Page 27 - Paper Library - AiPaper

Defining endogenous barcoding sites for CRISPR/Cas9-based cell lineage tracing in zebrafish

Published:2/1/2020

CRISPR/Cas9 Cell Lineage TracingEndogenous Barcoding SitesDevelopmental Biology in ZebrafishApplication of Genomic TechniquesTissue Organization Research

This study defines endogenous barcoding sites for CRISPR/Cas9based cell lineage tracing in zebrafish, utilizing advanced genomic techniques to identify reliable loci. The findings enhance lineage tracing precision, offering important insights for genetic and developmental resear

Align$^3$GR: Unified Multi-Level Alignment for LLM-based Generative Recommendation

Published:11/14/2025

LLM-based Recommendation SystemsMulti-Level Alignment MethodsBehavior Modeling and AlignmentDynamic Preference AdaptationSelf-Play Decision Optimization

Align

^3

GR effectively transforms LLMs into recommendation systems via a unified multilevel alignment approach, introducing dual tokenization, enhanced behavior modeling, and progressive decision optimization. It significantly outperforms stateoftheart metrics.

Multi-Aspect Cross-modal Quantization for Generative Recommendation

Published:11/19/2025

Generative Recommendation SystemsCross-modal QuantizationMultimodal Information IntegrationSemantic IDs LearningRecommendation Datasets

This paper introduces the MACRec model for generative recommendation, integrating multimodal information to improve semantic ID quality. It employs crossmodal quantization to reduce conflict rates and combines implicit and explicit alignments, enhancing the generative model's pe

Mitigating Exposure Bias in Online Learning to Rank Recommendation: A Novel Reward Model for Cascading Bandits

Published:10/20/2024

Mitigation of Exposure Bias in Online Recommendation SystemsLinear Cascading BanditsExposure-Aware Reward ModelFeedback Loop Issue in Recommender SystemsDiverse Recommendation Strategies

This study addresses exposure bias in recommender systems, introducing an ExposureAware reward model for online learningtorank via Linear Cascading Bandits. It adjusts item utility based on position in the list, enhancing exposure fairness while maintaining accuracy, outpacing

Bandits with Ranking Feedback

Published:1/1/2024

Bandits with Ranking FeedbackNo-Regret Algorithm DesignStochastic Reward MechanismBandits in Adversarial SettingsOptimizing Regret Bound

This paper introduces a novel variant of multiarmed bandits that provides ranking feedback instead of numerical rewards, suitable for scenarios like human preferences. It explores noregret algorithms in stochastic and adversarial settings, proving the impossibility of logarithm

Growth-coupled microbial biosynthesis of the animal pigment xanthommatin

Published:11/3/2025

Microbial Biosynthesis StrategyXanthommatin ProductionHeterologous Natural Product Pathway EngineeringBiosynthetic Feedback LoopGrowth-Driven Synthesis in Bacteria

This study introduces a growthcoupled biosynthetic strategy to tackle low initial yields in engineered bacterial pathways, using an excised C1 moiety to drive growth and enhance xanthommatin production in Pseudomonas putida, demonstrating the potential of adaptive laboratory e

016

RL-PINNs: Reinforcement Learning-Driven Adaptive Sampling for Efficient Training of PINNs

Published:4/17/2025

Reinforcement Learning-Driven Adaptive SamplingPhysics-Informed Neural NetworksPartial Differential Equations SolvingSingle-Round Sampling TrainingMarkov Decision Process

This paper introduces RLPINNs, a reinforcement learningdriven adaptive sampling framework that enhances the training efficiency of PhysicsInformed Neural Networks (PINNs) by enabling optimal training point selection in a single round of sampling, significantly improving accura

HyPINO: Multi-Physics Neural Operators via HyperPINNs and the Method of Manufactured Solutions

Published:9/5/2025

Multi-Physics Neural OperatorsZero-Shot GeneralizationSwin Transformer HypernetworkPhysics-Informed Neural NetworksPartial Differential Equations Solving

HyPINO is introduced as a multiphysics neural operator for zeroshot generalization across various PDEs without taskspecific finetuning, combining a Swin Transformer hypernetwork and mixed supervision for improved accuracy in benchmarks.

MscaleFNO: Multi-scale Fourier Neural Operator Learning for Oscillatory Function Spaces

Published:12/28/2024

Multi-Scale Fourier Neural OperatorHigh-Frequency Mapping LearningNonlinear Mapping of Helmholtz EquationSpectral Bias ReductionWave Scattering Problem

This paper introduces MscaleFNO, a multiscale Fourier neural operator that reduces spectral bias in learning mappings between highly oscillatory functions. It shows significant performance improvements in highfrequency wave scattering problems by employing parallel scaled FNOs.

Pearl: A Review-driven Persona-Knowledge Grounded Conversational Recommendation Dataset

Published:8/1/2024

Conversational Recommendation SystemsLLM-based Recommendation DatasetPersonalized Recommendation DatasetUser Preference ModelingKnowledge-Augmented Conversational Recommendation

The PEARL dataset addresses limitations in conversational recommendation systems by providing specific user preferences and explanations. Synthesized from real reviews, it includes over 57k dialogues, enabling more contextually relevant recommendations. Models trained on PEARL ou

LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation

Published:2/6/2020

GNN-based Recommender SystemsLightGCN ModelNeighborhood AggregationUser-Item Interaction GraphImprovement in Collaborative Filtering

This study introduces LightGCN, simplifying Graph Convolutional Networks for recommendations. We found that common features like transformation and nonlinear activation added complexity with little performance gain. LightGCN focuses on neighborhood aggregation, achieving a notabl

Neural Graph Collaborative Filtering

Published:5/20/2019

Neural Graph Collaborative FilteringUser Behavior ModelingGraph-Structured Recommendation SystemsHigh-Order Connectivity ModelingEmbedding Propagation Mechanism

The Neural Graph Collaborative Filtering (NGCF) framework integrates useritem interactions by propagating embeddings on a graph, effectively capturing collaborative signals and demonstrating significant improvements across several benchmark datasets.

以中国数学史为主线的教学设计探索

Teaching Design Focused on Chinese Mathematical History

This study explores incorporating Chinese mathematical history into junior high school teaching, showing a young teacher's process of integrating it into a lesson on the Pythagorean Theorem, enhancing cultural confidence and core competencies.

Asking Clarifying Questions for Preference Elicitation With Large Language Models

Published:10/14/2025

LLM-guided User Preference ElicitationClarifying Question Generation in Generative Recommendation SystDiffusion Model-driven Sequential Question GenerationUser Preference Mining MethodsMulti-Stage User Preference Modeling

This paper presents a novel method using large language models to generate clarifying questions for eliciting user preferences, particularly when user history is limited, significantly improving the model's effectiveness in guiding preferences.

Multi-Peptide: Multimodality Leveraged Language-Graph Learning of Peptide Properties

Published:12/19/2024

Transformer-based Peptide Property PredictionGNNs with Language ModelsMultimodal Learning in BioinformaticsContrastive Loss FrameworkPeptide Property Dataset Evaluation

The study introduces MultiPeptide, combining transformer models with graph neural networks to enhance peptide property prediction. Using a contrastive loss framework, it achieves a stateoftheart 88.057% accuracy in hemolysis prediction, showcasing multimodal learning's potent

M2oE: Multimodal Collaborative Expert Peptide Model

Published:12/3/2024

Multimodal Peptide Prediction ModelExpert Model and Cross-Attention MechanismIntegration of Peptide Structural and Sequence InformationComplex Task Prediction

The M2oE model integrates peptide sequence and structural information with expert models and crossattention mechanisms, significantly enhancing performance in complex task predictions, as demonstrated by experimental results.

Modelling Techniques to Improve the Quality of Food Using Artificial Intelligence

Published:7/28/2021

AI Applications in Food Quality ImprovementStrategies for Food Safety and Quality EnhancementAI Techniques in Food ProcessingAI in Supply Chain OptimizationCost-Effectiveness Analysis of Food Quality Improvement

This review examines AI's applications in enhancing food safety and quality, analyzing various modeling techniques and their costeffectiveness, aimed at informing policymakers to tackle challenges posed by population growth and environmental changes.

KungfuBot: Physics-Based Humanoid Whole-Body Control for Learning Highly-Dynamic Skills

Published:6/15/2025

Humanoid Whole-Body ControlHigh-Dynamic Motion ImitationPhysics-Based Dynamic TrackingBi-Level Optimization FrameworkRobot Skill Learning

This paper presents KungfuBot, a physicsbased humanoid control framework that learns highdynamic human behaviors like Kungfu and dance through multistep motion processing and adaptive tracking, achieving significantly lower tracking errors successfully implemented on a robot.

Flow-GRPO: Training Flow Matching Models via Online RL

Published:5/9/2025

Online Reinforcement Learning Training for Flow Matching ModelsODE to SDE ConversionDenoising Reduction StrategyText-to-Image Generation TasksStatistical Sampling in Generative Models

FlowGRPO integrates online policy gradient RL into flow matching models, enhancing sampling efficiency and output quality in texttoimage generation by converting ODEs to SDEs and employing a denoising reduction strategy, minimizing reward hacking.

Practical Use of ChatGPT in Psychiatry for Treatment Plan and Psychoeducation

Published:11/16/2023

Application of ChatGPT in Mental HealthAI Support in Mental HealthPsychotherapy and Online SupportPsychoeducation and Information DisseminationEthical and Privacy Considerations

This paper explores ChatGPT's applications in psychiatry, highlighting its potential in treatment planning, psychoeducation, selfhelp, and crisis management. It emphasizes that it should not replace professional guidance while addressing ethical concerns.

521 - 540 / 980

Papers