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.

RECUR: Resource Exhaustion Attack via Recursive-Entropy Guided Counterfactual Utilization and Reflection

Initial Risk Probing and Feasibility Testing of Glow: a Generative AI-Powered Dialectical Behavior Therapy Skills Coach for Substance Use Recovery and HIV Prevention

Interpretable Failure Analysis in Multi-Agent Reinforcement Learning Systems

Objective Decoupling in Social Reinforcement Learning: Recovering Ground Truth from Sycophantic Majorities

Securing Dual-Use Pathogen Data of Concern

Graph-Enhanced Deep Reinforcement Learning for Multi-Objective Unrelated Parallel Machine Scheduling

Free(): Learning to Forget in Malloc-Only Reasoning Models

Structure-Aware Robust Counterfactual Explanations via Conditional Gaussian Network Classifiers

Small Agent Group is the Future of Digital Health

Towards Adaptive, Scalable, and Robust Coordination of LLM Agents: A Dynamic Ad-Hoc Networking Perspective

Accelerating Social Science Research via Agentic Hypothesization and Experimentation

LOCA-bench: Benchmarking Language Agents Under Controllable and Extreme Context Growth

IV Co-Scientist: Multi-Agent LLM Framework for Causal Instrumental Variable Discovery

MePo: Meta Post-Refinement for Rehearsal-Free General Continual Learnin

Selective Fine-Tuning for Targeted and Robust Concept Unlearning

MedCoG: Maximizing LLM Inference Density in Medical Reasoning via Meta-Cognitive Regulation

GCN-MPPR: Enhancing the Propagation of Message Passing Neural Networks via Motif-Based Personalized PageRank

MemFly: On-the-Fly Memory Optimization via Information Bottleneck

ToolSelf: Unifying Task Execution and Self-Reconfiguration via Tool-Driven Intrinsic Adaptation

Emergent Misalignment is Easy, Narrow Misalignment is Hard

LQA: A Lightweight Quantized-Adaptive Framework for Vision-Language Models on the Edge

Time Series Reasoning via Process-Verifiable Thinking Data Synthesis and Scheduling for Tailored LLM Reasoning

Data Darwinism Part I: Unlocking the Value of Scientific Data for Pre-training

Do Multi-Agents Dream of Electric Screens? Achieving Perfect Accuracy on AndroidWorld Through Task Decomposition

Disentangled Instrumental Variables for Causal Inference with Networked Observational Data

Learning to Continually Learn via Meta-learning Agentic Memory Designs

Humanizing AI Grading: Student-Centered Insights on Fairness, Trust, Consistency and Transparency

Geo-Code: A Code Framework for Reverse Code Generation from Geometric Images Based on Two-Stage Multi-Agent Evolution

EventCast: Hybrid Demand Forecasting in E-Commerce with LLM-Based Event Knowledge

ONTrust: A Reference Ontology of Trust

Efficient Table Retrieval and Understanding with Multimodal Large Language Models

SleepMaMi: A Universal Sleep Foundation Model for Integrating Macro- and Micro-structures

M2A: Multimodal Memory Agent with Dual-Layer Hybrid Memory for Long-Term Personalized Interactions

VERIFY-RL: Verifiable Recursive Decomposition for Reinforcement Learning in Mathematical Reasoning

When Is Enough Not Enough? Illusory Completion in Search Agents

MSP-LLM: A Unified Large Language Model Framework for Complete Material Synthesis Planning

Joint Reward Modeling: Internalizing Chain-of-Thought for Efficient Visual Reward Models

GraphAgents: Knowledge Graph-Guided Agentic AI for Cross-Domain Materials Design

Computing the Reachability Value of Posterior-Deterministic POMDPs

Are Reasoning LLMs Robust to Interventions on Their Chain-of-Thought?

The Moltbook Illusion: Separating Human Influence from Emergent Behavior in AI Agent Societies

Can LLMs Truly Embody Human Personality? Analyzing AI and Human Behavior Alignment in Dispute Resolution

Progressive Multi-Agent Reasoning for Biological Perturbation Prediction

VGAS: Value-Guided Action-Chunk Selection for Few-Shot Vision-Language-Action Adaptation

NAAMSE: Framework for Evolutionary Security Evaluation of Agents

W&D:Scaling Parallel Tool Calling for Efficient Deep Research Agents

SupChain-Bench: Benchmarking Large Language Models for Real-World Supply Chain Management

RAPiD: Real-time Deterministic Trajectory Planning via Diffusion Behavior Priors for Safe and Efficient Autonomous Driving

Adaptive Scaffolding for Cognitive Engagement in an Intelligent Tutoring System

Steer2Adapt: Dynamically Composing Steering Vectors Elicits Efficient Adaptation of LLMs

TermiGen: High-Fidelity Environment and Robust Trajectory Synthesis for Terminal Agents

BRIDGE: Predicting Human Task Completion Time From Model Performance

Incentive-Aware AI Safety via Strategic Resource Allocation: A Stackelberg Security Games Perspective

From Out-of-Distribution Detection to Hallucination Detection: A Geometric View

Is there "Secret Sauce'' in Large Language Model Development?

PreFlect: From Retrospective to Prospective Reflection in Large Language Model Agents

ANCHOR: Branch-Point Data Generation for GUI Agents

Theory of Space: Can Foundation Models Construct Spatial Beliefs through Active Exploration?

Aster: Autonomous Scientific Discovery over 20x Faster Than Existing Methods

DLLM-Searcher: Adapting Diffusion Large Language Model for Search Agents

ST-Raptor: An Agentic System for Semi-Structured Table QA

LLM-FSM: Scaling Large Language Models for Finite-State Reasoning in RTL Code Generation

Can Post-Training Transform LLMs into Causal Reasoners?

Accelerating Vision Transformers on Brain Processing Unit

Can One-sided Arguments Lead to Response Change in Large Language Models?

GRP-Obliteration: Unaligning LLMs With a Single Unlabeled Prompt

Steering Safely or Off a Cliff? Rethinking Specificity and Robustness in Inference-Time Interventions

ASMa: Asymmetric Spatio-temporal Masking for Skeleton Action Representation Learning

REBEL: Hidden Knowledge Recovery via Evolutionary-Based Evaluation Loop

ATEX-CF: Attack-Informed Counterfactual Explanations for Graph Neural Networks

RuleSmith: Multi-Agent LLMs for Automated Game Balancing

SR4-Fit: An Interpretable and Informative Classification Algorithm Applied to Prediction of U.S. House of Representatives Elections

Coupled Local and Global World Models for Efficient First Order RL

Addressing the Waypoint-Action Gap in End-to-End Autonomous Driving via Vehicle Motion Models

Emergent Low-Rank Training Dynamics in MLPs with Smooth Activations

Multi-Way Representation Alignment

Learning Rate Scaling across LoRA Ranks and Transfer to Full Finetuning

AnyThermal: Towards Learning Universal Representations for Thermal Perception

Personagram: Bridging Personas and Product Design for Creative Ideation with Multimodal LLMs

Generics in science communication: Misaligned interpretations across laypeople, scientists, and large language models

Optimal rates for density and mode estimation with expand-and-sparsify representations

Stop the Flip-Flop: Context-Preserving Verification for Fast Revocable Diffusion Decoding

Protean Compiler: An Agile Framework to Drive Fine-grain Phase Ordering

Hear You in Silence: Designing for Active Listening in Human Interaction with Conversational Agents Using Context-Aware Pacing

Self-Improving World Modelling with Latent Actions

Urban Spatio-Temporal Foundation Models for Climate-Resilient Housing: Scaling Diffusion Transformers for Disaster Risk Prediction

Coding Agents with Environment Interaction: A Theoretical Perspective

NanoNet: Parameter-Efficient Learning with Label-Scarce Supervision for Lightweight Text Mining Model

SVRepair: Structured Visual Reasoning for Automated Program Repair

Transformer-Based Reinforcement Learning for Autonomous Orbital Collision Avoidance in Partially Observable Environments

Communication Enhances LLMs' Stability in Strategic Thinking

Allocate Marginal Reviews to Borderline Papers Using LLM Comparative Ranking

HQP: Sensitivity-Aware Hybrid Quantization and Pruning for Ultra-Low-Latency Edge AI Inference

iScheduler: Reinforcement Learning-Driven Continual Optimization for Large-Scale Resource Investment Problems

Analyzing Diffusion and Autoregressive Vision Language Models in Multimodal Embedding Space

Rethinking Memory Mechanisms of Foundation Agents in the Second Half

Recontextualizing Famous Quotes for Brand Slogan Generation

Git for Sketches: An Intelligent Tracking System for Capturing Design Evolution

EUGens: Efficient, Unified, and General Dense Layers

Agentic Uncertainty Reveals Agentic Overconfidence

Pseudo-Boolean Proof Logging for Optimal Classical Planning

Created by

Haebom

Category

Empty

저자

Simon Dold, Malte Helmert, Jakob Nordstrom, Gabriele Roger, Tanja Schindler

개요

본 논문은 클래식 계획 문제에 대한 하한선 인증서를 소개합니다. 이 인증서는 독립적인 제3자가 검증할 수 있는 방식으로 문제의 해결 불가능성 또는 계획의 최적성을 증명하는 데 사용될 수 있습니다. 논문에서는 계획 알고리즘과 무관한 의사 부울 제약 조건을 기반으로 하한선 인증서를 생성하는 일반적인 프레임워크를 설명합니다. 사례 연구로, 패턴 데이터베이스 휴리스틱과 h<sup>max</sup>를 구체적인 예로 사용하여, 적당한 오버헤드로 최적성 증명을 생성하도록 A* 알고리즘을 수정하는 방법을 보여줍니다. 동일한 증명 로깅 방식은 의사 부울 제약 조건에 대한 추론으로 효율적으로 표현될 수 있는 모든 휴리스틱에 대해 작동합니다.

시사점, 한계점

•

시사점:

◦

클래식 계획 문제의 해결 불가능성 또는 계획의 최적성을 독립적으로 검증 가능한 방식으로 증명할 수 있는 새로운 방법을 제시합니다.

◦

계획 알고리즘과 무관하게 적용 가능한 일반적인 프레임워크를 제공합니다.

◦

A* 알고리즘을 수정하여 최적성 증명을 생성하는 방법을 제시하고, 패턴 데이터베이스 휴리스틱과 h<sup>max</sup>를 통해 실용성을 보여줍니다.

◦

의사 부울 제약 조건으로 효율적으로 표현 가능한 모든 휴리스틱에 적용 가능합니다.

•

한계점:

◦

제시된 프레임워크의 효율성 및 확장성에 대한 보다 심도있는 분석이 필요합니다.

◦

다양한 종류의 계획 문제 및 휴리스틱에 대한 실험적 평가가 부족합니다.

◦

인증서 생성에 필요한 계산 오버헤드가 실제 문제에 적용 가능한 수준인지에 대한 추가 연구가 필요합니다.

Made with Slashpage