IJCAI 2023 Papers — Page 5
International Joint Conference on Artificial Intelligence · 639 papers
Neural Capacitated Clustering
Jonas K. Falkner (University of Hildesheim), Lars Schmidt-Thieme (University of Hildesheim)
OptimizationGraph Neural NetworkTransformerSupervised Fine-TuningBenchmark
🎯 What it does: Proposed a deep learning-based neural capacitated clustering method (Neural Capacitated Clustering, NCC), which uses graph neural networks to learn point-to-cluster assignment probabilities and dynamically satisfies capacity constraints during the iterative k-means process.
Neuro-Symbolic Class Expression Learning
Caglar Demir (Paderborn University), Axel-Cyrille Ngonga Ngomo (Paderborn University)
Representation LearningConvolutional Neural NetworkReinforcement LearningGraphBiomedical DataBenchmark
🎯 What it does: Propose a neuro-symbolic class expression learning method called DRILL, modeling description logic class expression learning as a reinforcement learning problem, using deep Q-learning to drive the search for faster convergence to the target expression.
Neuro-Symbolic Learning of Answer Set Programs from Raw Data
Daniel Cunnington (IBM Research Europe), Alessandra Russo (Imperial College London)
Explainability and InterpretabilityData-Centric LearningConvolutional Neural NetworkRecurrent Neural NetworkImage
🎯 What it does: Propose an end-to-end neuro-symbolic learning framework NSIL that can simultaneously train a general neural network and an Answer Set Programming (ASP) knowledge base using only raw data labels, learning the perceptual and inferential mapping from raw inputs to target labels.
New Algorithms for the Fair and Efficient Allocation of Indivisible Chores
Jugal Garg (University of Illinois Urbana Champaign), John Qin (University of Illinois Urbana Champaign)
Optimization
🎯 What it does: This paper studies how to achieve allocation in additive chores distribution problems while maintaining fairness (EF1 or EFX) and efficiency (fPO), and provides constructive polynomial-time algorithms for three agents, two-type agents, and three binary agents; meanwhile, it proves that EFX+fPO may not exist in two-type binary instances.
New Bounds and Constraint Programming Models for the Weighted Vertex Coloring Problem
Olivier Goudet (Université d'Angers), David Lesaint (Université d'Angers)
OptimizationGraphBenchmark
🎯 What it does: Designed and implemented three models, including iterative vertex reduction rules, upper bound calculation, and constraint-based programming, to solve the weighted vertex coloring problem.
New Fairness Concepts for Allocating Indivisible Items
Ioannis Caragiannis (Aarhus University), Giovanna Varricchio (Goethe University Frankfurt)
Optimization
🎯 What it does: This paper proposes two new fairness concepts for the indivisible item allocation problem—cognitive EFX (EEFX) and minimum EFX share (MXS), and proves that they always exist in any instance and can be efficiently computed using polynomial-time algorithms.
Non-Lambertian Multispectral Photometric Stereo via Spectral Reflectance Decomposition
Jipeng Lv (Peking University), Boxin Shi (Peking University)
Depth EstimationConvolutional Neural NetworkImage
🎯 What it does: Propose a multispectral photogrammetry method named NeuralMPS, which utilizes spectral reflectance decomposition (SRD) to convert non-Lambertian spectral MPS problems into non-Lambertian CPS problems with unknown illumination intensity, thereby recovering surface normals;
Non-Obvious Manipulability in Extensive-Form Mechanisms: The Revelation Principle for Single-Parameter Agents
Thomas Archbold (King's College London), Carmine Ventre (King's College London)
Optimization
🎯 What it does: This paper studies non-obvious manipulability (NOM) mechanisms under bounded rationality, proposing a generalized labeling framework that applies cyclic monotonicity to indirect mechanisms, thereby characterizing the implementability of social choice functions for single-parameter agents and two-agent-two-type domains, and proving the equivalence between indirect and direct mechanisms in these scenarios, forming a new revelation principle.
Norm Deviation in Multiagent Systems: A Foundation for Responsible Autonomy
Amika M. Singh (Harvard Law School), Munindar P. Singh (North Carolina State University)
Safty and PrivacyText
🎯 What it does: Propose a conceptual framework of norm deviation based on Habermas's theory of communicative action and case law, providing a theoretical foundation for responsible autonomous intelligence
Not Only Pairwise Relationships: Fine-Grained Relational Modeling for Multivariate Time Series Forecasting
Jinming Wu (Beijing University of Posts and Telecommunications), Jianxin Liao (Beijing University of Posts and Telecommunications)
Graph Neural NetworkTime Series
🎯 What it does: Proposed ReMo, a fine-grained relationship modeling method based on hypergraphs, to enhance the performance of multivariate time series forecasting.
Null-Space Diffusion Sampling for Zero-Shot Point Cloud Completion
Xinhua Cheng (Peking University), Jian Zhang (Peking University)
RestorationDiffusion modelPoint Cloud
🎯 What it does: Propose a zero-shot point cloud completion framework NSDS, which utilizes a pre-trained diffusion model combined with null-space sampling to achieve holistic reconstruction of degraded point clouds;
ODEE: A One-Stage Object Detection Framework for Overlapping and Nested Event Extraction
Jinzhong Ning (Dalian University of Technology), Hongfei Lin (Dalian University of Technology)
Object DetectionTransformerLarge Language ModelTextBiomedical DataFinance Related
🎯 What it does: Propose a one-stage object detection-based framework called ODEE, which utilizes vertex-based marking and trigger/argument span and type prediction to directly extract overlapping and nested events.
On a Voter Model with Context-Dependent Opinion Adoption
Luca Becchetti (Sapienza University of Rome), Francesco Pasquale (Tor Vergata University)
🎯 What it does: Investigated a context-aware voting model, analyzing the fixed probability and consensus time under both asynchronous and synchronous update patterns, and provided theoretical results for both unbiased and biased scenarios.
On Adversarial Robustness of Demographic Fairness in Face Attribute Recognition
Huimin Zeng (University of Illinois at Urbana-Champaign), Dong Wang (University of Illinois at Urbana-Champaign)
RecognitionSafty and PrivacyAdversarial AttackConvolutional Neural NetworkImage
🎯 What it does: Investigate the adversarial robustness of facial attribute recognition (FAR) models in terms of demographic fairness, propose cross-attribute, no test data, clean labels, group-level poisoning attacks, and design robust fair training defense.
On Approximating Total Variation Distance
Arnab Bhattacharyya (National University of Singapore), N. V. Vinodchandran (University of Nebraska-Lincoln)
Optimization
🎯 What it does: This paper studies the computational complexity of total variation distance between two product distributions (and finite parameter Bayesian networks), and provides a deterministic FPTAS for specific subclasses (uniform distributions or finite parameter distributions).
On Conditional and Compositional Language Model Differentiable Prompting
Jonathan Pilault (Mila - Quebec AI Institute, Polytechnique Montreal), Markus Dreyer (Amazon Alexa)
GenerationRepresentation LearningTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: This paper proposes a conditional differentiable prompt generation model, PROPS, based on a neural production system (ProdSys), which can automatically generate continuous prompts according to task instructions, input metadata, and other textual conditions, thereby completing various natural language generation tasks on frozen pre-trained language models;
On Discovering Interesting Combinatorial Integer Sequences
Martin Svatoš (Czech Technical University in Prague), Ondřej Kuželka (Czech Technical University in Prague)
GenerationData SynthesisSequential
🎯 What it does: Automatically generate integer sequences with combinatorial explanations and build a database containing over 26,000 unique sequences.
On Efficient Transformer-Based Image Pre-training for Low-Level Vision
Wenbo Li (Chinese University of Hong Kong), Jiangbo Lu (SmartMore Corporation)
RestorationExplainability and InterpretabilityTransformerImage
🎯 What it does: Investigate the role of pretraining Transformers in low-level vision tasks and propose an efficient Encoder-Decoder Transformer (EDT) model, conducting experiments on three tasks: super-resolution, denoising, and de-raining.
On Lower Bounds for Maximin Share Guarantees
Halvard Hummel (Norwegian University of Science and Technology)
Optimization
🎯 What it does: The study demonstrates that under multi-agent additive value settings, if the number of items does not exceed the number of agents plus a constant c, a maximum-minimum share (MMS) allocation exists; it also provides an upper bound for the minimal number of agents n_c satisfying this condition.
On Optimal Strategies for Wordle and General Guessing Games
Michael Cunanan (University of New South Wales), Michael Thielscher (University of New South Wales)
OptimizationText
🎯 What it does: This paper proposes a general framework for solving optimal strategies in guessing games without performing full search, and implements and verifies it using Wordle as an example.
On the Compilability of Bounded Numeric Planning
Nicola Gigante (Free University of Bozen-Bolzano), Enrico Scala (University of Brescia)
Optimization
🎯 What it does: Studied the compilability of finite numerical planning, constructed a hierarchical language framework, and provided compilability relations between different fragments and classical planning.
On the Complexity of Counterfactual Reasoning
Yunqiu Han (University of California, Los Angeles), Adnan Darwiche (University of California, Los Angeles)
Graph
🎯 What it does: This paper studies the computational complexity of counterfactual reasoning in structural causal models (SCM) and proves that it is no more difficult than association or intervention reasoning under the tree width and causal tree width frameworks.
On the Fairness Impacts of Private Ensembles Models
Cuong Tran (Syracuse University), Ferdinando Fioretto (University of Virginia)
Safty and PrivacyImageTabular
🎯 What it does: Investigated the impact of the PATE privacy-integrated model on fairness across different groups, analyzed the differential effects caused by algorithm parameters and data features, and proposed a soft label mitigation scheme.
On the Paradox of Learning to Reason from Data
Honghua Zhang (University of California, Los Angeles), Guy Van den Broeck (University of California, Los Angeles)
Data-Centric LearningTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: This paper trains BERT (and T5) models in a restricted logical reasoning problem space (SimpleLogic) and evaluates their reasoning ability across different data distributions; it finds that models achieve nearly 100% accuracy on training distributions but perform extremely poorly on other distributions, indicating that models have not truly learned reasoning but instead exploit statistical features in the data.
On the Reuse Bias in Off-Policy Reinforcement Learning
Chengyang Ying (Tsinghua University), Jun Zhu (Tsinghua University)
Reinforcement Learning
🎯 What it does: This paper studies the estimation bias caused by reusing data in offline replay buffers and introduces the concept of Reuse Bias.
On the Role of Memory in Robust Opinion Dynamics
Luca Becchetti (Sapienza University of Rome), Robin Vacus (CNRS)
OptimizationComputational Efficiency
🎯 What it does: Analyze the limits of memoryless opinion dynamics in information dissemination, proving that the voting model alone can converge in expected time O(n² log n), and subsequently demonstrate that introducing limited memory dynamics can reduce convergence time to nearly linear.
On the Study of Curriculum Learning for Inferring Dispatching Policies on the Job Shop Scheduling
Zangir Iklassov (Mohamed bin Zayed University of Artificial Intelligence), Martin Takac (Mohamed bin Zayed University of Artificial Intelligence)
OptimizationRecurrent Neural NetworkTransformerReinforcement LearningTabularBenchmark
🎯 What it does: This paper proposes a deep adaptive scheduling strategy based on reinforcement learning, enhancing its performance on large-scale job shop scheduling (JSP) through curriculum learning.
On Translations between ML Models for XAI Purposes
Alexis de Colnet (Technical University of Vienna), Pierre Marquis (University of Artois)
Explainability and Interpretability
🎯 What it does: This paper studies the translatability between various machine learning models (decision trees, random forests, boosted trees, binary neural networks, Boolean multi-layer perceptrons, etc.) under polynomial time and polynomial space, aiming to construct a transferability map for XAI.
One Model for All Domains: Collaborative Domain-Prefix Tuning for Cross-Domain NER
Xiang Chen (Zhejiang University), Huajun Chen (Zhejiang University)
RecognitionDomain AdaptationTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Propose a cross-domain named entity recognition (NER) method, CP-NER, which redefines NER as a text-to-text generation task and adapts to different domains by freezing a pre-trained language model (T5) combined with collaborative domain prefix fine-tuning.
One Model, Any CSP: Graph Neural Networks as Fast Global Search Heuristics for Constraint Satisfaction
Jan Tönshoff (RWTH Aachen University), Martin Grohe (RWTH Aachen University)
OptimizationGraph Neural NetworkReinforcement LearningGraph
🎯 What it does: Designed a general graph neural network architecture ANYCSP, which can be unsupervised trained to become a search heuristic for any constraint satisfaction problem (CSP).
Online Harmonizing Gradient Descent for Imbalanced Data Streams One-Pass Classification
Han Zhou (Chongqing University), Yuyu Huang (Chongqing University)
ClassificationOptimizationTabularSequential
🎯 What it does: Propose an online harmonized gradient descent (OHGD) algorithm for one-pass streaming data class imbalance classification;
Online Task Assignment with Controllable Processing Time
Ruoyu Wu (University of Sydney), Liming Ge (University of Sydney)
Optimization
🎯 What it does: Proposed an online task allocation model with controllable processing time and designed the OMLA (Online Machine and Level Assignment) algorithm;
Open Anomalous Trajectory Recognition via Probabilistic Metric Learning
Qiang Gao (Southwestern University of Finance and Economics), Fan Zhou (University of Electronic Science and Technology of China)
Anomaly DetectionKnowledge DistillationAuto EncoderContrastive LearningTime SeriesSequential
🎯 What it does: Propose the open abnormal trajectory identification task and design the ATROM model
Open-world Semi-supervised Novel Class Discovery
Jiaming Liu (University of Electronic Science and Technology of China), Junming Shao (University of Electronic Science and Technology of China)
ClassificationContrastive LearningImage
🎯 What it does: Propose a semi-supervised open-world new category discovery method called OpenNCD, which can simultaneously identify known categories and discover unknown categories in unlabeled data.
OptIForest: Optimal Isolation Forest for Anomaly Detection
Haolong Xiang (Macquarie University), Xiaolong Xu (Nanjing University of Information Science and Technology)
Anomaly DetectionTabularBenchmark
🎯 What it does: Proposed OptIForest, combining theoretical analysis and practical implementation, which achieves efficient and robust anomaly detection through the optimal isolation tree design based on the best branching factor e
Optimal Anytime Coalition Structure Generation Utilizing Compact Solution Space Representation
Redha Taguelmimt (Univ Lyon), Tuomas Sandholm (Carnegie Mellon University)
OptimizationComputational EfficiencyGraph
🎯 What it does: Proposed the CSS graph and hybrid algorithm ELIXIR based on compressed search space representation to efficiently solve optimal joint structure generation problems.
Optimal Decision Tree Policies for Markov Decision Processes
Daniël Vos, Sicco Verwer (Delft University of Technology)
Reinforcement LearningTabularBenchmark
🎯 What it does: Proposes OMDT, a framework based on mixed-integer linear programming (MILP), which directly solves the optimal decision tree strategy for Markov decision processes (MDPs) under given tree size constraints.
Optimal Decision Trees For Interpretable Clustering with Constraints
Pouya Shati (University of Toronto), Sheila McIlraith (University of Toronto)
OptimizationExplainability and InterpretabilityTabularBenchmark
🎯 What it does: Proposed an interpretable constrained clustering method based on SAT, using decision trees to achieve explainable clustering while satisfying must-link/cannot-link constraints;
Optimal Seat Arrangement: What Are the Hard and Easy Cases?
Esra Ceylan (TU Wien), Sanjukta Roy (Pennsylvania State University)
OptimizationGraph
🎯 What it does: This paper conducts a systematic complexity analysis of four optimal seating arrangement problems based on seat maps (maximum welfare, maximum minimum utility, envy-free arrangement, exchange-stable arrangement).
Ordinal Hedonic Seat Arrangement under Restricted Preference Domains: Swap Stability and Popularity
Anaëlle Wilczynski (CentraleSupélec)
Optimization
🎯 What it does: Studied the ordinal-type Hedonic Seat Arrangement problem, focusing on the existence and construction of two types of solutions: swap-stable and popular. The feasibility under different preference extensions (Fishburn, Gardenfors, Responsive Set) and graph structures (paths, cycles, clusters) was analyzed, along with polynomial-time construction algorithms and complexity results.
Orientation-Independent Chinese Text Recognition in Scene Images
Haiyang Yu (Fudan University), Xiangyang Xue (Fudan University)
RecognitionConvolutional Neural NetworkTransformerAuto EncoderImageBenchmark
🎯 What it does: A method for direction-agnostic recognition of Chinese text in scenes is studied by separating content and directional information.
Orion: Online Backdoor Sample Detection via Evolution Deviance
Huayang Huang (Wuhan University), Tao Wang (Wuhan University)
Anomaly DetectionAdversarial AttackImage
🎯 What it does: Construct a multi-exit side network (S-Net) to monitor forward prediction evolution, online identify backdoor samples, and recover original labels.
OSDP: Optimal Sharded Data Parallel for Distributed Deep Learning
Youhe Jiang (Peking University), Bin Cui (Peking University)
OptimizationComputational Efficiency
🎯 What it does: OSDP automatically generates distributed training plans, combining data parallelism and model parallelism to optimize memory usage and overall throughput.
OSP2B: One-Stage Point-to-Box Network for 3D Siamese Tracking
Jiahao Nie (Hangzhou Dianzi University), Jing Zhang (University of Sydney)
Object TrackingAutonomous DrivingGraph Neural NetworkTransformerPoint Cloud
🎯 What it does: Proposed a one-stage point-to-box network, OSP2B, for 3D LiDAR single-target tracking.
Outsourcing Adjudication to Strategic Jurors
Ioannis Caragiannis (Aarhus University), Nikolaj Schwartzbach (Aarhus University)
Optimization
🎯 What it does: This paper proposes a decentralized arbitration mechanism based on majority voting and payment functions, and designs an incentive scheme that minimizes payments through equilibrium analysis and linear programming.
Overlooked Implications of the Reconstruction Loss for VAE Disentanglement
Nathan Michlo (University of Witwatersrand), Steven James (University of Witwatersrand)
GenerationRepresentation LearningConvolutional Neural NetworkAuto EncoderImage
🎯 What it does: This paper analyzes the impact of reconstruction loss on perceptual data distance when VAEs learn separable representations. It constructs an adversarial dataset to demonstrate the non-separability caused by traditional pixel-level loss, and subsequently proposes a spatially aware reconstruction loss to restore separability.
Parameterized Local Search for Max c-Cut
Jaroslav Garvardt (Friedrich Schiller University Jena), Nils Morawietz (Friedrich Schiller University Jena)
OptimizationGraphBenchmark
🎯 What it does: This paper investigates the use of parameterized local search in the maximum c-cut problem, proving that LS MAX c-CUT is W[1]-hard under the k parameterization, and presents an algorithm with a runtime of O((3eΔ)^k · c · k³ · Δ · n); subsequently, this algorithm is incorporated as a post-processing step into the existing MOH approximation algorithm, significantly improving the solution quality on multiple benchmark instances.
Parametrized Gradual Semantics Dealing with Varied Degrees of Compensation
Dragan Doder (Utrecht University), Srdjan Vesic (University of Artois)
Graph
🎯 What it does: A family of parameterized progressive semantics is proposed to evaluate argument strength in weighted argumentation graphs, achieving control over different levels of compensation through parameter α, thereby unifying and extending existing semantics.
Part Aware Contrastive Learning for Self-Supervised Action Recognition
Yilei Hua (Wuhan University of Science and Technology), Shiqian Wu (University of North Carolina at Charlotte)
RecognitionRecurrent Neural NetworkGraph Neural NetworkTransformerContrastive LearningGraph
🎯 What it does: Designed and implemented a self-supervised contrastive learning framework called SkeAttnCLR for 3D skeleton action recognition, enhancing feature representation by integrating global and local contrastive learning.
Participatory Budgeting with Multiple Degrees of Projects and Ranged Approval Votes
Gogulapati Sreedurga (Indian Institute of Science)
Optimization
🎯 What it does: This paper introduces multiple feasible costs (multi-tiered projects) into the participatory budgeting framework and defines a voting mode called 'interval acceptance voting,' where voters specify acceptable cost intervals for each project. Subsequently, four utility functions based on this voting mode are proposed and analyzed (Carleton utility, cost utility, cost upper bound utility, and distance disutility), along with their corresponding feasible solution rules.
Participatory Budgeting: Data, Tools and Analysis
Piotr Faliszewski (AGH University), Nimrod Talmon (Ben Gurion University)
OptimizationTabularBenchmark
🎯 What it does: This study constructs a data and tool library for participatory budgeting (PB), including Pabulib, Pabutools, and Pabustats, and conducts large-scale experiments on over 600 real PB cases in Polish cities, comparing the fairness and efficiency of different voting rules;
PasCore: A Chinese Overlapping Relation Extraction Model Based on Global Pointer Annotation Strategy
Peng Wang (Southeast University), Wei Li (Beijing Institute of Computer Technology and Application)
TransformerLarge Language ModelText
🎯 What it does: Proposes PasCore, a Chinese overlapping relation extraction model based on a global pointer annotation strategy, which sequentially performs three stages: relation prediction, head entity annotation, and tail entity annotation.
PathLAD+: An Improved Exact Algorithm for Subgraph Isomorphism Problem
Yiyuan Wang (Northeast Normal University), Qingwei Lin (Microsoft Research)
OptimizationComputational EfficiencyImageMesh
🎯 What it does: Propose an improved exact algorithm called PathLAD+, which addresses the subgraph isomorphism problem through three novel heuristics: probe search, matching sorting based on probe information, and adaptive propagation.
PED-ANOVA: Efficiently Quantifying Hyperparameter Importance in Arbitrary Subspaces
Shuhei Watanabe (University of Freiburg), Frank Hutter (University of Freiburg)
Hyperparameter SearchImageBenchmark
🎯 What it does: Propose a fast local f-ANOVA method based on Pearson divergence (PED-ANOVA), which can efficiently compute hyperparameter importance in any subspace (local space);
Physics-Guided Human Motion Capture with Pose Probability Modeling
Jingyi Ju (Southeast University), Yangang Wang (Southeast University)
Pose EstimationDiffusion modelAuto EncoderImagePhysics Related
🎯 What it does: This paper proposes a physics-guided human motion capture method based on reverse diffusion, using the latent Gaussian distribution generated by VAE as the initial value, and gradually correcting the posture by introducing gradients of physical simulation and image projection error during the diffusion process, thereby obtaining three-dimensional actions that comply with both visual and physical constraints.
Poisoning the Well: Can We Simultaneously Attack a Group of Learning Agents?
Ridhima Bector (Nanyang Technological University), Zinovi Rabinovich (Nanyang Technological University)
Adversarial AttackReinforcement LearningAuto Encoder
🎯 What it does: Propose a collective environment poisoning (CEP) attack framework that poisons the environment during training of multi-agent learners, capable of simultaneously controlling the entire group of learners in black-box/ultra-black-box settings without internal information.
PowerBEV: A Powerful Yet Lightweight Framework for Instance Prediction in Bird’s-Eye View
Peizheng Li (Mercedes-Benz AG), Juergen Gall (University of Bonn)
Object DetectionSegmentationAutonomous DrivingConvolutional Neural NetworkFlow-based ModelOptical FlowImage
🎯 What it does: Proposed an end-to-end BEV prediction framework called PowerBEV, achieving instance prediction through only two outputs: semantic segmentation and backward clustering flow.
PPAT: Progressive Graph Pairwise Attention Network for Event Causality Identification
Zhenyu Liu (Harbin Institute of Technology), Min Zhang (Harbin Institute of Technology)
ClassificationGraph Neural NetworkTransformerText
🎯 What it does: Proposed the Progressive Graph Pairwise Attention Network (PPAT) for event causality recognition.
Prediction with Incomplete Data under Agnostic Mask Distribution Shift
Yichen Zhu (Shanghai Jiao Tong University), Chenghu Zhou (Shanghai Jiao Tong University)
Data-Centric LearningImageTabularTime Series
🎯 What it does: Propose the StableMiss method, which can predict under missing data and achieve robust generalization while keeping the mask distribution unchanged.
Preferences and Constraints in Abstract Argumentation
Gianvincenzo Alfano (University of Calabria), Irina Trubitsyna (University of Calabria)
🎯 What it does: Based on the original Dung abstract argumentation framework, an extension capable of expressing 'truth value hierarchy' preferences (i.e., preferences related to argument states) is proposed, which is combined with 3-value constraints (e.g., ϕ⇒v or v⇒ϕ) to form a new ePCAF framework; further considering multi-agent scenarios, the definition and reasoning problems of multi-agent ePCAF (mPCAF) are provided.
Principal-Agent Boolean Games
David Hyland (University of Oxford), Michael Wooldridge (University of Oxford)
Agentic AI
🎯 What it does: Proposes the multi-agent moral hazard (principal-agent) problem within the Boolean game framework, and provides formal definitions for contract design and verification.
Privacy-Preserving End-to-End Spoken Language Understanding
Yinggui Wang (Ant Group), Le Yang (University Of Canterbury)
Safty and PrivacyTransformerAudio
🎯 What it does: Developed an end-to-end privacy-preserving framework for speech language understanding (SLU), utilizing hidden layer separation and adversarial training to prevent speech recognition (ASR) and identity recognition (IR) attacks.
Probabilistic Masked Attention Networks for Explainable Sequential Recommendation
Huiyuan Chen (Visa Research), Hao Yang (Visa Research)
Recommendation SystemExplainability and InterpretabilityTransformerSequential
🎯 What it does: This paper proposes a probabilistic mask attention network (PMAN), which introduces learnable binary masks into the self-attention layer of Transformers to sparsify the attention distribution, thereby filtering out noise interactions in sequences and improving the accuracy and interpretability of sequential recommendations.
Probabilistic Planning with Prioritized Preferences over Temporal Logic Objectives
Lening Li (Worcester Polytechnic Institute), Jie Fu (University of Florida)
Robotic IntelligenceReinforcement Learning
🎯 What it does: This paper studies time planning in probabilistic environments based on user preferences, proposing a new specification language called Priority Qualitative Choice Linear Time Logic for Finite (PQCLTLf) to concisely specify temporal goals and their corresponding preferences.
Probabilistic Rule Induction from Event Sequences with Logical Summary Markov Models
Debarun Bhattacharjya (IBM Research), Keerthiram Murugesan (IBM Research)
Explainability and InterpretabilitySequential
🎯 What it does: Propose the Logical Summary Markov Model (LSuMM) and implement two novel models (Count SuMM and Shared Sequential SuMM) for interpretable event sequence prediction; design a greedy search learning algorithm that generates probabilistic logic rules using historical predicates.
Probabilistic Temporal Logic for Reasoning about Bounded Policies
Nima Motamed (Utrecht University), Brian Logan (Utrecht University)
Reinforcement Learning
🎯 What it does: This paper proposes Probabilistic Logic of Bounded Policies (PLBP), a novel probabilistic temporal logic for describing finite-step decision strategies and their uncertain outcomes, with semantic interpretations on Markov Decision Processes (MDPs).
Progressive Label Propagation for Semi-Supervised Multi-Dimensional Classification
Teng Huang (Southeast University), Min-Ling Zhang (Southeast University)
ClassificationImageTextMultimodalityBiomedical Data
🎯 What it does: This paper proposes a semi-supervised method for multi-dimensional classification using a small number of labeled samples and a large number of unlabeled samples, called PLAP (Progressive Label Propagation)
ProMix: Combating Label Noise via Maximizing Clean Sample Utility
Ruixuan Xiao (Zhejiang University), Junbo Zhao (Zhejiang University)
ClassificationData-Centric LearningImage
🎯 What it does: ProMix systematically improves the utilization of clean samples when learning from noisy labels through advanced sample selection and debiased semi-supervised learning;
Prompt Federated Learning for Weather Forecasting: Toward Foundation Models on Meteorological Data
Shengchao Chen (University of Technology Sydney), Jing Jiang (University of Technology Sydney)
Federated LearningGraph Neural NetworkTransformerPrompt EngineeringTime Series
🎯 What it does: Proposed MetePFL, a framework for weather prediction on multi-region heterogeneous meteorological time series data, combining a pre-trained Transformer base model with federated Prompt learning;
Prompt Learns Prompt: Exploring Knowledge-Aware Generative Prompt Collaboration For Video Captioning
Liqi Yan (Fudan University), Qifan Wang (Meta)
GenerationTransformerPrompt EngineeringVision Language ModelVideoTextMultimodality
🎯 What it does: Proposed a two-stage video captioning method based on knowledge-aware prompting, first using a pre-trained knowledge prompt to extract key actions and objects, then generating complete captions through a frozen main model and a small number of learnable video-language prompts
Proportionality Guarantees in Elections with Interdependent Issues
Markus Brill (University of Warwick), Jannik Peters (TU Berlin)
Optimization
🎯 What it does: This paper conducts theoretical analysis of proportional representation under a multi-issue, interdependent conditional election framework, and extends two classic voting rules, PAV and MES, to accommodate non-binary domains and conditional voting.
Proportionally Fair Online Allocation of Public Goods with Predictions
Siddhartha Banerjee (Cornell University), Nisarg Shah (University of Toronto)
Optimization
🎯 What it does: This paper studies improving fairness allocation algorithms in the online public goods allocation problem by leveraging predictions of each participant's future total value;
Pseudo-Labeling Enhanced by Privileged Information and Its Application to In Situ Sequencing Images
Marzieh Haghighi (Broad Institute of MIT and Harvard), Shantanu Singh (Broad Institute of MIT and Harvard)
Object DetectionConvolutional Neural NetworkVision Language ModelImageBiomedical DataBenchmark
🎯 What it does: Propose a semi-supervised object detection framework that integrates privileged information into pseudo labels, applying it to the barcode decoding task of In-Situ Sequencing (ISS) images in spatial transcriptomics, forming the PLePI-ISS two-stage SSOD (Semi-Supervised Object Detection) model.
pTSE: A Multi-model Ensemble Method for Probabilistic Time Series Forecasting
Yunyi Zhou (Ant Group), Sheng Li (University of Virginia)
Time Series
🎯 What it does: Propose a multi-model probabilistic time series ensemble method called pTSE based on the hidden Markov model.
Pyramid Diffusion Models for Low-light Image Enhancement
Dewei Zhou (Zhejiang University), Yi Yang (Zhejiang University)
RestorationDiffusion modelImage
🎯 What it does: Propose a low-light image enhancement method called PyDiff based on diffusion models, achieving faster and higher quality restoration through pyramid diffusion and a global corrector.
Quantifying Consistency and Information Loss for Causal Abstraction Learning
Fabio Massimo Zennaro (University of Warwick), Theodoros Damoulas (University of Warwick)
Explainability and InterpretabilityRepresentation LearningGraphBiomedical Data
🎯 What it does: This paper proposes an intervention-based abstract approximation metric, combining consistency (IC) with information loss (IIL, ISIL, ISC) to quantify differences between causal models at different hierarchical structural levels, and provides corresponding evaluation and learning algorithms;
Quantifying Harm
Sander Beckers (University of Amsterdam), Joseph Y. Halpern (Cornell University)
Safty and PrivacyExplainability and Interpretability
🎯 What it does: This paper proposes a quantitative harm measurement method based on a causal model, extending previous qualitative harm concepts and introducing factors such as uncertainty, probability weighting, and group fairness to calculate harm at individual and societal levels.
Quantitative Reasoning and Structural Complexity for Claim-Centric Argumentation
Johannes K. Fichte (Linkoping University), Arne Meier (Leibniz Universitat Hannover)
Computational Efficiency
🎯 What it does: This paper proposes a declarative quantified reasoning framework that counts the number of extended claims to achieve fine-grained reasoning between credible reasoning and strict reasoning.
Quick Multi-Robot Motion Planning by Combining Sampling and Search
Keisuke Okumura (National Institute of Advanced Industrial Science and Technology), Xavier Défago (Tokyo Institute of Technology)
Robotic Intelligence
🎯 What it does: Propose a new multi-robot motion planning algorithm called SSSP, which can simultaneously build individual roadmaps for each robot and find collision-free paths in a single search round.
RAIN: RegulArization on Input and Network for Black-Box Domain Adaptation
Qucheng Peng (University of Central Florida), Chen Chen (University of Central Florida)
Domain AdaptationKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: Propose the RAIN method in the black-box domain adaptation scenario, combining regularization at the input layer and network layer to enhance target domain performance.
Rainbow Cycle Number and EFX Allocations: (Almost) Closing the Gap
Shayan Chashm Jahan (Sharif University of Technology), Mohammad Sharifi (Sharif University of Technology)
Optimization
🎯 What it does: This paper studies the fair allocation problem of indivisible items, establishes the connection between rainbow cycle numbers and EFX (almost envy-free) allocation, and improves the upper bound of rainbow cycle numbers by introducing a new extremal combinatorial problem.
RaMLP: Vision MLP via Region-aware Mixing
Shenqi Lai (InsightFace.ai), Kaipeng Zhang (Shanghai AI Laboratory)
ClassificationObject DetectionSegmentationImage
🎯 What it does: Designed a Region-aware MLP (RaMLP) visual backbone, achieving efficient visual representations with variable sizes through the Region-aware Mixing (RaM) module, directly applicable to image classification and dense prediction tasks.
Random Assignment of Indivisible Goods under Constraints
Yasushi Kawase (University of Tokyo), Yu Yokoi (Tokyo Institute of Technology)
Optimization
🎯 What it does: This paper studies the feasibility of randomly allocating indivisible items under various constraints, exploring the scenarios in which random allocations can simultaneously satisfy sequential efficiency (sd-efficiency) and sequential envy-freedom (sd-envy-freedom).
Ranking-based Argumentation Semantics Applied to Logical Argumentation
Jesse Heyninck (Open Universiteit), Christian Straßer (Ruhr-Universitat Bochum)
🎯 What it does: This paper systematically studies the application of ranking-based semantics to structured argumentation frameworks, proving that these semantics can be viewed as blame metrics, and further evaluates their robustness against different attack forms and construction methods.
RaSa: Relation and Sensitivity Aware Representation Learning for Text-based Person Search
Yang Bai (Soochow University), Min Zhang (Soochow University)
RetrievalRepresentation LearningTransformerVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: Propose a representation learning method RaSa for person re-identification in text retrieval, which includes two tasks: relation-aware learning (RA) and sensitivity-aware learning (SA).
Realistic Cell Type Annotation and Discovery for Single-cell RNA-seq Data
Yuyao Zhai (Peking University), Minghua Deng (Peking University)
ClassificationAuto EncoderBiomedical Data
🎯 What it does: Proposed a new task — achieving realistic cell type annotation and discovery in single-cell RNA sequencing data, and designed an end-to-end algorithm scPOT to simultaneously identify known cell types and cluster unknown cell types
Recognizable Information Bottleneck
Yilin Lyu (Beijing Jiaotong University), Liping Jing (Beijing Jiaotong University)
Representation LearningImage
🎯 What it does: The paper proposes the Recognizable Information Bottleneck (RIB), which enhances model generalization performance by constraining the recognizability of representations.
Reconstruction-Aware Prior Distillation for Semi-supervised Point Cloud Completion
Zhaoxin Fan (Renmin University of China), Jun He (Renmin University of China)
RestorationAuto EncoderGenerative Adversarial NetworkPoint Cloud
🎯 What it does: Propose a semi-supervised point cloud completion framework named RaPD, which learns semantic priors using a large number of unpaired complete/incomplete point clouds and trains the completion network with only a small number of paired samples.
Recursive Small-Step Multi-Agent A* for Dec-POMDPs
Wietze Koops (Radboud University), Thiago D. Simão (Radboud University)
OptimizationReinforcement LearningBenchmark
🎯 What it does: Proposed a recursive small-step multi-agent A* (RS-MAA*) algorithm for solving optimal policies in finite-horizon Dec-POMDPs.
Reducing Communication for Split Learning by Randomized Top-k Sparsification
Fei Zheng (Zhejiang University), Binhui Yao (Midea Group)
Federated LearningConvolutional Neural NetworkRecurrent Neural NetworkImageTextSequential
🎯 What it does: Proposes a method to reduce communication in segmentation learning through random Topk sparsification, balancing training convergence and generalization.
Regularisation for Efficient Softmax Parameter Generation in Low-Resource Text Classifiers
Daniel Grießhaber (Stuttgart Media University), Ngoc Thang Vu (University of Stuttgart)
ClassificationMeta LearningText
🎯 What it does: This study improves the performance of meta-learning models in low-resource text classification by generating new tasks and applying regularization.
Reinforcement Learning Approaches for Traffic Signal Control under Missing Data
Hao Mei (New Jersey Institute of Technology), Hua Wei (New Jersey Institute of Technology)
OptimizationReinforcement Learning
🎯 What it does: Propose a method for traffic signal control in urban road networks with missing sensor data using reinforcement learning, and design two solutions based on data imputation;
Relation-enhanced DETR for Component Detection in Graphic Design Reverse Engineering
Xixuan Hao (University of Hong Kong), Chin-Yew Lin (Microsoft Research)
Object DetectionTransformerImage
🎯 What it does: This paper addresses reverse engineering in graphic design, studying the component detection task;
Relative Inconsistency Measures for Indefinite Databases with Denial Constraints
Francesco Parisi (University of Calabria), John Grant (University of Maryland)
Tabular
🎯 What it does: This paper proposes and studies relative inconsistency measures on infinite databases with negation constraints.
ReLiNet: Stable and Explainable Multistep Prediction with Recurrent Linear Parameter Varying Networks
Alexandra Baier (Analytic Computing, University of Stuttgart), Steffen Staab (Analytic Computing, University of Stuttgart)
Explainability and InterpretabilityRecurrent Neural NetworkTime SeriesSequential
🎯 What it does: Propose the ReLiNet framework, which maps RNN-generated hidden states to a time-varying linear parameter system, achieving interpretability and stability for multi-step prediction;
RePaint-NeRF: NeRF Editting via Semantic Masks and Diffusion Models
Xingchen Zhou (Shenzhen University), You Li (Guangdong Laboratory of Artificial Intelligence and Digital Economy)
Image HarmonizationGenerationDepth EstimationDiffusion modelNeural Radiance FieldImage
🎯 What it does: Propose the RePaint-NeRF framework, which edits NeRF scenes using semantic masks and diffusion models.
REPLACE: A Logical Framework for Combining Collective Entity Resolution and Repairing
Meghyn Bienvenu (University of Bordeaux), Víctor Gutiérrez-Basulto (Cardiff University)
OptimizationData-Centric Learning
🎯 What it does: Proposed a new unified framework called REPLACE for simultaneously performing collective entity resolution and data repairing, along with providing complete logical semantics and optimal solution determination;
Revenue Maximization Mechanisms for an Uninformed Mediator with Communication Abilities
Zhikang Fan (Renmin University of China), Weiran Shen (Renmin University of China)
Finance Related
🎯 What it does: The study introduces a mediator between sellers and buyers, designing a communication mechanism that can maximize the mediator's revenue.
Reverse Engineering of Temporal Queries Mediated by LTL Ontologies
Marie Fortin (Universit' e Paris Cit'), Michael Zakharyaschev (Birkbeck, University of London)
Sequential
🎯 What it does: This paper studies the reverse engineering (query-by-example) problem for linear temporal logic (LTL) queries combined with LTL ontologies, systematically analyzing the separability and complexity of different query languages (such as path/branch queries containing only U, *, ◇) under both ontology-free and ontology-informed scenarios.
Revisiting the Evaluation of Deep Learning-Based Compiler Testing
Yongqiang Tian (University of Waterloo), Shing-Chi Cheung (Hong Kong University of Science and Technology)
TextBenchmark
🎯 What it does: This paper proposes Kitten, a language-agnostic, mutation-based program generator designed to fairly evaluate deep learning-driven compiler testing tools.