AAAI 2024 Papers with AI Summaries
AAAI Conference on Artificial Intelligence · 2331 papers
→ AAAI 2024 papers with code (1014)
Each paper below shows an AI-generated one-line summary. Get the full 6-part summary (innovation, method, data, results, limitations) and search all 2331 AAAI 2024 papers by keyword, author or institution —
free trial on arXivSub.
‘Why Didn’t You Allocate This Task to Them?’ Negotiation-Aware Task Allocation and Contrastive Explanation Generation
Zahra Zahedi (Arizona State University), Subbarao Kambhampati (Arizona State University)
OptimizationExplainability and InterpretabilityAgentic AIText
🎯 What it does: This study proposes an Artificial Intelligence Task Allocator (AITA) that can generate negotiation-based, interpretable task allocations and provide comparative explanations through a negotiation tree when allocations are questioned by humans.
1/2-Approximate MMS Allocation for Separable Piecewise Linear Concave Valuations
Chandra Chekuri (University of Illinois), Ruta Mehta (University of Illinois)
Optimization
🎯 What it does: This paper studies the fair allocation of indivisible goods under separable piecewise linear concave (SPLC) preferences, proposing a polynomial-time solvable 1/2-MMS approximate allocation and providing a 1/3-APS greedy algorithm under submodular preferences.
3D Visibility-Aware Generalizable Neural Radiance Fields for Interacting Hands
Xuan Huang (Shenzhen Campus of Sun Yat-sen University), Xiaodan Liang (Tencent)
GenerationData SynthesisPose EstimationNeural Radiance FieldGenerative Adversarial NetworkImage
🎯 What it does: A visible perception NeRF model that can generalize from a single image is proposed for generating high-quality images of interactive two-hand scenes.
3D-STMN: Dependency-Driven Superpoint-Text Matching Network for End-to-End 3D Referring Expression Segmentation
Changli Wu (Xiamen University), Xiaoshuai Sun (Xiamen University)
Object DetectionSegmentationGraph Neural NetworkTextPoint Cloud
🎯 What it does: An end-to-end 3D Referring Expression Segmentation framework, 3D-STMN, is proposed, which directly performs semantic matching at the superpoint level, avoiding error propagation and redundant computation in a two-stage process.
A Brain-Inspired Way of Reducing the Network Complexity via Concept-Regularized Coding for Emotion Recognition
Han Lu (Fudan University), Qiang Luo (Fudan University)
RecognitionExplainability and InterpretabilityComputational EfficiencyConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: This paper proposes a brain-inspired emotional recognition framework that employs a dual pathway (conceptual pathway and perceptual pathway) to reduce network complexity and enhance interpretability through emotional concept regularization.
A Bregman Proximal Stochastic Gradient Method with Extrapolation for Nonconvex Nonsmooth Problems
Qingsong Wang (Xiangtan University), Deren Han (Beihang University)
OptimizationImage
🎯 What it does: A Bregman Approximate Stochastic Gradient Method (BPSGE) is proposed, which addresses non-convex and non-smooth optimization problems without global Lipschitz continuous gradients by incorporating extrapolation techniques, and provides convergence and complexity analysis from subsequences to full sequences.
A Class of Topological Pseudodistances for Fast Comparison of Persistence Diagrams
Rolando Kindelan Nuñez (Universidad de Chile), Nancy Hitschfeld (Universidad de Chile)
ClassificationComputational EfficiencyAuto EncoderImage
🎯 What it does: This paper proposes a class of adjustable complexity topological pseudodistance—Extended Topological Pseudodistance (ETD)—for quickly comparing persistence diagrams (PD) while maintaining stability guarantees.
A Closer Look at Curriculum Adversarial Training: From an Online Perspective
Lianghe Shi (Wuhan University), Weiwei Liu (Wuhan University)
OptimizationAdversarial AttackConvolutional Neural NetworkImageTime Series
🎯 What it does: This paper provides an in-depth analysis of Curriculum Adversarial Training (CAT) from the perspective of online learning, deriving a generalization error upper bound that includes intermediate iterations, and using this theory to explain the advantages of CAT in enhancing model robustness and alleviating the accuracy-robustness trade-off.
A Compiler for Weak Decomposable Negation Normal Form
Petr Illner (Charles University), Petr Kučera (Charles University)
🎯 What it does: This paper introduces weakly decomposable negation normal form (wDNNF) circuits into the knowledge compilation map and proves that they share the same properties as DNNF circuits in terms of queries and transformations, while being more concise than DNNF; it then proposes and implements a compiler named Bella that can compile CNF formulas into wDNNF circuits.
A Comprehensive Analysis of the Effectiveness of Large Language Models as Automatic Dialogue Evaluators
Chen Zhang (National University of Singapore), Haizhou Li (Chinese University of Hong Kong)
TransformerLarge Language ModelText
🎯 What it does: Conducted multi-dimensional automatic dialogue evaluation of 30 large language models on 12 meta-evaluation datasets.
A Comprehensive Augmentation Framework for Anomaly Detection
Jiang Lin (Southeast University), Yaping Yan (Southeast University)
Anomaly DetectionAuto EncoderImage
🎯 What it does: A comprehensive anomaly simulation framework and segmentation training strategy is proposed to enhance the performance of reconstruction-based anomaly detection.
A Computation-Aware Shape Loss Function for Point Cloud Completion
Shunran Zhang (University of Macau), Leong Hou U (Shenzhen University)
OptimizationComputational EfficiencyPoint Cloud
🎯 What it does: An adaptive auction algorithm based on initial prices is proposed, significantly accelerating and improving the computational efficiency and accuracy of the Earth Mover's Distance (EMD) loss function in point cloud completion tasks.
A Convolutional Neural Network Interpretable Framework for Human Ventral Visual Pathway Representation
Mufan Xue (Beijing Institute of Technology), Guoyuan Yang (Beijing Institute of Technology)
Explainability and InterpretabilityRepresentation LearningConvolutional Neural NetworkImageMagnetic Resonance Imaging
🎯 What it does: This study designs and validates an interpretable convolutional neural network framework (CNNIF) that encodes voxel responses of the human dorsal visual pathway by combining feature-weighted receptive fields (FWRF) with spatial pooling fields, and reveals the hierarchical representation of the concept of 'human' across layers such as V1–V2–V3–hV4–FFA through network dissection.
A Cross-View Hierarchical Graph Learning Hypernetwork for Skill Demand-Supply Joint Prediction
Wenshuo Chao (Hong Kong University of Science and Technology), Hao Liu (Hong Kong University of Science and Technology)
Graph Neural NetworkGraphTime SeriesFinance Related
🎯 What it does: This paper proposes a Cross-View Hierarchical Graph Learning Hypernetwork (CHGH) for jointly predicting changes in the demand and supply of skills in the labor market. The model encodes the graph structures of demand and supply, clusters hierarchical relationships, and constructs a conditional hyper-decoder using historical supply-demand gaps to achieve accurate predictions of future skill demand and supply.
A Diffusion Model with State Estimation for Degradation-Blind Inverse Imaging
Liya Ji (Hong Kong University of Science and Technology), Qifeng Chen (Hong Kong University of Science and Technology)
RestorationGenerationDiffusion modelImage
🎯 What it does: A diffusion model based on a learnable state estimator is proposed to achieve inverse image reconstruction under unknown degradation operators.
A Diffusion-Based Framework for Multi-Class Anomaly Detection
Haoyang He (Zhejiang University), Lei Xie (Zhejiang University)
Anomaly DetectionDiffusion modelAuto EncoderImage
🎯 What it does: A multi-class anomaly detection framework DiAD based on diffusion models is proposed, integrating pixel space autoencoders, a semantic guidance SG network, and a spatially aware feature fusion SFF block to achieve precise reconstruction and localization of anomalous regions.
A Diffusion-Based Pre-training Framework for Crystal Property Prediction
Zixing Song (Chinese University of Hong Kong), Irwin King (Chinese University of Hong Kong)
Graph Neural NetworkDiffusion modelScore-based ModelGraphPhysics Related
🎯 What it does: Through the self-supervised pre-training-fine-tuning framework CrysDiff, a diffusion model is utilized for the joint reconstruction of crystal structures, thereby improving the performance of crystal property predictions.
A Dual Stealthy Backdoor: From Both Spatial and Frequency Perspectives
Yudong Gao (China University of Petroleum), Weifeng Liu (China University of Petroleum)
Adversarial AttackImage
🎯 What it does: A dual hidden backdoor attack method (DUBA) that is concealed in both spatial and frequency domains is designed, utilizing wavelet transform, mixed smoothing with FFT and DCT, and a weak trigger training + strong attack strategy to achieve a high success rate and low detectability.
A Dual-Way Enhanced Framework from Text Matching Point of View for Multimodal Entity Linking
Shezheng Song (Hefei University of Technology), Meng Wang (National University of Defense Technology)
RetrievalRepresentation LearningTransformerContrastive LearningImageTextMultimodality
🎯 What it does: Proposed and implemented a Dual-Way Enhanced (DWE) framework, treating multimodal information (text and images) as a neural text matching task to improve the entity linking problem.
A Dynamic GCN with Cross-Representation Distillation for Event-Based Learning
Yongjian Deng (Beijing University of Technology), Youfu Li (City University of Hong Kong)
ClassificationRecognitionObject DetectionKnowledge DistillationGraph Neural NetworkContrastive LearningImageVideo
🎯 What it does: A dynamic graph convolutional network EDGCN based on event cameras is designed, and a cross-representation distillation framework CRD is proposed to enhance the performance of point-based event models.
A Dynamic Learning Method towards Realistic Compositional Zero-Shot Learning
Xiaoming Hu (University of Science and Technology of China), Zilei Wang (University of Science and Technology of China)
ClassificationDomain AdaptationConvolutional Neural NetworkTransformerGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes a more realistic task of RCZSL (Realistic Compositional Zero-Shot Learning) and constructs an RCZSL dataset based on MIT-States, utilizing GAN for multi-domain transformation; it introduces a dynamic learning framework that dynamically adapts to unknown attributes, objects, and image styles through visual and semantic modulators, addressing the issue of domain shift.
A Fast Exact Solver with Theoretical Analysis for the Maximum Edge-Weighted Clique Problem
Lu Liu (University of Electronic Science and Technology of China), Yi Zhou (University of Electronic Science and Technology of China)
OptimizationGraphBenchmark
🎯 What it does: This paper systematically studies the Maximum Edge Weight Clique Problem (MEWCP). It first proves that the problem remains NP-hard even when the minimum degree of the graph is $n-2$. Then, it proposes a branch-and-bound algorithm called MEWCat based on a new, tighter upper bound, and provides a theoretical time complexity upper bound of $O^*(1.4423^n)$. Finally, experiments are conducted on various benchmarks.
A Fixed-Parameter Tractable Algorithm for Counting Markov Equivalence Classes with the Same Skeleton
Vidya Sagar Sharma (Tata Institute of Fundamental Research)
Graph
🎯 What it does: This paper proposes a fixed-parameter tractable algorithm based on the tree width and maximum degree of a graph to count the number of Markov equivalence classes (MECs) with the same skeleton.
A Fixed-Point Approach to Unified Prompt-Based Counting
Wei Lin (City University of Hong Kong), Antoni B. Chan (City University of Hong Kong)
Object DetectionDomain AdaptationConvolutional Neural NetworkPrompt EngineeringContrastive LearningImageMultimodality
🎯 What it does: A unified prompt-based category-free counting framework is proposed, which can accept three types of prompts: boxes, points, and text, and output target density maps.
A General Implicit Framework for Fast NeRF Composition and Rendering
Xinyu Gao (Zhejiang University), Changqing Zou (Zhejiang University)
GenerationDepth EstimationComputational EfficiencyNeural Radiance FieldPoint Cloud
🎯 What it does: A general implicit framework is proposed, utilizing Neural Depth Fields (NeDF) to achieve fast synthesis of NeRF objects, real-time shadow casting, and interactive previews of multi-object scenes.
A General Search-Based Framework for Generating Textual Counterfactual Explanations
Daniel Gilo (Technion Israel Institute of Technology), Shaul Markovitch (Technion Israel Institute of Technology)
GenerationExplainability and InterpretabilityLarge Language ModelSupervised Fine-TuningText
🎯 What it does: A general search-based framework TCE-SEARCH is proposed to generate interpretable counterfactual texts for text classifiers.
A General Theoretical Framework for Learning Smallest Interpretable Models
Sebastian Ordyniak (University of Leeds), Stefan Szeider (TU Wien)
Explainability and InterpretabilityComputational Efficiency
🎯 What it does: This paper proposes a general framework for efficiently learning and enumerating the smallest interpretable models (decision trees, decision sets, decision lists, binary decision diagrams, and their ensembles) under the parameter s+δ.
A Generalized Neural Diffusion Framework on Graphs
Yibo Li (Beijing University of Posts and Telecommunications), Chuan Shi (Beijing University of Posts and Telecommunications)
Graph Neural NetworkDiffusion modelGraph
🎯 What it does: A general diffusion equation framework is proposed, and based on this framework, HiD-Net is designed to improve the expressive power of graph neural networks by utilizing the unimodal characteristics of 2-hop neighbors.
A Generalized Shuffle Framework for Privacy Amplification: Strengthening Privacy Guarantees and Enhancing Utility
E Chen (Zhejiang Lab), Yifei Ge (Xi'an Jiaotong-Liverpool University)
Safty and PrivacyImageTabular
🎯 What it does: A general Shuffle framework GSPA is proposed, supporting arbitrary (ε,δ) personalized local differential privacy randomizers, and a global privacy amplification analysis after shuffling is provided.
A Goal Interaction Graph Planning Framework for Conversational Recommendation
Xiaotong Zhang (Dalian University of Technology), Xianchao Zhang (Dalian University of Technology)
Recommendation SystemGraph Neural NetworkTransformerLarge Language ModelTextGraph
🎯 What it does: A target interaction graph planning framework is proposed for multi-target conversation recommendation, which can dynamically plan the target sequence in the dialogue and guide users to achieve the final recommendation.
A Graph Dynamics Prior for Relational Inference
Liming Pan (University of Science and Technology of China), Ivan Dokmanic (Universitat Basel)
Graph Neural NetworkGraphTime Series
🎯 What it does: A Graph Dynamics Prior (GDP) framework is proposed, utilizing multi-order polynomial graph filtering and shallow single-step GNN to reconstruct interaction graphs from observed dynamics by sharing graph structures in parallel.
A Hierarchical Network for Multimodal Document-Level Relation Extraction
Lingxing Kong (Nanjing University), Jiajun Chen (Nanjing University)
Convolutional Neural NetworkTransformerVideoTextMultimodality
🎯 What it does: This paper proposes a multi-modal document-level relation extraction (MDocRE) task, utilizing both text and corresponding video modalities to enhance the handling of long-distance dependencies and multi-mention selection issues.
A Hybrid Global-Local Perception Network for Lane Detection
Qing Chang (Nanjing University of Science and Technology), Yifei Tong (Nanjing University of Science and Technology)
Autonomous DrivingTransformerImage
🎯 What it does: A hybrid global-local perception network (HGLNet) is proposed and implemented for lane detection, achieving high-precision lane localization by parallel extraction of global context and local features.
A Joint Framework with Heterogeneous-Relation-Aware Graph and Multi-Channel Label Enhancing Strategy for Event Causality Extraction
Ruili Pu (Shanxi University), Jianxing Zheng (Shanxi University)
ClassificationGraph Neural NetworkLarge Language ModelTextGraph
🎯 What it does: Proposes a joint multi-label extraction framework that combines heterogeneous relationship-aware graphs and a multi-channel label enhancement strategy to achieve event causality extraction tasks.
A Label Disambiguation-Based Multimodal Massive Multiple Instance Learning Approach for Immune Repertoire Classification
Fan Xu (ShanghaiTech University), Jianhua Yao (Shanghai Jiao Tong University)
ClassificationTransformerSupervised Fine-TuningMultimodalityBiomedical Data
🎯 What it does: A multi-modal large-scale multi-instance learning model based on label disambiguation (LaDM³IL) is proposed for immune receptor library classification and related receptor identification.
A Learnable Discrete-Prior Fusion Autoencoder with Contrastive Learning for Tabular Data Synthesis
Rongchao Zhang (Peking University), Yu Huang (Peking University)
GenerationData SynthesisTransformerAuto EncoderContrastive LearningTabular
🎯 What it does: A learnable discrete prior fusion autoencoder named GTCoder is proposed, which utilizes Transformer to fuse multimodal tabular data and optimizes the latent representation of discrete features through contrastive learning to generate high-quality synthetic data.
A Local-Ascending-Global Learning Strategy for Brain-Computer Interface
Dongrui Gao (Chengdu University of Information Technology), Yongqing Zhang (University of Electronic Science and Technology of China)
RecognitionRecurrent Neural NetworkGraph Neural NetworkTime SeriesBiomedical DataElectrocardiogram
🎯 What it does: This study investigates a Local-Ascending-Global Learning strategy (LAG), which dynamically captures strong connectivity patterns between brain regions under different cognitive tasks through a K-level Adaptive Ascending Network (SALK), achieving learning and fusion of multi-scale functional connectivity.
A Multi-Modal Contrastive Diffusion Model for Therapeutic Peptide Generation
Yongkang Wang, Wen Zhang (Huazhong Agricultural University)
GenerationDrug DiscoveryTransformerDiffusion modelContrastive LearningMultimodalityBiomedical Data
🎯 What it does: A multi-modal contrastive diffusion model (MMCD) is proposed for simultaneously generating sequences and structures of therapeutic peptides.
A Multimodal, Multi-Task Adapting Framework for Video Action Recognition
Mengmeng Wang (Zhejiang University), Yong Liu (Zhejiang University)
RecognitionTransformerSupervised Fine-TuningContrastive LearningVideoMultimodality
🎯 What it does: The M-CLIP framework is proposed, which implements a complete method for transferring CLIP to video action recognition tasks by adding multimodal adapters and a multitask decoder to the visual and text branches of CLIP.
A New Benchmark and Model for Challenging Image Manipulation Detection
Zhenfei Zhang (University at Albany), Ming-Ching Chang (McGill University)
Anomaly DetectionConvolutional Neural NetworkImageBenchmark
🎯 What it does: A new dual-branch network is proposed, capable of detecting image tampering under difficult conditions such as small-scale tampering and double compression with the same quality factor.
A New Mechanism for Eliminating Implicit Conflict in Graph Contrastive Learning
Dongxiao He (Tianjin University), Zhiyong Feng (Tianjin University)
ClassificationRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: This paper proposes the Embedding-Ignoring Conflict (EIC) in information contrastive learning and designs the PiGCL method to dynamically capture and ignore certain negative samples to enhance graph contrastive learning performance.
A Non-parametric Graph Clustering Framework for Multi-View Data
Shengju Yu (National University of Defense Technology), En Zhu (National University of Defense Technology)
OptimizationComputational EfficiencyGraph Neural NetworkGraph
🎯 What it does: A parameter-free multi-view graph clustering framework (NpGC) is proposed, which introduces two types of anchors: view-related and view-independent. It directly constructs a consensus bidirectional graph on the anchor layer, thereby capturing the exclusive and common features of each view, improving clustering quality and achieving linear complexity.
A Novel Energy Based Model Mechanism for Multi-Modal Aspect-Based Sentiment Analysis
Tianshuo Peng (Wuhan University), Hai Zhao (Shanghai Jiao Tong University)
ClassificationRecognitionData-Centric LearningTransformerLarge Language ModelPrompt EngineeringImageTextMultimodality
🎯 What it does: A unified multimodal fine-grained sentiment analysis framework DQPSA is proposed to address the interrelationship issues of visual information attention differences, modality gaps, and span boundaries in multimodal sentiment analysis.
A Novel Skip Orthogonal List for Dynamic Optimal Transport Problem
Xiaoyang Xu (University of Science and Technology of China), Hu Ding (University of Science and Technology of China)
OptimizationTabular
🎯 What it does: A dynamic optimal transport algorithm is proposed, which can quickly update the optimal transport plan after changes in the weights or positions of data points;
A Perspective of Q-value Estimation on Offline-to-Online Reinforcement Learning
Yinmin Zhang (University of Sydney), Wanli Ouyang (Shanghai Artificial Intelligence Laboratory)
Reinforcement LearningTabular
🎯 What it does: This paper addresses the issue of Q-value estimation distortion in offline-to-online reinforcement learning (O2O RL) by proposing the SO2 method, which enhances Q-value accuracy through Perturbed Value Update by adding noise to target actions and smoothing the Q-network updates (increasing the frequency of Q-value updates), thereby accelerating online fine-tuning.
A Plug-and-Play Quaternion Message-Passing Module for Molecular Conformation Representation
Angxiao Yue (Renmin University of China), Hongteng Xu (Renmin University of China)
Drug DiscoveryGraph Neural NetworkGraph
🎯 What it does: A pluggable quaternion information transmission module (QMP) is proposed to enhance the representation of molecular conformations in 3D molecular graph neural networks.
A Positive-Unlabeled Metric Learning Framework for Document-Level Relation Extraction with Incomplete Labeling
Ye Wang (East China Normal University), Wenxin Hu (East China Normal University)
TransformerTextBiomedical Data
🎯 What it does: To address the issue of incomplete annotations in document-level relation extraction, a positive-negative sample unlabeled metric learning framework (P³M) is proposed. This framework enhances the model's generalization and robustness by applying dropout augmentation on positive samples and using unlabeled relations as pseudo-negative samples for mixup.
A Pre-convolved Representation for Plug-and-Play Neural Illumination Fields
Yiyu Zhuang (Nanjing University), Xun Cao (Nanjing University)
RestorationGenerationNeural Radiance FieldImage
🎯 What it does: Proposes the NeIF framework, which uses a differentiable lighting field to reconstruct environmental lighting and achieve physical rendering of objects.
A Primal-Dual Algorithm for Hybrid Federated Learning
Tom Overman (Northwestern University), Diego Klabjan (Northwestern University)
OptimizationFederated LearningSafty and PrivacyTabular
🎯 What it does: This paper proposes HyFDCA - a primal-dual coordinate ascent algorithm for mixed federated learning;
A Provably Accurate Randomized Sampling Algorithm for Logistic Regression
Agniva Chowdhury (Oak Ridge National Laboratory), Pradeep Ramuhalli (Oak Ridge National Laboratory)
ClassificationOptimizationComputational EfficiencyTabularFinance Related
🎯 What it does: A randomized sampling-based logistic regression algorithm is proposed, which obtains an approximate maximum likelihood estimate by randomly sampling and rescaling the observed values, thus quickly solving large-scale logistic regression problems.
A Reinforcement-Learning-Based Multiple-Column Selection Strategy for Column Generation
Haofeng Yuan (Tsinghua University), Shiji Song (Tsinghua University)
OptimizationGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: This paper proposes a multi-column selection strategy based on reinforcement learning to improve the convergence speed of the column generation algorithm.
A Robust Mutual-Reinforcing Framework for 3D Multi-Modal Medical Image Fusion Based on Visual-Semantic Consistency
Hao Zhang (Wuhan University), Jiayi Ma (Wuhan Institute of Technology)
Image TranslationSegmentationConvolutional Neural NetworkTransformerAuto EncoderImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A robust 3D multimodal medical image fusion framework is proposed, utilizing visual-semantic consistency to achieve complementary enhancement of visual fusion and lesion segmentation, significantly improving the accuracy and robustness against degradation of both tasks.
A Score-Based Deterministic Diffusion Algorithm with Smooth Scores for General Distributions
Karthik Elamvazhuthi (University of California, Riverside), Fabio Pasqualetti (University of California, Santa Barbara)
GenerationData SynthesisOptimizationDiffusion modelScore-based ModelImageOrdinary Differential Equation
🎯 What it does: A deterministic diffusion process is proposed for generative models to address the issue of score explosion in traditional stochastic diffusion;
A Separation and Alignment Framework for Black-Box Domain Adaptation
Mingxuan Xia (Zhejiang University), Haobo Wang (Zhejiang University)
Domain AdaptationContrastive LearningImage
🎯 What it does: This paper proposes a new black-box domain adaptation framework called SEAL, which separates target domain samples into two categories: well-adapted and under-adapted. It performs graph contrastive learning and nearest centroid supervision on the latter, achieving adaptive target domain classifiers without accessing source data.
A Sequentially Fair Mechanism for Multiple Sensitive Attributes
Francois Hu (University of Montreal), Arthur Charpentier (University of Quebec at Montreal)
Tabular
🎯 What it does: This paper proposes a sequential fairness mechanism that utilizes multiple Wasserstein barycenters for post-processing predictions with multiple sensitive attributes to achieve group fairness.
A Surprisingly Simple Continuous-Action POMDP Solver: Lazy Cross-Entropy Search Over Policy Trees
Marcus Hoerger (University of Queensland), Nan Ye (Australian National University)
OptimizationRobotic IntelligenceReinforcement Learning from Human FeedbackReinforcement LearningAgentic AITabular
🎯 What it does: An online POMDP solver LCEOPT is proposed, which uses lazy cross-entropy search to find near-optimal continuous action policies in the policy tree space.
A Theory of Non-acyclic Generative Flow Networks
Leo Brunswic (Huawei Shanghai Research Center), Lizhuang Ma (Shanghai Jiaotong University)
GenerationOptimizationReinforcement Learning from Human FeedbackFlow-based ModelGraph
🎯 What it does: This paper extends Generative Flow Networks (GFlowNets) from the original directed acyclic graph to measurable spaces that include cycles and continuous state spaces, and introduces the concept of stability along with corresponding stability loss and regularization to avoid infinite sampling time caused by getting stuck in cycles during training.
A Transfer Approach Using Graph Neural Networks in Deep Reinforcement Learning
Tianpei Yang (Tianjin University), Matthew E. Taylor (University of Alberta)
Robotic IntelligenceGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: The TURRET method is proposed to address the issue of mismatched robot state-action spaces in cross-domain multi-source transfer learning, utilizing graph neural networks to achieve unified state embedding and adaptive weighted multi-source strategies to accelerate target task learning.
A Twist for Graph Classification: Optimizing Causal Information Flow in Graph Neural Networks
Zhe Zhao (University of Science and Technology of China), Yang Wang (University of Science and Technology of China)
ClassificationOptimizationExplainability and InterpretabilityGraph Neural NetworkGraph
🎯 What it does: This paper proposes an ICL framework that combines information theory and causal learning, specifically designed for graph classification tasks. It can decompose graph features into causal and non-causal parts during the training process, maximizing causal information and minimizing non-causal information through multi-objective optimization, thereby enhancing the model's robustness and interpretability against out-of-distribution data.
A Two-Stage Information Extraction Network for Incomplete Multi-View Multi-Label Classification
Xin Tan (Harbin Institute of Technology), Zhanyan Tang (Harbin Institute of Technology)
ClassificationAuto EncoderMultimodality
🎯 What it does: A two-stage information extraction network is proposed to address the problem of incomplete multi-view multi-label classification;
A Unified Environmental Network for Pedestrian Trajectory Prediction
Yuchao Su (Shenzhen University), Xia Li (Shenzhen University)
Convolutional Neural NetworkAuto EncoderTime Series
🎯 What it does: A unified environment network (UEN) based on CNN is proposed, which simultaneously models social interactions and scene constraints using polar coordinate mapping and local multimodal windows to achieve multimodal trajectory prediction.
A Unified Knowledge Transfer Network for Generalized Category Discovery
Wenkai Shi (Xi'an Jiaotong University), Ping Chen (University of Massachusetts Boston)
ClassificationKnowledge DistillationTransformerContrastive LearningText
🎯 What it does: A unified Knowledge Transfer Network (KTN) is proposed, which utilizes labeled known categories and unlabeled mixed data to achieve knowledge transfer through entropy soft distinction and prototype weighting, enhancing the performance of general category discovery.
A Unified Masked Autoencoder with Patchified Skeletons for Motion Synthesis
Esteve Valls Mascaró (Technische Universität Wien), Dongheui Lee (German Aerospace Center)
GenerationData SynthesisPose EstimationTransformerAuto EncoderVideoTime Series
🎯 What it does: This paper proposes a unified pose-blocking method based on a masked autoencoder (UNIMASK-M), which can simultaneously predict, interpolate, and reconstruct human motion.
A Unified Self-Distillation Framework for Multimodal Sentiment Analysis with Uncertain Missing Modalities
Mingcheng Li (Fudan University), Lihua Zhang (Fudan University)
Knowledge DistillationRepresentation LearningTransformerTextMultimodalityAudio
🎯 What it does: A unified self-distillation framework (UMDF) is proposed to address the issue of uncertain missing modalities in multimodal sentiment analysis.
A Unified View on Forgetting and Strong Equivalence Notions in Answer Set Programming
Zeynep G. Saribatur (Institute of Logic and Computation TU Wien), Stefan Woltran (Institute of Logic and Computation TU Wien)
🎯 What it does: A new concept of relativized strong simplification is proposed, unifying various related paradigms in ASP such as forgetting, strong equivalence, strong persistence, and simplification, and providing the corresponding necessary and sufficient conditions and semantic characterization;
A User-Friendly Framework for Generating Model-Preferred Prompts in Text-to-Image Synthesis
Nailei Hei (Fudan University), Wenqiang Zhang (Fudan University)
GenerationData SynthesisTransformerLarge Language ModelPrompt EngineeringDiffusion modelImageText
🎯 What it does: A framework is proposed for automatically converting user-input rough prompts into fine-grained prompts preferred by the model, along with the construction of a corresponding training dataset.
A Variational Autoencoder for Neural Temporal Point Processes with Dynamic Latent Graphs
Sikun Yang (Great Bay University), Hongyuan Zha (Chinese University of Hong Kong)
Recurrent Neural NetworkGraph Neural NetworkAuto EncoderTime SeriesSequential
🎯 What it does: A dynamic structure neural point process based on variational autoencoders is proposed to capture temporal dependency relationships in asynchronous event sequences.
AACP: Aesthetics Assessment of Children’s Paintings Based on Self-Supervised Learning
Shiqi Jiang (East China Normal University), Chenhui Li (East China Normal University)
ClassificationRecognitionConvolutional Neural NetworkSupervised Fine-TuningContrastive LearningImage
🎯 What it does: A dataset for evaluating the aesthetics of children's drawings is constructed, and a multi-attribute evaluation model based on self-supervised learning is proposed, which can automatically provide scores for eight aesthetic attributes of children's drawings.
Abstract Action Scheduling for Optimal Temporal Planning via OMT
Stefan Panjkovic (Fondazione Bruno Kessler), Andrea Micheli (Fondazione Bruno Kessler)
OptimizationTabular
🎯 What it does: An improved abstract step encoding (φ SCHED) is proposed to address the performance bottleneck caused by the original abstraction being too loose by incorporating abstract scheduling constraints into the optimization model theory (OMT), thus achieving more efficient optimal temporal planning.
Abstract and Explore: A Novel Behavioral Metric with Cyclic Dynamics in Reinforcement Learning
Anjie Zhu (University of Electronic Science and Technology of China), Jie Shao (University of Electronic Science and Technology of China)
Reinforcement LearningSequential
🎯 What it does: The BCD (Behavioral Metric with Cyclic Dynamics) method is proposed to address the issues of state representation collapse and insufficient dynamics correlation in environments with program-generated, sparse rewards and high noise, thereby improving exploration efficiency.
Abstraction of Situation Calculus Concurrent Game Structures
Yves Lesperance, Shakil M. Khan (University of Regina)
🎯 What it does: This paper proposes an abstract framework for synchronous multi-agent game structures based on situational calculus, and proves that under certain constraints, this abstraction can be used to verify the properties of µATL-FO strategies.
ACAMDA: Improving Data Efficiency in Reinforcement Learning through Guided Counterfactual Data Augmentation
Yuewen Sun (Mohamed bin Zayed University of Artificial Intelligence), Kun Zhang (Carnegie Mellon University)
Domain AdaptationOptimizationReinforcement LearningGenerative Adversarial NetworkTime Series
🎯 What it does: Proposes the ACAMDA framework, which combines causal structure recovery with adversarial inverse generation to generate guiding counterfactual data augmentation to enhance the data efficiency of reinforcement learning in heterogeneous domains.
Accelerate Multi-Agent Reinforcement Learning in Zero-Sum Games with Subgame Curriculum Learning
Jiayu Chen (Tsinghua University), Yi Wu (Shanghai Qi Zhi Institute)
Reinforcement Learning
🎯 What it does: This paper proposes a subgame-based automatic curriculum learning framework (SACL) that helps multiple agents converge to Nash equilibrium faster in zero-sum games.
Accelerating Cutting-Plane Algorithms via Reinforcement Learning Surrogates
Kyle Mana (J.P. Morgan AI Research), Manuela Veloso (J.P. Morgan AI Research)
OptimizationReinforcement LearningTabular
🎯 What it does: A proxy trained using reinforcement learning replaces the traditional Mixed Integer Master Problem (MIMP), accelerating the convergence speed of the cutting-plane algorithm in two types of discrete optimization: stochastic optimization (Benders decomposition) and L0 regularized regression.
Accelerating Text-to-Image Editing via Cache-Enabled Sparse Diffusion Inference
Zihao Yu (Peking University), Bin Cui (Peking University)
Image TranslationGenerationComputational EfficiencyConvolutional Neural NetworkDiffusion modelImageText
🎯 What it does: This paper presents FISEdit, a framework that utilizes caching and sparse inference to accelerate text-to-image fine-tuning, allowing for quick edits only in the affected areas after subtle modifications to the text prompts by the user.
Accelerating the Global Aggregation of Local Explanations
Alon Mor (Technion Israel Institute of Technology), Benny Kimelfeld (Technion Israel Institute of Technology)
Explainability and InterpretabilityComputational EfficiencyText
🎯 What it does: A global aggregation acceleration method for Anchor local explanations is proposed, along with a new probability aggregation function and various runtime optimizations.
ACT: Empowering Decision Transformer with Dynamic Programming via Advantage Conditioning
Chen-Xiao Gao (Nanjing University), Yang Yu (Nanjing University)
TransformerReinforcement LearningSequential
🎯 What it does: This paper proposes the Advantage-Conditioned Transformer (ACT), which combines decision transformers with dynamic programming to generate actions conditioned on the advantage function, addressing the shortcomings of decision transformers in sparse rewards, random environments, and trajectory segments.
Active Learning Guided by Efficient Surrogate Learners
Yunpyo An (Ulsan National Institute of Science and Technology), Kwang In Kim (Pohang University of Science and Technology)
ClassificationOptimizationConvolutional Neural NetworkGaussian SplattingImage
🎯 What it does: An active learning algorithm utilizing Gaussian process surrogate learners is proposed, which can continuously update label information and reduce redundant sampling without retraining the deep network.
ADA-GAD: Anomaly-Denoised Autoencoders for Graph Anomaly Detection
Junwei He (Institute of Computing Technology), Qingming Huang (Institute of Computing Technology)
Anomaly DetectionGraph Neural NetworkAuto EncoderGraph
🎯 What it does: A two-stage graph anomaly detection framework ADA-GAD is proposed: first, an unsupervised 'anomaly denoising' is used to enhance and generate cleaner graphs, pre-training a graph autoencoder; then, the encoder is frozen, and the decoder is retrained to reconstruct on the original graph, utilizing attention to aggregate multi-level embeddings and regularize the anomaly distribution to discern anomalous nodes.
Ada-Retrieval: An Adaptive Multi-Round Retrieval Paradigm for Sequential Recommendations
Lei Li (Renmin University of China), Xing Xie (Microsoft Research Asia)
RetrievalRecommendation SystemRecurrent Neural NetworkReinforcement LearningSequential
🎯 What it does: An adaptive multi-round retrieval paradigm named Ada-Retrieval is proposed for sequence recommendation systems, which better captures potential candidates by iteratively optimizing user representations.
AdaCCD: Adaptive Semantic Contrasts Discovery Based Cross Lingual Adaptation for Code Clone Detection
Yangkai Du (Zhejiang University), Shouling Ji (Zhejiang University)
Domain AdaptationAI Code AssistantContrastive LearningText
🎯 What it does: This paper proposes AdaCCD, a cross-language adaptive code clone detection method that utilizes a pre-trained multilingual programming language model to discover semantically similar/dissimilar pairs in unlabeled data of the target language through adaptive contrastive learning, and improves performance through iterative training.
AdaFormer: Efficient Transformer with Adaptive Token Sparsification for Image Super-resolution
Xiaotong Luo (Xiamen University), Yun Fu (Northeastern University)
RestorationSuper ResolutionTransformerImage
🎯 What it does: This paper studies an efficient Transformer model for large-size images, AdaFormer, which achieves fast super-resolution inference through adaptive token sparsification and early exit.
AdapEdit: Spatio-Temporal Guided Adaptive Editing Algorithm for Text-Based Continuity-Sensitive Image Editing
Zhiyuan Ma (Tsinghua University), Bowen Zhou (Tsinghua University)
Image TranslationGenerationDiffusion modelImageText
🎯 What it does: This paper proposes a text-conditioned image editing algorithm called AdapEdit, based on time-space adaptive guidance, to achieve soft editing tasks (such as fine-grained changes in posture, actions, adjectives, etc.) without the need for additional training or optimization.
AdapterGNN: Parameter-Efficient Fine-Tuning Improves Generalization in GNNs
Shengrui Li (Tsinghua University), Jing Bai (Microsoft Research Asia)
Drug DiscoveryGraph Neural NetworkSupervised Fine-TuningGraphBiomedical Data
🎯 What it does: A parameter-efficient fine-tuning framework designed specifically for graph neural networks, called AdapterGNN, is proposed and evaluated.
Adaptive Anytime Multi-Agent Path Finding Using Bandit-Based Large Neighborhood Search
Thomy Phan (University of Southern California), Sven Koenig (University of Southern California)
OptimizationReinforcement LearningGraphBenchmark
🎯 What it does: An adaptive large neighborhood search framework called BALANCE based on multi-armed bandits is proposed for multi-agent path planning at arbitrary times.
Adaptive Discovering and Merging for Incremental Novel Class Discovery
Guangyao Chen (Peking University), Yonghong Tian (Peking University)
ClassificationKnowledge DistillationRepresentation LearningContrastive LearningImage
🎯 What it does: Proposes the Adaptive Discovering and Merging (ADM) framework, which first decouples representation learning through self-supervised contrastive learning and knowledge distillation in the incremental new class discovery task, and then generates high-quality pseudo-labels using Triple Comparison and Probability Regularization; subsequently, it employs Adaptive Feature Fusion (AFF) and Adaptive Model Merging (AMM) to incrementally integrate new knowledge into the model without parameters, avoiding catastrophic forgetting.
Adaptive Feature Imputation with Latent Graph for Deep Incomplete Multi-View Clustering
Jingyu Pu (University of Electronic Science and Technology of China), Lifang He (Lehigh University)
Graph Neural NetworkAuto EncoderMultimodality
🎯 What it does: A deep incomplete multi-view clustering method AGDIMC based on adaptive feature imputation and latent graphs is proposed.
Adaptive FSS: A Novel Few-Shot Segmentation Framework via Prototype Enhancement
Jing Wang (University of Science and Technology Beijing), Tianxiang Zhang (University of Science and Technology Beijing)
SegmentationTransformerImage
🎯 What it does: Proposes an adapter-based few-shot semantic segmentation framework called Adaptive FSS, which utilizes a Prototype Adaptive Module to enhance prototypes for new categories, achieving rapid adaptation.
Adaptive Graph Learning for Multimodal Conversational Emotion Detection
Geng Tu (Harbin Institute of Technology), Ruifeng Xu (Harbin Institute of Technology)
ClassificationRecognitionRecurrent Neural NetworkGraph Neural NetworkSupervised Fine-TuningTextMultimodalityAudio
🎯 What it does: This paper proposes an Adaptive Interactive Graph Network (AdaIGN) that models multimodal dialogue emotions through a learnable node and edge selection strategy.
Adaptive Hardness Negative Sampling for Collaborative Filtering
Riwei Lai (Harbin Engineering University), Li Chen (Hong Kong Baptist University)
Recommendation SystemTabular
🎯 What it does: A new negative sampling paradigm called Adaptive Hardness Negative Sampling (AHNS) is proposed, aimed at alleviating false positive and false negative issues by adaptively selecting negative samples of varying hardness during the training process.
Adaptive Integration of Partial Label Learning and Negative Learning for Enhanced Noisy Label Learning
Mengmeng Sheng (Nanjing University of Science and Technology), Yazhou Yao (Nanjing University of Science and Technology)
ClassificationContrastive LearningImage
🎯 What it does: The NPN method is proposed, which combines noise label learning through Partial Label Learning (PLL) and Negative Label Learning (NL) to automatically decompose candidate labels and complementary labels, adapting to the strength of noise.
Adaptive Meta-Learning Probabilistic Inference Framework for Long Sequence Prediction
Jianping Zhu (Dalian University of Technology), Fei Wu (Zhejiang University)
OptimizationMeta LearningTransformerAuto EncoderTime SeriesSequentialFinance Related
🎯 What it does: This paper proposes an adaptive meta-learning probabilistic inference framework based on sequence decomposition, called AMPIF, for long sequence prediction.
Adaptive Prompt Routing for Arbitrary Text Style Transfer with Pre-trained Language Models
Qingyi Liu (Sun Yat-sen University), Keze Wang (Datastory)
GenerationTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: An Adaptive Prompt Routing (APR) framework is proposed, which automatically selects the optimal prompt from a set of readable prompts for each input sentence for arbitrary text style conversion, and generates stylized text through LLM.
Adaptive Reactive Synthesis for LTL and LTLf Modulo Theories
Andoni Rodríguez (IMDEA Software Institute), César Sánchez (IMDEA Software Institute)
🎯 What it does: This paper proposes an instant reaction synthesis method based on Boolean abstraction and SMT solvers, which can generate controllers for Linear Temporal Logic model theory (LTL T).
Adaptive Shortcut Debiasing for Online Continual Learning
Doyoung Kim (KAIST), Jae-Gil Lee (KAIST)
ClassificationImage
🎯 What it does: The DropTop framework is proposed to enhance the model's transferability and stability in online continual learning by suppressing shortcut feature bias.
Adaptive Uncertainty-Based Learning for Text-Based Person Retrieval
Shenshen Li (University of Electronic Science and Technology of China), Heng Tao Shen (University of Electronic Science and Technology of China)
RetrievalTransformerContrastive LearningTextMultimodality
🎯 What it does: An Adaptive Uncertainty-based Learning (AUL) framework is proposed from the perspective of uncertainty to address the issues of matching ambiguity and unidirectional cross-modal alignment in text retrieval-based pedestrian retrieval.
Adv-Diffusion: Imperceptible Adversarial Face Identity Attack via Latent Diffusion Model
Decheng Liu (Xidian University), Xinbo Gao (Chongqing University of Posts and Telecommunications)
GenerationAdversarial AttackDiffusion modelImage
🎯 What it does: This paper studies a method for generating imperceptible adversarial facial identity attacks using a latent diffusion model in latent space.
Advancing Spatial Reasoning in Large Language Models: An In-Depth Evaluation and Enhancement Using the StepGame Benchmark
Fangjun Li (University of Leeds), Anthony G. Cohn (University of Leeds)
TransformerLarge Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: This paper addresses the template errors in the StepGame benchmark and conducts an in-depth evaluation and enhancement of the performance of large language models in spatial reasoning tasks.
Advancing Video Synchronization with Fractional Frame Analysis: Introducing a Novel Dataset and Model
Yuxuan Liu (Tsinghua University), Pin Tao (Beijing University of Technology)
RecognitionPose EstimationTransformerVideo
🎯 What it does: This study implements a fractional-frame synchronization method for multi-view videos, proposing the InSynFormer model and releasing the IFID dataset.
Adversarial Attacks on Federated-Learned Adaptive Bitrate Algorithms
Rui-Xiao Zhang (University of Hong Kong), Tianchi Huang (Sony Group Corporation)
Federated LearningAdversarial AttackReinforcement LearningVideo
🎯 What it does: The study conducts targeted attacks on the adaptive bitrate (ABR) algorithm implemented in deep reinforcement learning (DRL) within the federated learning framework, proposing a two-phase attack framework called MAFL to construct malicious models and contaminate the global model through model replacement, significantly reducing QoE on the target client while maintaining performance for other clients.