AAAI 2023 Papers — Page 9
AAAI Conference on Artificial Intelligence · 1578 papers
Learning Polysemantic Spoof Trace: A Multi-Modal Disentanglement Network for Face Anti-spoofing
Kaicheng Li (Beihang University), Di Huang (Beihang University)
RecognitionAnomaly DetectionGenerative Adversarial NetworkImageMultimodality
🎯 What it does: A multi-modal decoupling network is proposed to learn multi-semantic forgery traces from RGB and depth data for facial forgery detection.
Learning Program Synthesis for Integer Sequences from Scratch
Thibault Gauthier (Czech Technical University in Prague), Josef Urban (Czech Technical University in Prague)
GenerationGraph Neural NetworkReinforcement LearningSequential
🎯 What it does: Through self-learning cycles, programs are generated unsupervised from the integer sequences in OEIS using tree search and tree neural networks, ultimately discovering implementations for 27,987 sequences without any human annotations.
Learning Progressive Modality-Shared Transformers for Effective Visible-Infrared Person Re-identification
Hu Lu (Jiangsu University), Pingping Zhang (Dalian University of Technology)
RecognitionRetrievalTransformerImage
🎯 What it does: A progressive modality sharing framework PMT based on Transformer is proposed and implemented, using grayscale images as auxiliary for progressive learning, combined with modality sharing enhanced loss and discriminative center loss to improve visible-infrared person re-identification performance.
Learning Rational Subgoals from Demonstrations and Instructions
Zhezheng Luo (Massachusetts Institute of Technology), Leslie Pack Kaelbling (Massachusetts Institute of Technology)
OptimizationRobotic IntelligenceReinforcement Learning from Human FeedbackReinforcement LearningImage
🎯 What it does: In this paper, the authors propose a method for learning subgoals from weakly labeled demonstrations (RSG) and utilizing these subgoals for efficient long-range planning.
Learning Relational Causal Models with Cycles through Relational Acyclification
Ragib Ahsan (University of Illinois at Chicago), Elena Zheleva (University of Illinois at Chicago)
GraphTabular
🎯 What it does: This paper proposes the 'relation acyclic operation' and proves that under this condition, the existing relational causal discovery algorithm RCD can still maintain measurability and completeness in the presence of cyclic relational causal models, along with experimental validation.
Learning Representations of Bi-level Knowledge Graphs for Reasoning beyond Link Prediction
Chanyoung Chung (KAIST), Joyce Jiyoung Whang (KAIST)
Representation LearningGraph Neural NetworkGraph
🎯 What it does: This paper proposes the concept of a Bi-level Knowledge Graph (Bi-level KG), constructing a novel knowledge graph that includes base-level triples and hierarchical triples, and introduces the BiVE model for embedding learning on this graph; it also designs a random walk-based incremental data augmentation strategy; and proposes two new tasks: Triplet Prediction and Conditional Link Prediction.
Learning Revenue Maximization Using Posted Prices for Stochastic Strategic Patient Buyers
Eitan-Hai Mashiah (Tel Aviv University), Yishay Mansour (Google Research)
OptimizationReinforcement Learning
🎯 What it does: This paper studies the Stackelberg game where the seller maximizes revenue by posting a price sequence in the presence of buyers who can delay their purchases (patient buyers), and provides the characteristics and computational methods for optimal pure and mixed strategies.
Learning Safe Numeric Action Models
Argaman Mordoch (Ben Gurion University), Roni Stern (Washington University)
OptimizationRobotic IntelligenceTabular
🎯 What it does: A safe action model learning algorithm for numerical planning, N-SAM, is proposed, which can learn action models that comply with safety constraints based only on observed trajectories and can be used for planning.
Learning Second-Order Attentive Context for Efficient Correspondence Pruning
Xinyi Ye (Huazhong University of Science and Technology), Zhiguo Cao (Huazhong University of Science and Technology)
Computational EfficiencyTransformerImage
🎯 What it does: A correspondence pruning method based on second-order attention (Attention in Attention) is proposed, which achieves efficient and accurate outlier removal by integrating first-order feature-consistent context and second-order attention-consistent context.
Learning Semantic Alignment with Global Modality Reconstruction for Video-Language Pre-training towards Retrieval
Mingchao Li (Tsinghua University), Kuncai Zhang (Alibaba Group)
RetrievalTransformerContrastive LearningVideoTextMultimodality
🎯 What it does: This paper proposes the FEEL method, which achieves sequence-level semantic alignment during the video-language pre-training phase through global modality reconstruction and cross-modal self-contrast learning, thereby improving the performance of text-based video retrieval and video semantic moment retrieval tasks.
Learning Semantic Degradation-Aware Guidance for Recognition-Driven Unsupervised Low-Light Image Enhancement
Naishan Zheng (University of Science and Technology of China), Feng Zhao (University of Science and Technology of China)
ClassificationObject DetectionImage
🎯 What it does: This paper proposes a semantic degradation-aware guidance (SDAG) method for unsupervised low-light image enhancement, which learns the impact of low-light degradation on semantic features through self-supervised reconstruction and embeds it into the existing ULLIE network to improve downstream visual recognition performance.
Learning Similarity Metrics for Volumetric Simulations with Multiscale CNNs
Georg Kohl (Technical University of Munich), Nils Thuerey (Technical University of Munich)
OptimizationConvolutional Neural NetworkMultimodalityPhysics Related
🎯 What it does: A similarity model based on entropy is proposed, and a multi-scale convolutional network VolSiM is trained to evaluate the similarity of three-dimensional physical simulation data.
Learning Single Image Defocus Deblurring with Misaligned Training Pairs
Yu Li (Harbin Institute of Technology), Wangmeng Zuo (Peng Cheng Laboratory)
RestorationOptical FlowImage
🎯 What it does: This paper proposes a Joint De-blurring and Re-blurring Learning framework (JDRL) that addresses the deformation artifacts caused by training pairing errors in the single image defocus deblurring task through optical flow-based deformation and a spatially invariant re-blurring module.
Learning Temporal-Ordered Representation for Spike Streams Based on Discrete Wavelet Transforms
Jiyuan Zhang (Peking University), Tiejun Huang (Peking University)
RestorationSegmentationConvolutional Neural NetworkSpiking Neural NetworkSupervised Fine-TuningTime Series
🎯 What it does: A waveform-guided spike enhancement (WGSE) module based on discrete wavelet transform (DWT) and lightweight CNN is proposed to learn and generate more effective spike flow representations.
Learning the Finer Things: Bayesian Structure Learning at the Instantiation Level
Chase Yakaboski (Dartmouth College), Jr
TabularBiomedical Data
🎯 What it does: Proposes an instance hierarchy learning method based on Bayesian Knowledge Base (BKB), constructing the BKBSL algorithm using minimum entropy inference and MDL scoring;
Learning to Break Symmetries for Efficient Optimization in Answer Set Programming
Alice Tarzariol (University of Klagenfurt), Mark Law (ILASP Limited)
OptimizationTabular
🎯 What it does: A new method is proposed to enhance the symmetry breaking constraints in Answer Set Programming (ASP) for optimization problems through Inductive Logic Programming (ILP), aiming to improve optimization efficiency.
Learning to Count Isomorphisms with Graph Neural Networks
Xingtong Yu (University of Science and Technology of China), Xinming Zhang (University of Science and Technology of China)
Graph Neural NetworkGraph
🎯 What it does: This paper proposes a new graph neural network model, Count-GNN, to approximate the isomorphic count of a given query subgraph in the input graph.
Learning to Defer with Limited Expert Predictions
Patrick Hemmer (Karlsruhe Institute of Technology), Niklas Kühl (Karlsruhe Institute of Technology)
ClassificationData-Centric LearningImageBiomedical Data
🎯 What it does: This paper proposes a three-step method that utilizes a small number of expert predictions and semi-supervised learning to generate artificial expert predictions, supporting the learning-to-defer algorithm to work efficiently even when expert labels are scarce.
Learning to Generate an Unbiased Scene Graph by Using Attribute-Guided Predicate Features
Lei Wang (Xi'an Jiaotong University), Badong Chen (Xi'an Jiaotong University)
ClassificationObject DetectionGenerationTransformerAuto EncoderImage
🎯 What it does: A framework for generating unbiased scene graphs is proposed, utilizing attribute-guided predicate features to generate balanced data and enhance the fairness of predicate classification.
Learning to Imagine: Distillation-Based Interactive Context Exploitation for Dialogue State Tracking
Jinyu Guo (Beijing University of Posts and Telecommunications), Zihan Wang (University of Tokyo)
Knowledge DistillationTransformerLarge Language ModelText
🎯 What it does: A general module DICE-DST is proposed, which utilizes a teacher encoder to learn dialogue context knowledge and enhances the encoder of historical DST models through transcription attention alignment distillation, addressing the issue of missing key context when only part of the dialogue history is used.
Learning to Know Myself: A Coarse-to-Fine Persona-Aware Training Framework for Personalized Dialogue Generation
Yunpeng Li (Institute of Information Engineering), Wei Peng (Institute of Information Engineering)
GenerationTransformerLarge Language ModelContrastive LearningText
🎯 What it does: A two-stage coarse-to-fine persona-aware training framework is proposed, which first enhances the model's sensitivity to persona by constructing persona-aware questions, and then improves fine-grained consistency through contrastive learning to generate negative samples.
Learning to Learn Better for Video Object Segmentation
Meng Lan (Wuhan University), Dacheng Tao (JD Explore Academy)
Object TrackingSegmentationTransformerVideo
🎯 What it does: A new joint learning framework LLB is proposed, combining target transfer and matching two branches, and a Discriminative Label Generation Module (DLGM) and an Adaptive Fusion Module (AFM) are designed to improve target feature representation.
Learning to Memorize Entailment and Discourse Relations for Persona-Consistent Dialogues
Ruijun Chen (Yunnan University), Xuejie Zhang (Yunnan University)
GenerationTransformerLarge Language ModelText
🎯 What it does: A method is proposed for generating character-consistent dialogues by utilizing memory entailment relationships and discourse-level relationships.
Learning to Play General-Sum Games against Multiple Boundedly Rational Agents
Eric Zhao (Salesforce Research), Stephan Zheng (MosaicML)
OptimizationReinforcement LearningTabularTime Series
🎯 What it does: The paper proposes an algorithmic framework based on no-regret dynamics to train robust chief strategies (i.e., mechanisms) in general multi-agent settings and games, applying it to automated mechanism design, particularly dynamic tax policies.
Learning to Select from Multiple Options
Jiangshu Du (University of Illinois at Chicago), Philip S. Yu (Salesforce Research)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: For multi-choice tasks, two text entailment-based models, Context-TE and Parallel-TE, are proposed;
Learning to Select Pivotal Samples for Meta Re-weighting
Yinjun Wu (University of Pennsylvania), Mayur Naik (University of Pennsylvania)
ClassificationMeta LearningImage
🎯 What it does: A learning framework is proposed to automatically select high-quality meta samples (key samples) from a large-scale defect-laden training set to enhance the performance of the meta reweighting algorithm.
Learning to Select Prototypical Parts for Interpretable Sequential Data Modeling
Yifei Zhang (Institute of Information Engineering Chinese Academy of Sciences), Cunqing Ma (Institute of Information Engineering Chinese Academy of Sciences)
ClassificationExplainability and InterpretabilityConvolutional Neural NetworkRecurrent Neural NetworkTextSequentialBiomedical DataElectrocardiogram
🎯 What it does: This paper proposes a Self-Explaining Selection Model (SESM), which encodes each subsequence as a concept by using a learnable subsequence selection mechanism on the sequence, and then aggregates these concepts with linear weighting to make predictions, thereby achieving interpretability for sequence models.
Learning to Shape Rewards Using a Game of Two Partners
David Mguni (Huawei Research and Development), Yaodong Yang (Peking University)
Reinforcement LearningTabular
🎯 What it does: This paper proposes an automatic reward shaping framework called ROSA based on a two-agent Markov game, where the reward shaping function is learned and dynamically adjusted by an independent Shaper during the training process.
Learning to Super-resolve Dynamic Scenes for Neuromorphic Spike Camera
Jing Zhao (Peking University), Tiejun Huang (University of Pennsylvania)
RestorationData SynthesisSuper ResolutionConvolutional Neural NetworkSpiking Neural NetworkImageVideo
🎯 What it does: This paper studies an end-to-end SpikeSR-Net network for super-resolution reconstruction of high-quality high-resolution images from low-resolution binary spike stream.
Learning Topology-Specific Experts for Molecular Property Prediction
Suyeon Kim (Pohang University of Science and Technology), Hwanjo Yu (Pohang University of Science and Technology)
ClassificationDrug DiscoveryGraph Neural NetworkMixture of ExpertsGraphTabular
🎯 What it does: We propose TopExpert, which utilizes a clustering gating module to divide molecules into several expert groups based on topological similarity, allowing each expert to learn features within the corresponding topological subset for molecular property prediction.
Learning towards Selective Data Augmentation for Dialogue Generation
Xiuying Chen (King Abdullah University of Science and Technology), Rui Yan (Renmin University of China)
GenerationData SynthesisRecurrent Neural NetworkAuto EncoderGenerative Adversarial NetworkText
🎯 What it does: A Selective Data Augmentation framework (SDA) is proposed, which selects low-quality yet representative data samples for subsequent data augmentation from the perspectives of generation quality and representativeness through a dual adversarial network, thereby enhancing the performance of dialogue generation models.
Learning with Partial Labels from Semi-supervised Perspective
Ximing Li (Jilin University), Jihong Ouyang (Jilin University)
ClassificationConvolutional Neural NetworkImage
🎯 What it does: Transforming partial label learning problems into semi-supervised learning tasks, using high-confidence pseudo-labels and consistency regularization to train the model.
Learning-Assisted Algorithm Unrolling for Online Optimization with Budget Constraints
Jianyi Yang (University of California), Shaolei Ren (University of California)
OptimizationTime Series
🎯 What it does: This paper proposes a learning-assisted online optimization framework LAAU based on algorithm unrolling, aimed at solving online optimization problems with strict short-term budget constraints.
Learning-Augmented Algorithms for Online TSP on the Line
Themistoklis Gouleakis (National University of Singapore), Golnoosh Shahkarami (Max Planck Institute for Informatics)
Optimization
🎯 What it does: The online one-dimensional TSP problem was studied, and machine learning predictions were introduced, designing three algorithms: FARFIRST, NEARFIRST, and PIVOT;
LeNo: Adversarial Robust Salient Object Detection Networks with Learnable Noise
He Wang (Huazhong University of Science and Technology), He Tang (Huazhong University of Science and Technology)
Object DetectionAdversarial AttackConvolutional Neural NetworkImage
🎯 What it does: A lightweight learnable noise module, LeNo, has been designed and implemented to enhance the robustness of salient object detection models against adversarial attacks while maintaining high accuracy on clean images.
Less Is More Important: An Attention Module Guided by Probability Density Function for Convolutional Neural Networks
Jingfen Xie (Peking University), Jian Zhang (Peking University)
Convolutional Neural NetworkImageBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A probability density function-based attention module, PdfAM, is designed to generate 3D attention maps.
Let Graph Be the Go Board: Gradient-Free Node Injection Attack for Graph Neural Networks via Reinforcement Learning
Mingxuan Ju (University of Notre Dame), Yanfang Ye (Case Western Reserve University)
Adversarial AttackGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: This study investigates node injection attacks on black-box graph neural networks and proposes a gradient-free reinforcement learning attack framework called G2A2C.
Let the Data Choose: Flexible and Diverse Anchor Graph Fusion for Scalable Multi-View Clustering
Pei Zhang (Huawei Poisson Lab), Lei Luo (National University of Defense Technology)
OptimizationGraph Neural NetworkMultimodality
🎯 What it does: A multi-size, variable anchor-based anchor graph fusion method is proposed to achieve scalable multi-view clustering.
Leveraging Contaminated Datasets to Learn Clean-Data Distribution with Purified Generative Adversarial Networks
Bowen Tian (Sun Yat-sen University), Jianxing Yu (Sun Yat-sen University)
GenerationData-Centric LearningGenerative Adversarial NetworkImage
🎯 What it does: The paper proposes a PuriGAN model that learns the distribution of clean data using a small dataset containing only polluted instances.
Leveraging Modality-Specific Representations for Audio-Visual Speech Recognition via Reinforcement Learning
Chen Chen (Nanyang Technological University), Eng Siong Chng (Nanyang Technological University)
RecognitionTransformerReinforcement LearningVideoMultimodalityAudio
🎯 What it does: A reinforcement learning-based audio-video speech recognition framework MSRL is proposed, which utilizes modality-specific representations to dynamically compensate for noise pollution in audio information, enhancing the robustness of speech recognition.
Leveraging Structure for Improved Classification of Grouped Biased Data
Daniel Zeiberg (Northeastern University), Predrag Radivojac (Northeastern University)
ClassificationTabular
🎯 What it does: This paper studies the semi-supervised binary classification problem under the condition of naturally grouped data and biased labeled data, proposing an algorithm that utilizes data structure to improve classification performance.
Leveraging Sub-class Discimination for Compositional Zero-Shot Learning
Xiaoming Hu (University of Science and Technology of China), Zilei Wang (University of Science and Technology of China)
ClassificationRecognitionConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: A two-stage method is proposed for zero-shot learning through subclass discrimination, which first aligns synthetic embeddings using contrastive learning at the feature level, and then implements dynamic prototype updates using prototype modulation at the classifier level to enhance the recognition of unseen attribute-object combinations.
Leveraging Weighted Cross-Graph Attention for Visual and Semantic Enhanced Video Captioning Network
Deepali Verma (Indian Institute of Technology Banaras Hindu University), Tanima Dutta (Indian Institute of Technology Banaras Hindu University)
Object DetectionGenerationConvolutional Neural NetworkRecurrent Neural NetworkGraph Neural NetworkVideoText
🎯 What it does: A cross-graph attention mechanism that simultaneously utilizes visual region graphs and semantic knowledge graphs is proposed to achieve video description generation.
LidarMultiNet: Towards a Unified Multi-Task Network for LiDAR Perception
Dongqiangzi Ye (TuSimple), Hassan Foroosh (University of Central Florida)
Object DetectionSegmentationAutonomous DrivingPoint Cloud
🎯 What it does: A unified LiDAR multi-task network called LidarMultiNet is proposed, capable of simultaneously performing 3D object detection, semantic segmentation, and panoptic segmentation.
Lifelong Compression Mixture Model via Knowledge Relationship Graph
Fei Ye (University of York), Adrian G. Bors (University of York)
CompressionOptimizationAuto EncoderImage
🎯 What it does: This paper proposes a lifelong generative mixture model (LGMM) based on dynamic expansion and compression, addressing the issues of network forgetting and model size inflation in task-agnostic continual learning (TFCL).
Lifelong Embedding Learning and Transfer for Growing Knowledge Graphs
Yuanning Cui (Nanjing University), Wei Hu (Nanjing University)
Representation LearningGraph Neural NetworkAuto EncoderGraph
🎯 What it does: This paper proposes a lifelong knowledge graph embedding model named LKGE, which can quickly learn new facts, transfer existing knowledge, and retain old knowledge as the knowledge graph continues to grow.
Lifelong Person Re-identification via Knowledge Refreshing and Consolidation
Chunlin Yu (ShanghaiTech University), Jingya Wang (ShanghaiTech University)
RecognitionKnowledge DistillationConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: A lifecycle person re-identification method KRC is proposed, which combines knowledge replay, knowledge refresh, and knowledge consolidation in three stages, allowing the model to maintain old knowledge while improving the performance of both old and new tasks as new tasks are continuously added.
Lifelong Variational Autoencoder via Online Adversarial Expansion Strategy
Fei Ye (University of York), Adrian G. Bors (University of York)
GenerationData SynthesisAuto EncoderGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes an Online Adversarial Expansion Strategy (OAES) that dynamically increases the capacity of the VAE model to mitigate catastrophic forgetting in the scenario of Task-Free Continual Learning (TFCL).
Lifted Inference with Linear Order Axiom
Jan Tóth (Czech Technical University in Prague), Ondřej Kuželka (Czech Technical University in Prague)
Graph
🎯 What it does: This paper proposes a dynamic programming algorithm that can compute the weighted model counting (WFOMC) of two-variable first-order logic (and its counting quantifier extension C₂) with linear order axioms in polynomial time.
Lifting (D)QBF Preprocessing and Solving Techniques to (D)SSAT
Che Cheng (National Taiwan University), Jie-Hong R. Jiang (National Taiwan University)
TabularBenchmark
🎯 What it does: The first DSSAT solver has been implemented (by transforming DSSAT into SSAT and using dependency elimination) and an independent DSSATpre preprocessor has been developed, which simplifies SSAT/DSSAT formulas based on HQSpre.
LIMIP: Lifelong Learning to Solve Mixed Integer Programs
Sahil Manchanda (Indian Institute of Technology), Sayan Ranu (Indian Institute of Technology)
OptimizationKnowledge DistillationGraph Neural NetworkTabular
🎯 What it does: The study introduces a lifelong learning framework in Mixed Integer Programming (MIP) and proposes the LIMIP method to continuously learn and make branching decisions in multi-task sequences.
Linear Regularizers Enforce the Strict Saddle Property
Matthew Ubl (University of Florida), Kasra Yazdani (University of Florida)
Optimization
🎯 What it does: Proposes an improved gradient descent method based on linear regularization, which can enforce strict saddle point properties near non-strict saddle points and achieve escape.
Linking People across Text and Images Based on Social Relation Reasoning
Yang Lei (Guangxi University), Qingbao Huang (Guangxi University)
RecognitionRetrievalGraph Neural NetworkTransformerImageTextMultimodality
🎯 What it does: A model based on social relationship reasoning (SRR) is proposed to locate corresponding characters between images and text.
Linking Sketch Patches by Learning Synonymous Proximity for Graphic Sketch Representation
Sicong Zang (Shanghai Jiao Tong University), Lei Xu (Shanghai Jiao Tong University)
RestorationGenerationConvolutional Neural NetworkRecurrent Neural NetworkGraph Neural NetworkImage
🎯 What it does: The SP-gra2seq method is proposed, which generates robust graphical sketch representations by learning semantic similarity (synonym proximity) to dynamically connect sketch patches and perform graph convolution message passing, supplemented by clustering constraints.
LIQUID: A Framework for List Question Answering Dataset Generation
Seongyun Lee (Korea University), Jaewoo Kang (Korea University)
GenerationData SynthesisTransformerLarge Language ModelSupervised Fine-TuningTextBiomedical Data
🎯 What it does: Proposes the LIQUID framework, which can automatically generate multi-answer list question-answer datasets from unlabeled texts (such as Wikipedia and PubMed), significantly reducing the cost of manual annotation.
Loan Fraud Users Detection in Online Lending Leveraging Multiple Data Views
Sha Zhao (Zhejiang University), Gang Pan (FinVolution Group)
ClassificationRecommendation SystemAnomaly DetectionRecurrent Neural NetworkGenerative Adversarial NetworkTabularFinance Related
🎯 What it does: An end-to-end deep multi-view learning framework is proposed to detect online lending fraud users from multi-source heterogeneous data such as user attributes, APP installation logs, application installation behaviors, and application lists.
Local Explanations for Reinforcement Learning
Ronny Luss (IBM Research), Miao Liu (IBM Research)
Explainability and InterpretabilityReinforcement Learning
🎯 What it does: A local explanation method based on reinforcement learning policy dynamics is proposed—Strategic State eXplanation (SSX), which explains deep RL policies by learning meta-states and selecting strategic states with maximum path coverage within each meta-state.
Local Intrinsic Dimensional Entropy
Rohan Ghosh (National University of Singapore), Mehul Motani (National University of Singapore)
Convolutional Neural NetworkImage
🎯 What it does: A continuous space entropy measure based on local intrinsic dimension, called ID-Entropy, is proposed, and it is proven to satisfy the main properties of discrete entropy, thereby constructing a new information bottleneck criterion and a generalized error upper bound.
Local Justice and Machine Learning: Modeling and Inferring Dynamic Ethical Preferences toward Allocations
Violet (Xinying) Chen (Stevens Institute of Technology), Hoda Heidari (Carnegie Mellon University)
OptimizationReinforcement Learning from Human FeedbackReinforcement LearningTabularElectronic Health Records
🎯 What it does: A dynamic moral preference modeling framework based on Markov Decision Processes (MDP) and piecewise linear reward functions is proposed, utilizing active preference-based reward learning to infer the ethical orientation of social planners in allocating scarce resources from human feedback.
Local Path Integration for Attribution
Peiyu Yang (University of Western Australia), Ajmal Mian (University of Western Australia)
Explainability and InterpretabilityConvolutional Neural NetworkImage
🎯 What it does: This study proposes the Local Path Integration (LPI) method, which improves traditional path attribution methods by integrating gradients using reference samples from the local distribution of the input, addressing shortcomings in weak dependence and reference selection.
Local-Global Defense against Unsupervised Adversarial Attacks on Graphs
Di Jin (Tianjin University), Zhen Wang (Northwestern Polytechnical University)
Representation LearningAdversarial AttackGraph Neural NetworkGraph
🎯 What it does: This paper proposes an unsupervised graph pre-training method based on local-global defense to resist adversarial attacks on graph structures.
Locate Then Generate: Bridging Vision and Language with Bounding Box for Scene-Text VQA
Yongxin Zhu (University of Science and Technology of China), Linli Xu (University of Science and Technology of China)
RecognitionObject DetectionGenerationTransformerVision Language ModelImageTextMultimodality
🎯 What it does: This paper proposes a 'Locate-Then-Generate' (LTG) framework for the Scene Text Visual Question Answering (STVQA) task.
Logic and Commonsense-Guided Temporal Knowledge Graph Completion
Guanglin Niu (Beihang University), Bo Li (Beihang University)
GraphTime Series
🎯 What it does: This paper proposes a temporal knowledge graph completion model LCGE that simultaneously considers event timeliness, causality, and common sense.
Logical Satisfiability of Counterfactuals for Faithful Explanations in NLI
Suzanna Sia (Johns Hopkins University), Lambert Mathias (Meta AI Research)
Explainability and InterpretabilityPrompt EngineeringTextMultimodality
🎯 What it does: A method is proposed to evaluate the credibility of free-text explanations in natural language inference tasks by generating counterfactual hypotheses and testing their logical satisfiability.
LoNe Sampler: Graph Node Embeddings by Coordinated Local Neighborhood Sampling
Konstantin Kutzkov (Teva Pharmaceuticals)
ClassificationRepresentation LearningGraph Neural NetworkGraph
🎯 What it does: This paper proposes LONE SAMPLER, a method for generating discrete node embeddings through coordinated local neighborhood sampling, utilizing the local structure of the graph and sampling theory to construct interpretable and scalable node representations.
Long-Tail Cross Modal Hashing
Zijun Gao (Shandong University), Jinglin Zhang (George Mason University)
RetrievalAuto EncoderImageTextMultimodality
🎯 What it does: This paper proposes LtCMH, a cross-modal hashing method for long-tail multimodal data, which utilizes autoencoders to mine the individuality and commonality of modalities, and enhances the representation and discriminability of tail labels through dynamic meta-feature fusion.
LORE: Logical Location Regression Network for Table Structure Recognition
Hangdi Xing (Zhejiang University), Zhi Yu (Zhejiang University)
RecognitionTransformerTabular
🎯 What it does: This paper proposes a new table structure recognition framework called LORE, which transforms the table recognition problem into simultaneously regressing the spatial and logical positions of cells, directly outputting complete logical coordinates without the need for post-processing.
Losses over Labels: Weakly Supervised Learning via Direct Loss Construction
Dylan Sam (Carnegie Mellon University), J. Zico Kolter (Carnegie Mellon University)
ClassificationConvolutional Neural NetworkImageTextBenchmark
🎯 What it does: A weakly supervised learning framework (Losses over Labels, LoL) is proposed that directly converts weak labelers into loss functions, eliminating the need for generating pseudo-labels.
Lottery Pools: Winning More by Interpolating Tickets without Increasing Training or Inference Cost
Lu Yin (Eindhoven University of Technology), Mykola Pechenizkiy (University of Liverpool)
OptimizationConvolutional Neural NetworkImage
🎯 What it does: By performing linear interpolation on Lottery Ticket subnetworks obtained from different iterations, stronger sparse subnetworks are constructed, and these interpolated subnetworks are then interpolated back to the dense network, resulting in performance improvements.
Low Resource Quantitative Information Extraction via Structure Searching and Prefix-Based Text Generation
Tongliang Li (Beihang University), Zhoujun Li (Beihang University)
GenerationData SynthesisTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: In low-resource environments, this paper proposes to automatically generate training samples through constituency parse tree search and use a prefix text generation model to extract quantitative information.
Low-Light Image Enhancement Network Based on Multi-Scale Feature Complementation
Yong Yang (Tiangong University), Weiguo Wan (Jiangxi University of Finance and Economics)
RestorationConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: Design and implement a multi-scale feature complementary low-light image enhancement network (LIEN-MFC) based on a four-branch U-shaped structure.
Low-Light Video Enhancement with Synthetic Event Guidance
Lin Liu, Qi Tian (University of Science and Technology of China)
RestorationTransformerVideo
🎯 What it does: This paper proposes a three-stage low-light video enhancement framework that first restores normal light events using synthetic events, then fuses them with video frames, and finally completes the enhancement through a dual-branch network.
Low-Resource Personal Attribute Prediction from Conversations
Yinan Liu (Nankai University), Jiaoyan Chen (University of Manchester)
ClassificationRecommendation SystemTransformerLarge Language ModelText
🎯 What it does: Under low-resource conditions, the PEARL framework is proposed to predict user personal attributes using unannotated dialogue data in conversations.
LWSIS: LiDAR-Guided Weakly Supervised Instance Segmentation for Autonomous Driving
Xiang Li (Beijing Institute of Technology), Jianbing Shen (University of Macau)
SegmentationAutonomous DrivingPoint Cloud
🎯 What it does: Using LiDAR point clouds and 3D boxes as weak supervision to train a 2D instance segmentation model, reducing the need for 2D mask annotations.
M-sense: Modeling Narrative Structure in Short Personal Narratives Using Protagonist’s Mental Representations
Prashanth Vijayaraghavan (Massachusetts Institute of Technology Media Lab), Deb Roy (Massachusetts Institute of Technology Media Lab)
TransformerSupervised Fine-TuningTextMultimodality
🎯 What it does: A Reddit personal narrative dataset named STORIES was constructed, and the M-sense model was proposed to automatically identify the climax and conclusion in narratives.
M3AE: Multimodal Representation Learning for Brain Tumor Segmentation with Missing Modalities
Hong Liu (Xiamen University), Yefeng Zheng (Tencent Healthcare)
SegmentationRepresentation LearningConvolutional Neural NetworkAuto EncoderMultimodalityBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A full-process framework based on a multi-modal masked autoencoder (M³AE) is proposed for brain tumor segmentation with missing modalities, including self-supervised pre-training, model inversion to complete missing modalities, and memory-efficient self-distillation.
MA-GCL: Model Augmentation Tricks for Graph Contrastive Learning
Xumeng Gong (Beijing University of Posts and Telecommunications), Chuan Shi (Beijing University of Posts and Telecommunications)
ClassificationRepresentation LearningGraph Neural NetworkContrastive LearningGraph
🎯 What it does: Introducing Model Augmentation (MA-GCL) in graph contrastive learning, which generates more diverse contrastive views by altering the architecture of the view encoder, enhancing the effectiveness of unsupervised node representation learning.
Machines of Finite Depth: Towards a Formalization of Neural Networks
Pietro Vertechi, Mattia G. Bergomi
OptimizationComputational EfficiencyConvolutional Neural NetworkRecurrent Neural Network
🎯 What it does: A unified framework called 'Machines of Finite Depth' is proposed, which mathematically formalizes neural networks and their variants (fully connected, convolutional, recurrent, and networks with shortcut connections), and shows that both forward and backward computations can be viewed as solving the same resolvent.
MAGIC: Mask-Guided Image Synthesis by Inverting a Quasi-robust Classifier
Mozhdeh Rouhsedaghat (University of Southern California), Iacopo Masi (Sapienza University of Rome)
GenerationData SynthesisAuto EncoderGenerative Adversarial NetworkImage
🎯 What it does: MAGIC is proposed, a 'One-Shot' image synthesis method based on a single image and a binary mask.
MAPS-KB: A Million-Scale Probabilistic Simile Knowledge Base
Qianyu He (Fudan University), Yanghua Xiao (Fudan University)
GenerationRetrievalTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: The first million-level probabilistic personification (simile) knowledge base MAPS-KB has been constructed, covering 4.3M (topic, property, vehicle) triples, providing multi-dimensional probabilistic information (credibility, typicality), and achieving SOTA in three types of downstream tasks (personification explanation, generation, and text polishing).
Markov Decision Processes with Time-Varying Geometric Discounting
Jiarui Gan (University of Oxford), Goran Radanovic (Max Planck Institute for Software Systems)
OptimizationReinforcement Learning
🎯 What it does: This study investigates the time-varying geometric discount in infinite-horizon Markov decision processes, establishes a subgame perfect equilibrium (SPE) framework, proves its existence, computes EXPTIME-hardness, and proposes an ε-SPE approximation algorithm.
MaskBooster: End-to-End Self-Training for Sparsely Supervised Instance Segmentation
Shida Zheng (Hikvision Research Institute), Wenming Tan (Hikvision Research Institute)
Object DetectionSegmentationKnowledge DistillationImage
🎯 What it does: This paper proposes MaskBooster, an end-to-end self-training framework for sparse supervised instance segmentation (where only a small portion of instances have mask annotations).
Materialisation-Based Reasoning in DatalogMTL with Bounded Intervals
Przemysław A. Wałęga (University of Oxford), Bernardo Cuenca Grau (University of Oxford)
TabularTime SeriesBenchmark
🎯 What it does: For finite interval programs in DatalogMTL, a materialized and terminating reasoning algorithm is proposed, which achieves model unfolding and fact reasoning by detecting saturation states.
Maximizing the Probability of Fixation in the Positional Voter Model
Petros Petsinis (Aarhus University), Panagiotis Karras (Aarhus University)
OptimizationGraph Neural NetworkGraph
🎯 What it does: Introduce a voting model with location bias in networks and study which nodes to bias under a given budget to maximize the fixed probability of new features;
Maximum Entropy Population-Based Training for Zero-Shot Human-AI Coordination
Rui Zhao (Tencent AI Lab), Wei Yang (Tsinghua University)
Robotic IntelligenceReinforcement LearningSequential
🎯 What it does: The study trains RL agents that can collaborate with humans without using human data, proposing the Maximum Entropy Population Training (MEP) method.
MCL: Multi-Granularity Contrastive Learning Framework for Chinese NER
Shan Zhao (HeFei University of Technology), Meng Wang (PLA Academy of Military Science)
RecognitionRecurrent Neural NetworkContrastive LearningText
🎯 What it does: This paper proposes a Multi-Granularity Contrastive Learning (MCL) framework that integrates dictionary information into the character-word grid structure for Chinese named entity recognition tasks, achieving mutual calibration of character and word representations and highlighting keywords through Cross-Granularity Contrastive Learning (CCL) and Bi-Granularity Contrastive Learning (BCL).
MDM: Molecular Diffusion Model for 3D Molecule Generation
Lei Huang (City University of Hong Kong), Ka-Chun Wong (Tencent AI Lab)
GenerationData SynthesisDrug DiscoveryGraph Neural NetworkDiffusion modelPoint Cloud
🎯 What it does: A diffusion model for 3D molecular generation, MDM, has been developed, capable of generating high-quality 3D molecules from scratch.
Mean Estimation of Truncated Mixtures of Two Gaussians: A Gradient Based Approach
Sai Ganesh Nagarajan (École Polytechnique Fédérale de Lausanne), Samson Yu (National University of Singapore)
Optimization
🎯 What it does: This paper proposes a gradient-based variant of the EM algorithm for estimating the means of two groups in a truncated Gaussian mixture model with symmetric means.
Mean-Shifted Contrastive Loss for Anomaly Detection
Tal Reiss (Hebrew University of Jerusalem), Yedid Hoshen (Hebrew University of Jerusalem)
Anomaly DetectionConvolutional Neural NetworkTransformerContrastive LearningImage
🎯 What it does: Fine-tune normal samples using the new Mean-Shifted Contrastive Loss on pre-trained ImageNet features to enhance single-class anomaly detection performance.
Mediated Cheap Talk Design
Itai Arieli (Technion Israel Institute of Technology), Moshe Tennenholtz (Technion Israel Institute of Technology)
Optimization
🎯 What it does: This paper studies the problem of 'mediated cheap talk' where two information senders communicate with one receiver through a trusted intermediary in the absence of commitment rights. It clarifies the achievable strategy distribution and provides an O(n log n) algorithm for solving the optimal sender equilibrium, while also proposing a simple mechanism for the receiver's optimal strategy.
MEID: Mixture-of-Experts with Internal Distillation for Long-Tailed Video Recognition
Xinjie Li (Pennsylvania State University), Huijuan Xu (Pennsylvania State University)
RecognitionKnowledge DistillationMixture of ExpertsVideoBenchmark
🎯 What it does: A two-expert mixture framework based on internal distillation has been designed and implemented to address the frame-level imbalance problem in multi-label long-tail video recognition using frame-level attention and complementary frame selection.
Memorization Weights for Instance Reweighting in Adversarial Training
Jianfu Zhang (RIKEN AIP), Qibin Zhao (Shanghai Jiao Tong University)
OptimizationAdversarial AttackConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: This paper proposes a memory pool-based instance weighting method that uses self-supervised feature quantization to evaluate the typicality of samples, reducing overfitting to anomalous samples and thereby enhancing the robustness of adversarial training.
Memory-Aided Contrastive Consensus Learning for Co-salient Object Detection
Peng Zheng (Nanjing University of Aeronautics and Astronautics), Huan Xiong (Mohamed bin Zayed University of Artificial Intelligence)
Object DetectionTransformerGenerative Adversarial NetworkContrastive LearningImage
🎯 What it does: This paper proposes a memory contrast-based co-salient object detection framework called MCCL, which achieves real-time high-precision detection using a Transformer encoder along with three main modules: GCAM, MCM, and AIL.
Memory-Augmented Theory of Mind Network
Dung Nguyen (Deakin University), Truyen Tran (Deakin University)
Recurrent Neural NetworkSequential
🎯 What it does: A memory-enhanced theoretical mind-reading network called ToMMY has been constructed to attribute the psychological state of an observing agent from its historical and current trajectories, and to predict its future behavior;
Memory-Oriented Structural Pruning for Efficient Image Restoration
Xiangsheng Shi (Tsinghua University), Yu Wang (Tsinghua University)
RestorationSuper ResolutionOptimizationConvolutional Neural NetworkImage
🎯 What it does: A memory-oriented structure pruning method for image restoration models, MOSP, has been designed. By adding compressors on skip connections and grouping iterative pruning based on memory consumption, a significant reduction in peak memory is achieved.
Meta-Auxiliary Learning for Adaptive Human Pose Prediction
Qiongjie Cui (Nanjing University of Science and Technology), Weiqing Li (Nanjing University of Science and Technology)
Pose EstimationMeta LearningTransformerVideo
🎯 What it does: A test-time adaptive skeletal pose prediction framework is proposed, which quickly fine-tunes model parameters through self-supervised auxiliary tasks during inference.
Meta-Learning for Simple Regret Minimization
Javad Azizi (University of Southern California), Sumeet Katariya (Amazon)
Recommendation SystemMeta Learning
🎯 What it does: A meta-learning framework for simple regret minimization is proposed, which allows learning agents to interact with a series of bandit tasks sampled independently and identically from an unknown prior distribution, and learn their meta-parameters to perform better in future tasks.
Meta-Reinforcement Learning Based on Self-Supervised Task Representation Learning
Mingyang Wang (Technical University Munich), Alois Knoll (Technical University Munich)
Meta LearningRecurrent Neural NetworkReinforcement LearningAuto EncoderContrastive LearningSequential
🎯 What it does: Proposes the MoSS algorithm, which utilizes self-supervised task representation learning to achieve Meta-RL for non-parametric, non-stationary, and OOD tasks;
Meta-Sketch: A Neural Data Structure for Estimating Item Frequencies of Data Streams
Yukun Cao (University of Science and Technology of China), Xike Xie (University of Science and Technology of China)
Meta LearningTime Series
🎯 What it does: A 'Meta-Sketch' data structure based on meta-learning and memory-augmented neural networks is proposed for estimating the frequency of items in data streams within limited space.
MetaTPTrans: A Meta Learning Approach for Multilingual Code Representation Learning
Weiguo Pian (University of Luxembourg), Tegawendé F. Bissyandé (Université Virtuelle du Burkina Faso)
Representation LearningMeta LearningAI Code AssistantTransformerText
🎯 What it does: The multi-language code representation learning method MetaTPTrans, based on meta-learning, utilizes a meta-learner to dynamically generate feature extractor parameters for each programming language, achieving joint learning of language-independent and language-specific information.