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.

Integrating Generative Artificial Intelligence in ADRD: A Roadmap for Streamlining Diagnosis and Care in Neurodegenerative Diseases

Logits are All We Need to Adapt Closed Models

Oracular Programming: A Modular Foundation for Building LLM-Enabled Software

Learning Traffic Anomalies from Generative Models on Real-Time Observations

BiDepth: A Bidirectional-Depth Neural Network for Spatio-Temporal Prediction

Not all tokens are created equal: Perplexity Attention Weighted Networks for AI generated text detection

Prune 'n Predict: Optimizing LLM Decision-making with Conformal Prediction

Relation-aware Hierarchical Prompt for Open-vocabulary Scene Graph Generation

An Interoperable Machine Learning Pipeline for Pediatric Obesity Risk Estimation

BEExformer: A Fast Inferencing Binarized Transformer with Early Exits

VIVID-10M: A Dataset and Baseline for Versatile and Interactive Video Local Editing

Tiny-Align: Bridging Automatic Speech Recognition and Large Language Model on the Edge

DuSEGO: Dual Second-order Equivariant Graph Ordinary Differential Equation

Knowledge-Augmented Multimodal Clinical Rationale Generation for Disease Diagnosis with Small Language Models

Foundation Model for Composite Microstructures: Reconstruction, Stiffness, and Nonlinear Behavior Prediction

Provably Adaptive Average Reward Reinforcement Learning for Metric Spaces

TabDPT: Scaling Tabular Foundation Models on Real Data

Leveraging Skills from Unlabeled Prior Data for Efficient Online Exploration

UTF:Undertrained Tokens as Fingerprints A Novel Approach to LLM Identification

EVOLvE: Evaluating and Optimizing LLMs For In-Context Exploration

Defense-as-a-Service: Black-box Shielding against Backdoored Graph Models

Task-Agnostic Pre-training and Task-Guided Fine-tuning for Versatile Diffusion Planner

Dataset Distillation-based Hybrid Federated Learning on Non-IID Data

TheraGen: Therapy for Every Generation

SEAL: Towards Safe Autonomous Driving via Skill-Enabled Adversary Learning for Closed-Loop Scenario Generation

Insuring Uninsurable Risks from AI: Government as Insurer of Last Resort

DiPT: Enhancing LLM reasoning through diversified perspective-taking

Political Bias in LLMs: Unaligned Moral Values in Agent-centric Simulations

A Pairwise Comparison Relation-assisted Multi-objective Evolutionary Neural Architecture Search Method with Multi-population Mechanism

Causal Discovery-Driven Change Point Detection in Time Series

Learning Decentralized Multi-Biped Control for Payload Transport

Intuitive Fine-Tuning: Towards Simplifying Alignment into a Single Process

TKAN: Temporal Kolmogorov-Arnold Networks

A Review of Reward Functions for Reinforcement Learning in the context of Autonomous Driving

Integrated Gradient Correlation: a Dataset-wise Attribution Method

Continuous Spiking Graph Neural Networks

An In-depth Evaluation of Large Language Models in Sentence Simplification with Error-based Human Assessment

Pinning "Reflection" on the Agenda: Investigating Reflection in Human-LLM Co-Creation for Creative Coding

Dynamic Spiking Framework for Graph Neural Networks

Bounding the Worst-class Error: A Boosting Approach

CoCo: A Coupled Contrastive Framework for Unsupervised Domain Adaptive Graph Classification

An Epistemic and Aleatoric Decomposition of Arbitrariness to Constrain the Set of Good Models

Capsule Networks Do Not Need to Model Everything

DrugMCTS: a drug repurposing framework combining multi-agent, RAG and Monte Carlo Tree Search

Continuous Classification Aggregation

MedGemma Technical Report

Measuring Scientific Capabilities of Language Models with a Systems Biology Dry Lab

A Fuzzy Approach to the Specification, Verification and Validation of Risk-Based Ethical Decision Making Models

Thinking Beyond Tokens: From Brain-Inspired Intelligence to Cognitive Foundations for Artificial General Intelligence and its Societal Impact

Individual Causal Inference with Structural Causal Model

SymRAG: Efficient Neuro-Symbolic Retrieval Through Adaptive Query Routing

Scientists' First Exam: Probing Cognitive Abilities of MLLM via Perception, Understanding, and Reasoning

Reinforced Reasoning for Embodied Planning

Unearthing Gems from Stones: Policy Optimization with Negative Sample Augmentation for LLM Reasoning

Access Controls Will Solve the Dual-Use Dilemma

HiBayES: A Hierarchical Bayesian Modeling Framework for AI Evaluation Statistics

Leanabell-Prover: Posttraining Scaling in Formal Reasoning

Synthesizing world models for bilevel planning

MAPoRL: Multi-Agent Post-Co-Training for Collaborative Large Language Models with Reinforcement Learning

Practical Principles for AI Cost and Compute Accounting

Instantiation-based Formalization of Logical Reasoning Tasks using Language Models and Logical Solvers

Fourier Position Embedding: Enhancing Attention's Periodic Extension for Length Generalization

CATP-LLM: Empowering Large Language Models for Cost-Aware Tool Planning

Bongard in Wonderland: Visual Puzzles that Still Make AI Go Mad?

A Comprehensive Survey of Direct Preference Optimization: Datasets, Theories, Variants, and Applications

GI-NAS: Boosting Gradient Inversion Attacks through Adaptive Neural Architecture Search

Faster Reinforcement Learning by Freezing Slow States

Self-supervised Learning on Camera Trap Footage Yields a Strong Universal Face Embedder

EmbRACE-3K: Embodied Reasoning and Action in Complex Environments

Disentangling Neural Disjunctive Normal Form Models

ScaffoldAvatar: High-Fidelity Gaussian Avatars with Patch Expressions

CodeJudgeBench: Benchmarking LLM-as-a-Judge for Coding Tasks

WildFX: A DAW-Powered Pipeline for In-the-Wild Audio FX Graph Modeling

Reasoning or Memorization? Unreliable Results of Reinforcement Learning Due to Data Contamination

Accurate generation of chemical reaction transition states by conditional flow matching

Chat with AI: The Surprising Turn of Real-time Video Communication from Human to AI

Benchmarking and Evaluation of AI Models in Biology: Outcomes and Recommendations from the CZI Virtual Cells Workshop

Scene-Aware Conversational ADAS with Generative AI for Real-Time Driver Assistance

Cameras as Relative Positional Encoding

BenchReAD: A systematic benchmark for retinal anomaly detection

Can You Detect the Difference?

Privacy-Preserving Multi-Stage Fall Detection Framework with Semi-supervised Federated Learning and Robotic Vision Confirmation

An Empirical Evaluation of AI-Powered Non-Player Characters' Perceived Realism and Performance in Virtual Reality Environments

AudioMAE++: learning better masked audio representations with SwiGLU FFNs

RAPNet: A Receptive-Field Adaptive Convolutional Neural Network for Pansharpening

Logic layer Prompt Control Injection (LPCI): A Novel Security Vulnerability Class in Agentic Systems

CoralVQA: A Large-Scale Visual Question Answering Dataset for Coral Reef Image Understanding

Evaluating Fake Music Detection Performance Under Audio Augmentations

Referential ambiguity and clarification requests: comparing human and LLM behaviour

Response Wide Shut? Surprising Observations in Basic Vision Language Model Capabilities

From Sequence to Structure: Uncovering Substructure Reasoning in Transformers

Efficient Federated Learning with Heterogeneous Data and Adaptive Dropout

Multiple Choice Learning of Low Rank Adapters for Language Modeling

Energy Efficiency in AI for 5G and Beyond: A DeepRx Case Study

Devanagari Handwritten Character Recognition using Convolutional Neural Network

TAT: Temporal-Aligned Transformer for Multi-Horizon Peak Demand Forecasting

Feature Distillation is the Better Choice for Model-Heterogeneous Federated Learning

Toolsuite for Implementing Multiagent Systems Based on Communication Protocols

Recognizing Dementia from Neuropsychological Tests with State Space Models

FaceLLM: A Multimodal Large Language Model for Face Understanding

Learning in Chaos: Efficient Autoscaling and Self-healing for Distributed Training at the Edge

Created by

Haebom

저자

Wenjiao Feng, Rongxing Xiao, Zonghang Li, Hongfang Yu, Gang Sun, Long Luo, Mohsen Guizani, Qirong Ho

개요

본 논문은 에지 AI 클러스터에서 빈번한 노드 및 링크 변화로 인한 분산 학습의 중단 문제를 해결하기 위해 자가 치유 및 자동 확장 기능을 갖춘 탄력적이고 확장 가능한 에지 분산 학습 시스템인 Chaos를 제안합니다. Chaos는 다중 이웃 복제와 빠른 조각 스케줄링을 사용하여 확장 속도를 높이고, 클러스터 모니터를 통해 리소스 및 토폴로지 변화를 추적하여 스케줄러 결정을 지원하며, 피어 협상 프로토콜을 통해 중앙 관리자 없이 완전한 자가 관리 자동 확장을 가능하게 합니다. 실험 결과, Chaos는 Pollux, EDL 및 Autoscaling보다 훨씬 낮은 확장 지연 시간을 달성하고, scale-in, connect-link 및 disconnect-link 이벤트를 1밀리초 이내에 처리하여 노드 참여, 탈퇴 및 실패를 원활하게 처리합니다. 또한 가장 낮은 유휴 시간을 제공하여 클러스터가 성장함에 따라 우수한 리소스 사용 및 확장성을 보여줍니다.

시사점, 한계점

•

시사점:

◦

에지 AI 클러스터의 불안정한 환경에서도 탄력적이고 확장 가능한 분산 학습을 가능하게 하는 새로운 시스템을 제시합니다.

◦

다중 이웃 복제 및 빠른 조각 스케줄링을 통해 기존 시스템보다 훨씬 빠른 확장 속도를 제공합니다.

◦

자가 관리 자동 확장 기능을 통해 중앙 관리자의 개입 없이 효율적인 리소스 관리를 가능하게 합니다.

◦

극히 짧은 시간 내에 노드의 연결 및 분리 이벤트를 처리하여 시스템의 안정성을 크게 향상시킵니다.

•

한계점:

◦

본 논문에서 제시된 실험 결과는 특정 환경에서 수행되었으므로, 다른 환경에서는 성능이 다를 수 있습니다.

◦

Chaos 시스템의 복잡성이 높아 구현 및 유지보수에 어려움이 있을 수 있습니다.

◦

대규모 클러스터에 대한 성능 평가가 추가적으로 필요합니다.

◦

다양한 에지 환경에 대한 적용성 검증이 필요합니다.

Made with Slashpage