Daily Arxiv

전 세계에서 발간되는 인공지능 관련 논문을 정리하는 페이지 입니다.
본 페이지는 Google Gemini를 활용해 요약 정리하며, 비영리로 운영 됩니다.
논문에 대한 저작권은 저자 및 해당 기관에 있으며, 공유 시 출처만 명기하면 됩니다.

TabTune: A Unified Library for Inference and Fine-Tuning Tabular Foundation Models

Scalable Evaluation and Neural Models for Compositional Generalization

In Situ Training of Implicit Neural Compressors for Scientific Simulations via Sketch-Based Regularization

LA-MARRVEL: A Knowledge-Grounded and Language-Aware LLM Reranker for AI-MARRVEL in Rare Disease Diagnosis

CudaForge: An Agent Framework with Hardware Feedback for CUDA Kernel Optimization

Reg-DPO: SFT-Regularized Direct Preference Optimization with GT-Pair for Improving Video Generation

AI for Requirements Engineering: Industry adoption and Practitioner perspectives

HAFixAgent: History-Aware Automated Program Repair Agent

Quantifying truth and authenticity in AI-assisted candidate evaluation: A multi-domain pilot analysis

On Improvisation and Open-Endedness: Insights for Experiential AI

Beyond Synthetic Benchmarks: Evaluating LLM Performance on Real-World Class-Level Code Generation

WOD-E2E: Waymo Open Dataset for End-to-End Driving in Challenging Long-tail Scenarios

Using latent representations to link disjoint longitudinal data for mixed-effects regression

LLMComp: A Language Modeling Paradigm for Error-Bounded Scientific Data Compression (Technical Report)

Toward Humanoid Brain-Body Co-design: Joint Optimization of Control and Morphology for Fall Recovery

Enabling Robust In-Context Memory and Rapid Task Adaptation in Transformers with Hebbian and Gradient-Based Plasticity

The Mirror Loop: Recursive Non-Convergence in Generative Reasoning Systems

A Foundational Theory of Quantitative Abstraction: Adjunctions, Duality, and Logic for Probabilistic Systems

AgenticMath: Enhancing LLM Reasoning via Agentic-based Math Data Generation

StutterZero and StutterFormer: End-to-End Speech Conversion for Stuttering Transcription and Correction

ADPO: Anchored Direct Preference Optimization

A Survey on Collaborating Small and Large Language Models for Performance, Cost-effectiveness, Cloud-edge Privacy, and Trustworthiness

FaStfact: Faster, Stronger Long-Form Factuality Evaluations in LLMs

DE3S: Dual-Enhanced Soft-Sparse-Shape Learning for Medical Early Time-Series Classification

Efficient Latent Variable Causal Discovery: Combining Score Search and Targeted Testing

Training Optimal Large Diffusion Language Models

CardioForest: An Explainable Ensemble Learning Model for Automatic Wide QRS Complex Tachycardia Diagnosis from ECG

Automatic Discovery of One-Parameter Subgroups of Lie Groups: Compact and Non-Compact Cases of $\mathbf{SO(n)}$ and $\mathbf{SL(n)}$

A Unified Formal Theory on the Logical Limits of Symbol Grounding

Evaluating Large Language Models for Detecting Antisemitism

SME-TEAM: Leveraging Trust and Ethics for Secure and Responsible Use of AI and LLMs in SMEs

GDS Agent for Graph Algorithmic Reasoning

Activation Transport Operators

"Accessibility people, you go work on that thing of yours over there": Addressing Disability Inclusion in AI Product Organizations

AnalogSeeker: An Open-source Foundation Language Model for Analog Circuit Design

Automated Segmentation of Coronal Brain Tissue Slabs for 3D Neuropathology

Geometry-Aware Global Feature Aggregation for Real-Time Indirect Illumination

Fast weight programming and linear transformers: from machine learning to neurobiology

Voost: A Unified and Scalable Diffusion Transformer for Bidirectional Virtual Try-On and Try-Off

ViFP: A Framework for Visual False Positive Detection to Enhance Reasoning Reliability in VLMs

CoTox: Chain-of-Thought-Based Molecular Toxicity Reasoning and Prediction

Decentralized Aerial Manipulation of a Cable-Suspended Load using Multi-Agent Reinforcement Learning

TensorHyper-VQC: A Tensor-Train-Guided Hypernetwork for Robust and Scalable Variational Quantum Computing

PhysicsEval: Inference-Time Techniques to Improve the Reasoning Proficiency of Large Language Models on Physics Problems

Enhancing Fatigue Detection through Heterogeneous Multi-Source Data Integration and Cross-Domain Modality Imputation

Automatic Road Subsurface Distress Recognition from Ground Penetrating Radar Images using Deep Learning-based Cross-verification

Flow matching for reaction pathway generation

DiffSpectra: Molecular Structure Elucidation from Spectra using Diffusion Models

Omni-Router: Sharing Routing Decisions in Sparse Mixture-of-Experts for Speech Recognition

FedRef: Communication-Efficient Bayesian Fine-Tuning using a Reference Model

Layer Importance for Mathematical Reasoning is Forged in Pre-Training and Invariant after Post-Training

Benchmarking Foundation Models and Parameter-Efficient Fine-Tuning for Prognosis Prediction in Medical Imaging

Dense SAE Latents Are Features, Not Bugs

AlphaDecay: Module-wise Weight Decay for Heavy-Tailed Balancing in LLMs

Balancing Tails when Comparing Distributions: Comprehensive Equity Index (CEI) with Application to Bias Evaluation in Operational Face Biometrics

SLED: A Speculative LLM Decoding Framework for Efficient Edge Serving

R2R: Efficiently Navigating Divergent Reasoning Paths with Small-Large Model Token Routing

Large Language Models Miss the Multi-Agent Mark

Distilling LLM Agent into Small Models with Retrieval and Code Tools

This Time is Different: An Observability Perspective on Time Series Foundation Models

RoboRAN: A Unified Robotics Framework for Reinforcement Learning-Based Autonomous Navigation

Traversal Verification for Speculative Tree Decoding

Deep Learning Warm Starts for Trajectory Optimization on the International Space Station

A data-driven framework for team selection in Fantasy Premier League

SecRepoBench: Benchmarking Code Agents for Secure Code Completion in Real-World Repositories

Revisiting semi-supervised learning in the era of foundation models

Assessing the Macro and Micro Effects of Random Seeds on Fine-Tuning Large Language Models

A Survey on Text-Driven 360-Degree Panorama Generation

Beyond Covariance Matrix: The Statistical Complexity of Private Linear Regression

From Haystack to Needle: Label Space Reduction for Zero-shot Classification

REFA: Reference Free Alignment for multi-preference optimization

RAG-IT: Retrieval-Augmented Instruction Tuning for Automated Financial Analysis

Do Automatic Factuality Metrics Measure Factuality? A Critical Evaluation

Matryoshka Pilot: Learning to Drive Black-Box LLMs with LLMs

Graph Sampling for Scalable and Expressive Graph Neural Networks on Homophilic Graphs

Intelligent Computing Social Modeling and Methodological Innovations in Political Science in the Era of Large Language Models

Mastering Contact-rich Tasks by Combining Soft and Rigid Robotics with Imitation Learning

Inverse Entropic Optimal Transport Solves Semi-supervised Learning via Data Likelihood Maximization

MSDNet: Multi-Scale Decoder for Few-Shot Semantic Segmentation via Transformer-Guided Prototyping

Autonomous Robotic Drilling System for Mice Cranial Window Creation

A Reliable Cryptographic Framework for Empirical Machine Unlearning Evaluation

A Survey of Graph Neural Networks in Real world: Imbalance, Noise, Privacy and OOD Challenges

Neural Physics: Using AI Libraries to Develop Physics-Based Solvers for Incompressible Computational Fluid Dynamics

Survey on AI Ethics: A Socio-technical Perspective

Transfer Learning-based Real-time Handgun Detection

Emotion Detection From Social Media Posts

Agent-Omni: Test-Time Multimodal Reasoning via Model Coordination for Understanding Anything

Kosmos: An AI Scientist for Autonomous Discovery

The ORCA Benchmark: Evaluating Real-World Calculation Accuracy in Large Language Models

TabDSR: Decompose, Sanitize, and Reason for Complex Numerical Reasoning in Tabular Data

Mirror-Neuron Patterns in AI Alignment

Agentic Meta-Orchestrator for Multi-task Copilots

Misalignment Bounty: Crowdsourcing AI Agent Misbehavior

PlanU: Large Language Model Reasoning through Planning under Uncertainty

VoiceAgentBench: Are Voice Assistants ready for agentic tasks?

ForTIFAI: Fending Off Recursive Training Induced Failure for AI Model Collapse

Reinforcement Learning Foundations for Deep Research Systems: A Survey

Data Dependency-Aware Code Generation from Enhanced UML Sequence Diagrams

LLM-Driven Collaborative Model for Untangling Commits via Explicit and Implicit Dependency Reasoning

Why Isn't Relational Learning Taking Over the World?

JEEM: Vision-Language Understanding in Four Arabic Dialects

Created by

Haebom

저자

Karima Kadaoui, Hanin Atwany, Hamdan Al-Ali, Abdelrahman Mohamed, Ali Mekky, Sergei Tilga, Natalia Fedorova, Ekaterina Artemova, Hanan Aldarmaki, Yova Kementchedjhieva

개요

JEEM은 요르단, 아랍에미리트, 이집트, 모로코 등 4개 아랍어 사용 국가의 문화적 다양성을 반영한 이미지 자막 생성 및 시각적 질문 응답 과제를 포함하는 시각-언어 모델(VLM) 평가 벤치마크입니다. 5개의 주요 오픈소스 아랍어 VLM과 GPT-4V를 평가한 결과, 아랍어 VLM은 시각적 이해와 방언 특징적 생성 모두에서 성능이 저조했으며, GPT-4V가 가장 우수했지만 방언별 언어 능력과 시각적 이해 능력에는 차이가 있음을 보였습니다. 이는 더 포괄적인 모델과 문화적으로 다양한 평가 방식의 필요성을 강조합니다.

시사점, 한계점

•

시사점:

◦

문화적 다양성을 고려한 VLM 평가의 중요성 부각

◦

아랍어 VLM의 성능 개선 필요성 제기

◦

다양한 방언에 대한 VLM의 언어 능력 개선 필요성 제기

◦

더욱 포괄적인 VLM 개발 필요성 강조

•

한계점:

◦

평가에 사용된 VLM의 종류가 제한적일 수 있음

◦

JEEM 벤치마크의 범용성에 대한 추가 연구 필요

◦

GPT-4V의 성능이 우수하나, 여전히 시각적 이해 및 방언별 언어 능력에 한계가 존재

Made with Slashpage