes the 3DAlign-DAER framework to achieve fine-grained cross-modal alignment between 3D shapes and text.
3DDM: Physically-based Anisotropic 3D Diffusion Model with 3D Gaussian for Point Cloud Completion
Long Xi (Xi'an Polytechnic University), Wen Lv (Bournemouth University)
RestorationTransformerDiffusion modelPoint CloudStochastic Differential Equation
🎯 What it does: Proposed a physics-based anisotropic three-dimensional diffusion model (3DDM) for recovering complete 3D models from partial point clouds.
3DTeethSAM: Taming SAM2 for 3D Teeth Segmentation
Zhiguo Lu, Kun Zhou (Tianjin University)
SegmentationConvolutional Neural NetworkTransformerPoint CloudMeshBiomedical DataBenchmark
🎯 What it does: Using the pre-trained SAM2 model, first render 3D dental meshes into multi-view 2D images, segment these images, and then project the 2D segmentation results back into 3D space through voting and Graph Cut, achieving automated 3D dental instance segmentation and semantic annotation.
3One2: One-Step Regression plus One-Step Diffusion for One-Hot Modulation in Dual-Path Video Snapshot Compressive Imaging
Ge Wang (Zhejiang University), Xin Yuan (Westlake Institute for Optoelectronics)
RestorationCompressionConvolutional Neural NetworkDiffusion modelVideoStochastic Differential Equation
🎯 What it does: For snapshot compressive imaging with a thermal mask, this paper proposes a hybrid regression + diffusion reconstruction framework called RegDif, and implements dual optical channels at the hardware level to enhance reconstruction quality.
4D Point Cloud Segmentation via Active Test-Time Adaptation
Mingrong Gong, Sergio Escalera (University Of Hong Kong)
SegmentationDomain AdaptationAutonomous DrivingPoint Cloud
🎯 What it does: Proposes ATTA-4DSeg, an active test-time adaptation framework for 4D point cloud segmentation, enabling online model adjustment under extremely low annotation budgets.
4D Scaffold Gaussian Splatting with Dynamic-Aware Anchor Growing for Efficient and High-Fidelity Dynamic Scene Reconstruction
Woong Oh Cho (Yonsei University), Seon Joo Kim (Yonsei University)
GenerationNeural Radiance FieldGaussian SplattingVideo
🎯 What it does: Propose a framework based on 4D anchors, generating neural 4D Gaussians via sparse grid anchors to achieve high-quality dynamic scene reconstruction with significant storage compression.
4DSTR: Advancing Generative 4D Gaussians with Spatial-Temporal Rectification for High-Quality and Consistent 4D Generation
Mengmeng Liu (University of Twente), Hao Cheng (University of Bath)
GenerationData SynthesisRecurrent Neural NetworkDiffusion modelScore-based ModelGaussian SplattingVideoText
🎯 What it does: Propose the 4DSTR network to achieve high-quality and highly consistent generation from video to 4D, combining space-time correction technology to enhance motion realism.
6DAttack: Backdoor Attacks in the 6DoF Pose Estimation
Jihui Guo (University of Hong Kong), Xinlei He (Hong Kong University of Science and Technology)
Pose EstimationAdversarial AttackConvolutional Neural NetworkImageVideo
🎯 What it does: Proposed the 6DAttack framework to perform backdoor attacks on 6DoF pose estimation models.
A Better Start: Sensitivity-Aware Warm-Up for Robust and Efficient Fine-Tuning
Yile Chen (South China University of Technology), Jin Huang (South China Normal University)
OptimizationComputational EfficiencyTransformerSupervised Fine-TuningTextBenchmark
🎯 What it does: Proposes a learning rate sensitivity-based adaptive warm-up strategy (SAWU) for the warm-up phase in large model fine-tuning, dynamically adjusting loss, learning rate, and phase switching;
A Boundary Token Graph for Zero-Shot Relation Triplet Extraction Involving Discontinuous Entities
Kailun Lyu (Northeastern University), Jingwei Cheng (Northeastern University)
RecognitionTransformerPrompt EngineeringText
🎯 What it does: This paper studies the challenges of discrete entities in zero-shot relation triplet extraction (ZSRTE), proposing a new framework based on the Boundary Marking Graph (BoG) that transforms triplet extraction into a graph path search problem.
A Brain-Inspired Saliency Prediction Framework for Human-AI Cognitive Consistency in AIGC Content via Multi-Region Liquid Neurons
Shibo Wang (Jilin University), Shuo Li (Jilin University)
GenerationData SynthesisKnowledge DistillationTransformerMultimodality
🎯 What it does: This paper proposes an attention prediction framework based on a liquid neuron model in the brain's visual regions and a dual teacher distillation mechanism, aimed at achieving human-machine cognitive consistency in AIGC content evaluation.
A Catalyst Framework for the Quantum Linear System Problem via the Proximal Point Algorithm
Junhyung Lyle Kim (Rice University), Anastasios Kyrillidis (Rice University)
OptimizationPhysics Related
🎯 What it does: Designed a meta-algorithm for solving quantum linear systems based on the proximal point algorithm, which can wrap any existing QLSP solver and use a tunable step size for preprocessing, significantly reducing dependence on the condition number without sacrificing accuracy.
A Causal Target for Learning to Defer Under Hidden Confounding
Yanmin Li (National University of Defense Technology), Weidong Bao (National University of Defense Technology)
OptimizationTabular
🎯 What it does: Proposes a causal objective based on sharp interval estimation to learn a rejectable (pending) strategy under hidden confounding;
A Closer Look at Knowledge Distillation in Spiking Neural Network Training
Xu Liu, Dan Guo (School Of Computer Science And Information Engineering Hefei University Of Technology)
Computational EfficiencyKnowledge DistillationSpiking Neural NetworkImage
🎯 What it does: Proposed two novel knowledge distillation strategies—Significance-Scaled Activation Map Distillation (SAMD) and Noise-Smoothed Logit Distillation (NLD)—to effectively transfer knowledge from pre-trained artificial neural networks (ANNs) to spiking neural networks (SNNs) and enhance their training performance.
A Content-Preserving Secure Linguistic Steganography
Lingyun Xiang (Changsha University of Science and Technology), Yuling Liu (Hunan University)
Safty and PrivacyTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: Propose CLstega, achieving content-preserving linguistic steganography without altering the original text, utilizing controllable distribution transformations of masked language models to embed and extract secret information.
A Differential Perspective on Distributional Reinforcement Learning
Juan Sebastian Rojas (University of Toronto), Chi-Guhn Lee (University of Toronto)
Reinforcement Learning
🎯 What it does: Propose a distributed reinforcement learning framework under the average reward setting, and design D2 and D3 algorithms to learn or optimize the distribution of long-term per-step rewards.
A Disease-Aware Dual-Stage Framework for Chest X-ray Report Generation
Puzhen Wu (Weill Cornell Medicine), Yifan Peng (Weill Cornell Medicine)
GenerationTransformerLarge Language ModelVision Language ModelContrastive LearningBiomedical DataRetrieval-Augmented Generation
🎯 What it does: Propose a two-stage disease-aware framework that first learns disease semantic tags (DASTs) through vision-text alignment, then fuses them with visual features to generate chest X-ray reports.
A Domain-specific Heuristic for PDDL+-based Traffic Signal Optimisation
Francesco Doria (University of Calabria), Mauro Vallati (University of Huddersfield)
OptimizationTabularTime Series
🎯 What it does: This paper proposes a domain-specific heuristic hCAFE for PDDL+ traffic signal optimization and combines it with Greedy Best-First Search to build a planning framework deployable on actual traffic control infrastructure.
A Fast Heuristic Search Approach for Energy-Optimal Profile Routing for Electric Vehicles
Saman Ahmadi (RMIT University), Mahdi Jalili (RMIT University)
Autonomous DrivingOptimizationGraph
🎯 What it does: This paper proposes a multi-objective A* search method based on label setting to solve the energy-optimal path archive for electric vehicles in large-scale road networks (i.e., solving for all possible initial energy levels simultaneously).
A Flat Minima Perspective on Understanding Augmentations and Model Robustness
Weebum Yoo (Ulsan National Institute of Science and Technology), Sung Whan Yoon (Ulsan National Institute of Science and Technology)
Domain AdaptationAdversarial AttackData-Centric LearningImage
🎯 What it does: This paper theoretically derives how label-preserving data augmentation enhances model robustness under distribution shift through the perspective of flat minima.
A General Anchor-Based Framework for Scalable Fair Clustering
Shengfei Wei (National University of Defense Technology), Lei Luo (National University of Defense Technology)
Computational EfficiencyTabularFinance Related
🎯 What it does: Designed a generic anchor-based fundamental framework (AFCF) to scale any fair clustering algorithm to linear time.
A General Highly Accurate Online Planning Method Integrating Large Language Models into Nested Rollout Policy Adaptation for Dialogue Tasks
Hui Wang (Anhui University), Chaoxu Mu (Anhui University)
OptimizationLarge Language ModelReinforcement LearningText
🎯 What it does: Proposed an online planner called NRPA-GD based on NRPA, which uses LLM to simulate user-system interactions and adaptively optimizes strategies directly during the conversation;
A Geometric Perspective on Optimizing Vector Quantized Latent Diffusion Model for Image Restoration
Chen Hang (East China Normal University), Haichuan Song (East China Normal University)
RestorationOptimizationDiffusion modelImageVideo
🎯 What it does: Proposed and implemented two geometry-based vector quantization latent space initial state optimization methods (interpolation method and Chebyshev center method), significantly enhancing the performance of VQ-LDM in image and video denoising, super-resolution, and other restoration tasks.
A GPU-based Constraint Programming Solver
Pierre Talbot (University of Luxembourg)
Optimization
🎯 What it does: Proposed and implemented Turbo, a discrete constraint solver that runs entirely on the GPU, employing integer interval boundary propagation and backtracking search. The core technology involves converting the constraint network into a ternary constraint network (TCN) and performing parallel propagation and on-demand subproblem search on the GPU.
A Graph-Theoretical Perspective on Law Design for Multiagent Systems
Qi Shi (University of Southampton), Pavel Naumov (University of Southampton)
OptimizationGraph
🎯 What it does: Studies the legal design problem in multi-agent systems, proposes two types of laws (useful laws and no-responsibility gap laws), and solves their minimization.
A Hybrid Space Model for Misaligned Multi-modality Image Fusion
Yi Xiao (Dalian University of Technology), Risheng Liu (Dalian University of Technology)
Convolutional Neural NetworkTransformerContrastive LearningMultimodality
🎯 What it does: Proposes a dual-space framework that jointly optimizes image registration and fusion to address error alignment and information fusion in infrared and visible images.
A Knowledge Compilation Map for Quantum Information
Lieuwe Vinkhuijzen (Leiden University), Alfons Laarman (Leiden University)
Review/Survey PaperPhysics Related
🎯 What it does: This paper constructs a knowledge compilation map for quantum information, systematically comparing the conciseness, operability, and speed of matrix product states (MPS), decision diagrams (ADD, SLDD×, LIMDD), and restricted Boltzmann machines (RBM).
A Logical Analysis of an Information Filtering Architecture Based on Epistemic Trust Inference
Xu Li (University of Luxembourg), Liuwen Yu (University of Luxembourg)
🎯 What it does: Studied an information filtering architecture based on cognitive trust reasoning and conducted a formal logical analysis of it.
A More Efficient Reduction from Outlier-Aware to Outlier-Free k-Median
Zhen Zhang (Hunan University of Technology and Business), Qilong Feng (Central South University)
OptimizationComputational Efficiency
🎯 What it does: Propose a sampling-based anchor point method that simplifies the k-means (with outliers) problem into an outlier-free problem, leading to a faster FPT approximation algorithm.
A Multi-Agent LLM Framework for Multi-Domain Low-Resource In-Context NER via Knowledge Retrieval, Disambiguation and Reflective Analysis
Wenxuan Mu, Yijia Zhang (Dalian Minzu University)
RecognitionLarge Language ModelAgentic AIPrompt EngineeringTextRetrieval-Augmented Generation
🎯 What it does: Propose a multi-agent large language model framework called KDR-Agent, specifically designed for context learning in named entity recognition in multi-domain low-resource scenarios.
A Multi-Objective Optimization Framework for Adaptive Weighting in Physics-Informed Machine Learning
Guoquan Wu (National University of Singapore), Zhe Wu (National University of Singapore)
OptimizationPhysics Related
🎯 What it does: This paper proposes an adaptive weight framework based on multi-objective optimization, which uses NSGA-II to search the Pareto frontier of PINN and dynamically allocates loss weights using the VARI method, thereby achieving a better balance between physical constraints and data errors.
A Multimodal EEG-Eye Movement Model for Automatic Depression Detection
Hao-Long Yin (Shanghai Jiao Tong University), Bao-Liang Lu (Shanghai Jiao Tong University)
ClassificationTransformerMixture of ExpertsContrastive LearningMultimodalityBiomedical Data
🎯 What it does: This study proposes a multimodal EEG-eye movement model called E Mo, which automatically detects depression using EEG and eye movement signals.
A Natural-Gradient Approach for Nonlinear Stochastic Systems with Parameter Uncertainty
Liang Cao (University of British Columbia)
OptimizationPhysics Related
🎯 What it does: Propose a nonlinear stochastic control framework based on polynomial chaos (PCE) and natural gradient, which can design robust state feedback controllers for nonlinear systems under parameter uncertainties and noise.
A Network of Biologically Inspired Rectified Spectral Units (ReSUs) Learns Hierarchical Features Without Error Backpropagation
Shanshan Qin (Flatiron Institute), Dmitri Chklovskii (Flatiron Institute)
Image TranslationRepresentation LearningTime SeriesBiomedical Data
🎯 What it does: Designed and trained a self-supervised neural network based on past-future Canonical Correlation Analysis (CCA) - Rectified Spectral Units (ReSUs) - achieving hierarchical features analogous to fly visual motion detection in a 1D natural image translation task.
A Novel Approach to Evaluating Evaluation Metrics for Multi-Output Structured Prediction
Akshay Vyas (University of Texas at Dallas), Nicholas Ruozzi (University of Texas at Dallas)
Image TranslationSegmentationGenerationTransformerContrastive LearningImageTextMultimodality
🎯 What it does: Built an evaluation framework based on task approximate invariance, using this framework to systematically evaluate the reliability of existing evaluation metrics in multi-output structured prediction tasks (image colorization, machine translation, image captioning), and added invariance auxiliary loss during model training, comparing model performance across multiple metrics before and after invariance regularization.
A Novel Fine-Tuned CLIP-OOD Detection Method with Double Loss Constraint Through Optimal Transport Semantic Alignment
Hengyang Lu (Jiangnan University), Chenyou Fan (Anhui University)
Anomaly DetectionRepresentation LearningTransformerSupervised Fine-TuningPrompt EngineeringVision Language ModelContrastive LearningMultimodality
🎯 What it does: This paper proposes a Fine-Tuned Out-of-Distribution (OOD) detection method called DOT-OOD for Vision-Language models (CLIP), addressing the low-focus attention problem.
A Novel Retrieve-Read-Group Paradigm for Open Knowledge Base Canonicalization
Binhan Yang (Nankai University), Han Tian (Nankai University)
Representation LearningTransformerContrastive LearningTextBenchmarkRetrieval-Augmented Generation
🎯 What it does: Proposed the Retrieve-Read-Group three-phase paradigm, and implemented the normalization of entity phrases in open knowledge bases (OKB) using the self-supervised DUVK framework.
A Novel Sliced Fused Gromov-Wasserstein Distance
Moritz Piening (Technical University Berlin), Robert Beinert (Technical University Berlin)
ClassificationOptimizationPoint CloudMeshGraph
🎯 What it does: Propose a novel sliced fused Gromov-Wasserstein distance (SFTLB) and achieve an efficient FGW lower bound based on it.
A Paradigm Shift in High-Resolution Depth Estimation Using SPAD-Based LiDAR Histograms: From Signal Filtering to Lightweight Similarity Learning
Minsung Lee (Ajou University), Jongmin Lee (Ajou University)
Depth EstimationComputational EfficiencyConvolutional Neural NetworkPoint Cloud
🎯 What it does: This paper proposes a lightweight similarity learning network called LITOFNET, which directly estimates depth from SPAD LiDAR histograms, avoiding traditional signal filtering methods.
A Parallel CPU-GPU Framework for Batching Heuristic Operations in Depth-First Heuristic Search
Ehsan Futuhi (University of Alberta), Nathan R. Sturtevant (University of Alberta)
OptimizationComputational EfficiencyBenchmark
🎯 What it does: Designed a CPU–GPU parallel framework that combines batched neural network heuristic evaluation with depth-first search (IDA*, BTS), achieving significant improvements in search efficiency.
A Phase Transition for Opinion Dynamics with Competing Biases
Federico Capannoli (Leiden University), Matteo Quattropani (Roma Tre University)
GraphPhysics Related
🎯 What it does: Studied binary opinion dynamics with competitive bias and stubbornness on a directed configuration model (DCM), and proved the existence of phase transitions;
A Principle-Driven Adaptive Policy for Group Cognitive Stimulation Dialogue for Elderly with Cognitive Impairment
Jiyue Jiang (Chinese University of Hong Kong), Chuan Wu (University of Hong Kong)
TransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningPrompt EngineeringText
🎯 What it does: Proposed GCSD—a principle-driven adaptive strategy for multi-party cognitive stimulation dialogues targeting elderly patients with cognitive impairments, integrating multi-speaker context control, dynamic participant cognitive state modeling, cognitive stimulation attention loss, and multi-dimensional reward optimization;
A Pseudo-Label Optimization Method Based on Polar Coordinate Modeling and Prior Constraints
Yudi Wang (Central South University), Zailiang Chen (Central South University)
SegmentationConvolutional Neural NetworkBiomedical DataMagnetic Resonance Imaging
🎯 What it does: Proposed a pseudo-label optimization method (PMPC) based on polar coordinate modeling and prior constraints to enhance the accuracy of medical image semi-supervised segmentation and reduce false positives.
A Reasoning Paradigm for Named Entity Recognition
Hui Huang (Guizhou University), Yongbin Qin (Guizhou University)
RecognitionTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningPrompt EngineeringTextRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: Proposed a Chain-of-Thought (CoT) based named entity recognition framework called ReasoningNER, which achieves explicit reasoning for entities through a three-phase process (CoT generation, CoT fine-tuning, and inference enhancement);
A Robust Unlearning Method with Adaptive Knowledge Guidance and Memory Preservation
Jingyuan Tian (Chinese Academy of Sciences), Xiaofei Zhou (University of Chinese Academy of Sciences)
Safty and PrivacyTransformerLarge Language ModelText
🎯 What it does: Proposed an Adaptive Local Memory Perturbation Unlearning (ALMPU) method to safely remove harmful knowledge in large language models.
A Rolling Stone Gathers No Moss: Adaptive Policy Optimization for Stable Self-Evaluation in Large Multimodal Models
Wenkai Wang (Zhejiang University), Shengyu Zhang (Zhejiang University)
OptimizationTransformerLarge Language ModelReinforcement LearningMultimodality
🎯 What it does: This paper studies and enhances the self-assessment capabilities of large multimodal models (LMMs) through the online reinforcement learning framework AdaPO.
A Scalable and Exact Relaxation for Densest k-Subgraph via Error Bounds
Ya Liu (Chinese University of Hong Kong), Aritra Konar (Chinese University of Hong Kong)
OptimizationGraph
🎯 What it does: Proposes an exact penalty continuous relaxation method based on the error bound principle for solving the dense k-subgraph (DkS) problem.
A Solution Space Transformation-Guided Co-Evolution for Energy-Saving Distributed Heterogeneous Flexible Job Shop Scheduling
Tao Li (Henan Normal University), Zhi-Hui Zhan (Henan Normal University)
OptimizationReinforcement LearningTabularBenchmark
🎯 What it does: Proposes a Solution Space Transformation Guided Coevolutionary Algorithm (SSTCE) for the Energy-Saving Distributed Heterogeneous Flexible Job Shop Scheduling Problem (ES-DHFJSP), significantly improving scheduling quality and reducing energy consumption.
A Solver-in-the-Loop Framework for Improving LLMs on Answer Set Programming for Logic Puzzle Solving
Timo Pierre Schrader (Bosch Center for AI), Annemarie Friedrich (ScaDS.AI)
OptimizationTransformerLarge Language ModelSupervised Fine-TuningReinforcement LearningPrompt EngineeringTextBenchmark
🎯 What it does: A Solver-in-the-Loop framework was constructed by introducing feedback from ASP solvers during the training and inference stages of large language models (LLMs) when generating Answer Set Programming (ASP) code;
A Stage-Aware Mixture of Experts Framework for Neurodegenerative Disease Progression Modelling
Tiantian He (University College London), Daniel C. Alexander (University College London)
Graph Neural NetworkMixture of ExpertsAuto EncoderTime SeriesBiomedical DataPositron Emission TomographyAlzheimer's Disease
🎯 What it does: Propose a phase-aware hybrid expert framework to model the long-term progression of neurodegenerative diseases.
A Study of Finetuning Video Transformers for Multi-view Geometry Tasks
Huimin Wu (Hong Kong University of Science and Technology), Zhirong Wu (Microsoft Research Asia)
Depth EstimationTransformerSupervised Fine-TuningOptical FlowVideo
🎯 What it does: This paper achieves multi-view geometry tasks (optical flow, stereo matching, depth estimation) by fine-tuning video pre-trained visual Transformers (ViT), without relying on dedicated cost volumes or complex task-specific structures, using a simple framework with linear decoders or iterative refinement;
A Switching Framework for Online Interval Scheduling with Predictions
Antonios Antoniadis (University of Twente), Abolfazl Soltani (Max Planck Institute for Informatics, University of Saarland)
OptimizationTime Series
🎯 What it does: This paper proposes a learning-enhanced algorithm for online interval scheduling, designs semi-trust-switching and fully-trust-switching frameworks, and presents a random merge algorithm with smooth decay;
A Text-Routed Sparse Mixture-of-Experts Model with Explanation and Temporal Alignment for Multi-Modal Sentiment Analysis
Dongning Rao (Guangdong University of Technology), Jujian Lv (Guangdong Polytechnic Normal University)
ClassificationExplainability and InterpretabilityLarge Language ModelMixture of ExpertsMultimodality
🎯 What it does: This paper proposes a text routing sparse expert model that combines multi-modal large language models to generate explanations and time alignment for multi-modal sentiment analysis.
A Theoretical Analysis of Detecting Large Model-Generated Time Series
Junji Hou (Xi'an Jiaotong University), Pinghui Wang (Xi'an Jiaotong University)
Anomaly DetectionTime Series
🎯 What it does: Studied how to detect synthetic sequences generated by time series large models (TSLM), proposed a theoretical hypothesis, and implemented a white-box detection method.
A Theory of Adaptive Scaffolding for LLM-Based Pedagogical Agents
Clayton Cohn (Vanderbilt University), Gautam Biswas (Vanderbilt University)
TransformerLarge Language ModelPrompt EngineeringTextChain-of-Thought
🎯 What it does: Develop and evaluate an adaptive teaching agent, Inquizzitor, based on large language models, integrating Evidence-Centered Design, Social Cognitive Theory, and the Zone of Proximal Development framework, to implement immediate formative assessment and personalized feedback in middle school Earth science STEM+C courses.
A Theory-Inspired Framework for Few-Shot Cross-Modal Sketch Person Re-Identification
Yunpeng Gong (Xiamen University), Min Jiang (Xiamen University)
RecognitionRetrievalDomain AdaptationMeta LearningGenerative Adversarial NetworkContrastive LearningImageMultimodality
🎯 What it does: To address the challenge of matching hand-drawn sketches with RGB surveillance images, the KTCAA framework is proposed, which achieves transfer from the data-rich RGB domain to the sketch domain with limited samples through local sketch-style enhancement and adversarial perturbations of RGB images under a meta-learning framework.
A Topological Rewriting of Tarski’s Mereogeometry
Richard Dapoigny (University Savoie Mont-Blanc)
🎯 What it does: This paper extends the λ-MM library, constructing topological structures based on regular open sets within the Coq proof assistant, and reformalizes Tarski's solid geometry within this framework, proving that it satisfies Hausdorff and other topological axioms;
A TSP-Based Algorithm for Multi-League Traveling Tournament
Jingyang Zhao (University of Electronic Science and Technology of China), Ken-Ichi Kawarabayashi
Optimization
🎯 What it does: This paper studies the p-partite Traveling Tournament Problem (p-partite TTP), proposes an efficient algorithm based on the Traveling Salesman Problem (TSP), and proves that the problem is NP-hard when p≥3.
A Unified Convergence Analysis for Semi-Decentralized Learning: Sampled-to-Sampled vs. Sampled-to-All Communication
Angelo Rodio (Linkping University), Erik G. Larsson (Centre Inria d'Universit' e Cˆte d'Azur)
Federated LearningConvolutional Neural NetworkImage
🎯 What it does: This paper conducts a unified convergence analysis of two server-to-device communication primitives in semi-decentralized federated learning—Sampled-to-Sampled (S2S) and Sampled-to-All (S2A)—and provides both theoretical and experimental comparisons;
A Unified Self-Regulating Training Framework for Federated Deep Reinforcement Learning
Meng Xu, Jianping Wang (Southeast University)
Federated LearningReinforcement Learning
🎯 What it does: This paper proposes a unified self-adjusting training framework that combines sparse training with auxiliary models, enabling efficient training of federated deep reinforcement learning in dynamic state transition environments.
A Unified Shape-Aware Foundation Model for Time Series Classification
Zhen Liu (South China University of Technology), Qianli Ma (South China University of Technology)
ClassificationExplainability and InterpretabilityRepresentation LearningConvolutional Neural NetworkTransformerContrastive LearningTime Series
🎯 What it does: Proposes UniShape, a unified shape-aware foundation model for time series classification, which enhances cross-domain generalization capability through pre-training + transfer learning.
A-FloPS: Accelerating Diffusion Models via Adaptive Flow Path Sampler
Cheng Jin (Tsinghua University), Yuantao Gu (Tsinghua University)
GenerationComputational EfficiencyDiffusion modelFlow-based ModelImageTextOrdinary Differential Equation
🎯 What it does: Propose the A-FloPS framework, which accelerates the sampling process of diffusion models through training-free flow-matching reparameterization and adaptive velocity decomposition.
A²Flow: Automating Agentic Workflow Generation via Self-Adaptive Abstraction Operators
Mingming Zhao, Zhitang Chen (Huawei Noah's Ark Lab)
Large Language ModelAgentic AITextBenchmarkChain-of-Thought
🎯 What it does: Designed and implemented the A²Flow framework, which automatically generates agentic workflows from expert examples without requiring manual definition of operators.
A²LC: Active and Automated Label Correction for Semantic Segmentation
Youjin Jeon (Yonsei University), Euntai Kim (Yonsei University)
SegmentationTransformerVision Language ModelImage
🎯 What it does: Proposed a two-stage active and automatic label correction framework ALC2 for semantic segmentation.
A3D: Adaptive Affordance Assembly with Dual-Arm Manipulation
Jiaqi Liang (Peking University), Ruihai Wu (Peking University)
Robotic IntelligenceConvolutional Neural NetworkReinforcement LearningAuto EncoderPoint CloudBenchmark
🎯 What it does: Developed the A3D framework, which learns adaptive support and stability points through dual-arm coordination to complete furniture assembly tasks.
AbductiveMLLM: Boosting Visual Abductive Reasoning Within MLLMs
Boyu Chang, Tianfei Zhou (Beijing Institute of Technology)
GenerationTransformerLarge Language ModelPrompt EngineeringVision Language ModelDiffusion modelContrastive LearningVideoTextMultimodality
🎯 What it does: Proposed a multimodal large language model called AbductiveMLLM that integrates language and visual imagination to enhance visual abductive reasoning (VAR) capabilities.
Accelerating Controllable Generation via Hybrid-grained Cache
Lin Liu (University Of Science And Technology Of China), Yanbin Hao
GenerationComputational EfficiencyImageVideo
🎯 What it does: Introduce a mixed-grained caching technique in controllable generation models, utilizing coarse-grained block-level caching and fine-grained prompt-level caching to accelerate the inference process
Accelerating LLM Inference Throughput via Asynchronous KV Cache Prefetching
Yanhao Dong (Alibaba Cloud), Feng Lyu (Central South University)
Computational EfficiencyTransformerText
🎯 What it does: Propose an asynchronous KV Cache prefetch method based on GPU L2 cache, significantly improving LLM inference throughput
AccKV: Towards Efficient Audio-Video LLMs Inference via Adaptive-Focusing and Cross-Calibration KV Cache Optimization
Zhonghua Jiang (Zhejiang University), Shengyu Zhang (Alibaba Group)
OptimizationComputational EfficiencyTransformerLarge Language ModelVideoMultimodalityAudio
🎯 What it does: Designed and implemented a KV cache optimization framework named AccKV to enhance the computational and storage efficiency of audio-visual large language model (AV-LLM) inference.
Accommodate Knowledge Conflicts in Retrieval-augmented LLMs: Towards Robust Response Generation in the Wild
Jiatai Wang (Nankai University), Tao Li (Eigen AI)
RetrievalTransformerTextRetrieval-Augmented Generation
🎯 What it does: Proposed the Swin-VIB framework, which uses variational information bottleneck and sliding windows to adaptively filter retrieval contexts, alleviating knowledge conflicts in retrieval-enhanced LLMs and improving response reliability.
Achieving Equilibrium Under Utility Heterogeneity: An Agent-Attention Framework for Multi-Agent Multi-Objective Reinforcement Learning
Zhuhui Li (University of Exeter), Geyong Min (University of Exeter)
TransformerReinforcement LearningBenchmark
🎯 What it does: Propose a multi-objective multi-agent reinforcement learning framework based on agent attention to achieve Bayesian Nash Equilibrium in the presence of utility heterogeneity.
ACID-Style: An Adaptive Condition Injection Diffusion Model for Arbitrary Style Transfer
Ting Yang (Central South University), Hui Fang (Sony (China) Limited)
Image TranslationGenerationDiffusion modelImage
🎯 What it does: Propose the ACID-Style framework, which employs a lightweight content and style injection adapter to achieve efficient arbitrary style transfer on Stable Diffusion.
AcoustoReinforce: Multi-Particle Acoustophoretic Path Planning with Deep Reinforcement Learning
Pengyuan Wei (University College London), Prateek Mittal (University College London)
Reinforcement LearningPhysics Related
🎯 What it does: Propose a decentralized path planning method called AcoustoReinforce based on multi-agent reinforcement learning, specifically designed for multi-particle acoustic suspension control, capable of generating collision-free, acoustically stable, and physically realizable trajectories.
Action-and-object Aware Alignment for Partially Relevant Video Retrieval
Chuanshen Chen (South China University of Technology), Mingkui Tan (South China University of Technology)
RetrievalConvolutional Neural NetworkTransformerVision Language ModelContrastive LearningVideoTextMultimodality
🎯 What it does: Propose a dual-branch framework A3PRVR for retrieving partially relevant video segments from untrimmed videos based on text queries, employing both action and object visual features, and incorporating query-specific deformable temporal attention (Q-DTA) and action/object-aware contrastive loss to achieve fine-grained video-text alignment.
ActiShade: Activating Overshadowed Knowledge to Guide Multi-Hop Reasoning in Large Language Models
Huipeng Ma (Beijing Institute of Technology), Shuhao Zhang (Beijing Institute of Technology)
RetrievalRepresentation LearningLarge Language ModelContrastive LearningTextRetrieval-Augmented Generation
🎯 What it does: Propose the ActiShade framework, which iteratively detects and activates key knowledge masked in multi-hop reasoning to enhance the synergy between retrieval and generation.
Activating Visual Context and Commonsense Reasoning Through Masked Prediction in VLMs
Jiaao Yu (East China Normal University), Man Lan (East China Normal University)
Supervised Fine-TuningReinforcement LearningPrompt EngineeringVision Language ModelImageMultimodalityBenchmark
🎯 What it does: Propose a new fine-tuning task MPCC (Masked Prediction via Context and Commonsense), which obstructs key visual entities in images to force visual language models (VLMs) to integrate visual context and commonsense reasoning, thereby enhancing their cross-modal reasoning ability.
Activation Manipulation Attack: Penetrating and Harmful Jailbreak Attack Against Large Vision-Language Models
Haojie Hao (Beihang University), Xianglong Liu (Beihang University)
Adversarial AttackVision Language ModelImageTextMultimodality
🎯 What it does: Proposed ActMan, an activation manipulation attack framework that achieves stronger penetration and harmful destruction in large-scale vision-language models through fine-grained control of visual and language activations.
Activation-wise Propagation: A One-Timestep Strategy for Spiking Neural Networks
Jian Song, Donglin Wang (Westlake University)
ClassificationRecognitionObject DetectionSegmentationComputational EfficiencyConvolutional Neural NetworkSpiking Neural NetworkTransformerImagePoint Cloud
🎯 What it does: Propose Activation Layer Membrane Potential Propagation (AMP2), a spiking neural network (SNN) framework enabling training and inference within a single clock cycle.
Activations as Features: Probing LLMs for Generalizable Essay Scoring Representations
Jinwei Chi, Qiang Xu (Jinan University)
ClassificationRepresentation LearningTransformerLarge Language ModelPrompt EngineeringContrastive LearningTextBenchmark
🎯 What it does: Studied and verified that intermediate layer activations of large language models can effectively distinguish essay quality through linear probes, achieving significant performance in cross-prompt automatic essay scoring tasks.
Active Learning of Symbolic Automata over Rational Numbers
Sebastian Hagedorn Gaete, Rodrigo Toro Icarte (Pontificia Universidad Catolica De Chile)
Computational EfficiencyTextTime Series
🎯 What it does: This paper proposes an active learning approach based on the L* algorithm, capable of learning symbolic finite automata that use inequality predicates over the rational number domain.
Active Multi-source Domain Adaptation for Multimodal Fake News Detection
Yanping Chen (Soochow University), Jiajie Xu (Zhejiang Normal University)
ClassificationDomain AdaptationMixture of ExpertsContrastive LearningMultimodality
🎯 What it does: Proposed ADOSE, a framework combining multi-source active domain adaptation with multi-modal fake news detection, addressing cross-domain semantic and deceptive pattern differences by actively annotating a small number of target samples and building a multi-expert fusion network.
Active3D: Active High-Fidelity 3D Reconstruction via Multi-Level Uncertainty Quantification
Yan Li (National University Of Singapore), Gim Hee Lee (National University Of Singapore)
GenerationNeural Radiance FieldGaussian SplattingImagePoint Cloud
🎯 What it does: Propose the Active3D framework, which integrates implicit neural fields with explicit 3D Gaussian primitives to achieve active high-precision 3D reconstruction, constructing a hierarchical uncertainty space and planning the next best view based on expected mixed information gain.
Actor-Critic for Continuous Action Chunks: A Reinforcement Learning Framework for Long-Horizon Robotic Manipulation with Sparse Reward
Jiarui Yang (Fudan University), Yu-Gang Jiang (Fudan University)
Robotic IntelligenceRecurrent Neural NetworkReinforcement LearningContrastive LearningSequentialBenchmark
🎯 What it does: Designed a new Actor-Critic framework AC3 specifically for directly learning continuous action blocks to address long-horizon robotic manipulation tasks with sparse rewards
AD-FM: Multimodal LLMs for Anomaly Detection via Multi-Stage Reasoning and Fine-Grained Reward Optimization
Jingyi Liao (Nanyang Technological University), Xulei Yang (Nanyang Technological University)
Anomaly DetectionTransformerSupervised Fine-TuningReinforcement LearningVision Language ModelMultimodalityChain-of-Thought
🎯 What it does: Propose the AD-FM framework, which achieves more accurate visual judgment and localization in industrial anomaly detection tasks for general-purpose multimodal LLMs through multi-stage reasoning and fine-grained reward optimization.
AdaCuRL: Adaptive Curriculum Reinforcement Learning with Invalid Sample Mitigation and Historical Revisiting
Renda Li (Alibaba Group), Xiangxiang Chu (Alibaba Group)
Supervised Fine-TuningReinforcement LearningTextMultimodalityBenchmark
🎯 What it does: Propose the AdaCuRL framework, integrating coarse and fine difficulty estimation with adaptive curriculum reinforcement learning, to address gradient exhaustion and policy degradation in RL training, significantly enhancing reasoning capabilities in LLM/MLLM.
AdaDepth: Exploiting Inherent Scene Information for Self-Supervised Depth Estimation in Dynamic Scenes
Xuanang Gao (Shanghai Jiao Tong University), Wei Liu (Shanghai Jiao Tong University)
Depth EstimationAutonomous DrivingOptical FlowVideo
🎯 What it does: Proposes the AdaDepth framework, which generates more accurate monocular depth maps in dynamic scenes using self-supervised learning.
AdaField: Generalizable Surface Pressure Modeling with Physics-Informed Pre-training and Flow-Conditioned Adaptation
Junhong Zou (MAIS, Institute of Automation, Chinese Academy of Sciences), Xiangyu Zhu (MAIS, Institute of Automation, Chinese Academy of Sciences)
Autonomous DrivingConvolutional Neural NetworkTransformerSupervised Fine-TuningPoint CloudPhysics Related
🎯 What it does: Propose the AdaField framework for predicting pressure fields on vehicle surfaces, achieving efficient transfer from pre-trained public car data to data-scarce train and aircraft scenarios.
AdaFuse: Accelerating Dynamic Adapter Inference via Token-Level Pre-Gating and Fused Kernel Optimization
Qiyang Li (Baidu Inc.), Dawei Yin (Shanghai Jiao Tong University)
Computational EfficiencyTransformerLarge Language ModelMixture of ExpertsTextBenchmark
🎯 What it does: Introduce the AdaFuse framework on large language models, utilizing token-level pre-gating and fusion kernels to accelerate dynamic adapter inference.
AdaMCoT: Rethinking Cross-Lingual Factual Reasoning Through Adaptive Multilingual Chain-of-Thought
Zheng Weihua (Institute for Infocomm Research, A*STAR), Roy Ka-Wei Lee (Singapore University of Technology and Design)
Explainability and InterpretabilityTransformerLarge Language ModelSupervised Fine-TuningTextChain-of-Thought
🎯 What it does: Propose the AdaMCOT framework, which leverages Adaptive Multilingual Chain-of-Thought to enhance cross-lingual fact reasoning performance;
Adapt Before Continual Learning
Aojun Lu (Sichuan University), Yanan Sun (Sichuan University)
Computational EfficiencyRepresentation LearningTransformerSupervised Fine-TuningContrastive LearningImage
🎯 What it does: This paper proposes a framework called ACL that adaptively fine-tunes the pre-trained model (PTM) before each incremental task to enhance plasticity in continual learning while maintaining stability.
Adapt-As-You-Walk Through the Clouds: Training-Free Online Test-Time Adaptation of 3D Vision-Language Foundation Models
Mehran Tamjidi (University of Technology Sydney), Morteza Saberi (Macquarie University)
Domain AdaptationVision Language ModelPoint Cloud
🎯 What it does: Proposed a training-agnostic online testing-time adaptation framework called Uni-Adapter, specifically designed for 3D vision-language foundation models (VLFMs);
ADAPT: Adaptive Decentralized Architecture with Perception-Aligned Training for Structural Generalization in Multi-Agent RL
Zhixiang Zhang (Wuhan University), Feng Wang (BIGAI)
Recurrent Neural NetworkTransformerReinforcement LearningSequential
🎯 What it does: Propose the ADAPT framework to achieve structural generalization of multi-agent reinforcement learning in DTDE environments, supporting observation and execution for any number of objects;
AdaptCLIP: Adapting CLIP for Universal Visual Anomaly Detection
Bin-Bin Gao (Tencent Youtu Lab), Chengjie Wang (Tencent Youtu Lab)
Anomaly DetectionTransformerPrompt EngineeringContrastive LearningImage
🎯 What it does: Propose the AdaptCLIP framework, which adds three simple adapters to CLIP, achieving zero/few-shot visual anomaly detection and segmentation in cross-domain scenarios without requiring target domain fine-tuning.