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.

ADVERSA: Measuring Multi-Turn Guardrail Degradation and Judge Reliability in Large Language Models

The Epistemic Support-Point Filter: Jaynesian Maximum Entropy Meets Popperian Falsification

Training Language Models via Neural Cellular Automata

Where Do Flow Semantics Reside? A Protocol-Native Tabular Pretraining Paradigm for Encrypted Traffic Classification

InFusionLayer: a CFA-based ensemble tool to generate new classifiers for learning and modeling

Revisiting Sharpness-Aware Minimization: A More Faithful and Effective Implementation

Toward Epistemic Stability: Engineering Consistent Procedures for Industrial LLM Hallucination Reduction

Gated Adaptation for Continual Learning in Human Activity Recognition

Safety Under Scaffolding: How Evaluation Conditions Shape Measured Safety

AMB-DSGDN: Adaptive Modality-Balanced Dynamic Semantic Graph Differential Network for Multimodal Emotion Recognition

Targeted Bit-Flip Attacks on LLM-Based Agents

Evaluating Progress in Graph Foundation Models: A Comprehensive Benchmark and New Insights

HTM-EAR: Importance-Preserving Tiered Memory with Hybrid Routing under Saturation

Architecture-Aware LLM Inference Optimization on AMD Instinct GPUs: A Comprehensive Benchmark and Deployment Study

The DMA Streaming Framework: Kernel-Level Buffer Orchestration for High-Performance AI Data Paths

How to Count AIs: Individuation and Liability for AI Agents

A Governance and Evaluation Framework for Deterministic, Rule-Based Clinical Decision Support in Empiric Antibiotic Prescribing

RedFuser: An Automatic Operator Fusion Framework for Cascaded Reductions on AI Accelerators

Defining AI Models and AI Systems: A Framework to Resolve the Boundary Problem

Prompts and Prayers: the Rise of GPTheology

DeliberationBench: A Normative Benchmark for the Influence of Large Language Models on Users' Views

Assessing Cognitive Biases in LLMs for Judicial Decision Support: Virtuous Victim and Halo Effects

Measuring and Eliminating Refusals in Military Large Language Models

FERRET: Framework for Expansion Reliant Red Teaming

Personalized Group Relative Policy Optimization for Heterogenous Preference Alignment

GATech at AbjadMed: Bidirectional Encoders vs. Causal Decoders: Insights from 82-Class Arabic Medical Classification

SENS-ASR: Semantic Embedding injection in Neural-transducer for Streaming Automatic Speech Recognition

SpreadsheetArena: Decomposing Preference in LLM Generation of Spreadsheet Workbooks

Leveraging Wikidata for Geographically Informed Sociocultural Bias Dataset Creation: Application to Latin America

A Retrieval-Augmented Language Assistant for Unmanned Aircraft Safety Assessment and Regulatory Compliance

Automated evaluation of LLMs for effective machine translation of Mandarin Chinese to English

Empathy Is Not What Changed: Clinical Assessment of Psychological Safety Across GPT Model Generations

There Are No Silly Questions: Evaluation of Offline LLM Capabilities from a Turkish Perspective

Context Over Compute Human-in-the-Loop Outperforms Iterative Chain-of-Thought Prompting in Interview Answer Quality

Evaluating Adjective-Noun Compositionality in LLMs: Functional vs Representational Perspectives

CEI: A Benchmark for Evaluating Pragmatic Reasoning in Language Models

TAMUSA-Chat: A Domain-Adapted Large Language Model Conversational System for Research and Responsible Deployment

PoultryLeX-Net: Domain-Adaptive Dual-Stream Transformer Architecture for Large-Scale Poultry Stakeholder Modeling

A Two-Stage Architecture for NDA Analysis: LLM-based Segmentation and Transformer-based Clause Classification

The System Hallucination Scale (SHS): A Minimal yet Effective Human-Centered Instrument for Evaluating Hallucination-Related Behavior in Large Language Models

Causally Grounded Mechanistic Interpretability for LLMs with Faithful Natural-Language Explanations

Evolving Demonstration Optimization for Chain-of-Thought Feature Transformation

Quantifying Hallucinations in Language Language Models on Medical Textbooks

The Dunning-Kruger Effect in Large Language Models: An Empirical Study of Confidence Calibration

MoE-SpAc: Efficient MoE Inference Based on Speculative Activation Utility in Heterogeneous Edge Scenarios

AraModernBERT: Transtokenized Initialization and Long-Context Encoder Modeling for Arabic

Explainable LLM Unlearning Through Reasoning

One Model, Many Skills: Parameter-Efficient Fine-Tuning for Multitask Code Analysis

Decoupling Reasoning and Confidence: Resurrecting Calibration in Reinforcement Learning from Verifiable Rewards

A Hybrid Knowledge-Grounded Framework for Safety and Traceability in Prescription Verification

Nurture-First Agent Development: Building Domain-Expert AI Agents Through Conversational Knowledge Crystallization

Emulating Clinician Cognition via Self-Evolving Deep Clinical Research

FAME: Formal Abstract Minimal Explanation for Neural Networks

Trajectory-Informed Memory Generation for Self-Improving Agent Systems

Does LLM Alignment Really Need Diversity? An Empirical Study of Adapting RLVR Methods for Moral Reasoning

CUAAudit: Meta-Evaluation of Vision-Language Models as Auditors of Autonomous Computer-Use Agents

Adaptive RAN Slicing Control via Reward-Free Self-Finetuning Agents

IH-Challenge: A Training Dataset to Improve Instruction Hierarchy on Frontier LLMs

Resource-constrained Amazons chess decision framework integrating large language models and graph attention

Verbalizing LLM's Higher-order Uncertainty via Imprecise Probabilities

Beyond Scalars: Evaluating and Understanding LLM Reasoning via Geometric Progress and Stability

HEAL: Hindsight Entropy-Assisted Learning for Reasoning Distillation

Hybrid Self-evolving Structured Memory for GUI Agents

Agentic Control Center for Data Product Optimization

Turn: A Language for Agentic Computation

Permutation-Equivariant 2D State Space Models: Theory and Canonical Architecture for Multivariate Time Series

Architectural Design and Performance Analysis of FPGA based AI Accelerators: A Comprehensive Review

Sensitivity-Guided Framework for Pruned and Quantized Reservoir Computing Accelerators

Autonomous Edge-Deployed AI Agents for Electric Vehicle Charging Infrastructure Management

ARKV: Adaptive and Resource-Efficient KV Cache Management under Limited Memory Budget for Long-Context Inference in LLMs

PhD Thesis Summary: Methods for Reliability Assessment and Enhancement of Deep Neural Network Hardware Accelerators

Alignment Is the Disease: Censorship Visibility and Alignment Constraint Complexity as Determinants of Collective Pathology in Multi-Agent LLM Systems

ALADIN: Accuracy-Latency-Aware Design-space Inference Analysis for Embedded AI Accelerators

SiliconMind-V1: Multi-Agent Distillation and Debug-Reasoning Workflows for Verilog Code Generation

CktEvo: Repository-Level RTL Code Benchmark for Design Evolution

Design Conductor: An agent autonomously builds a 1.5 GHz Linux-capable RISC-V CPU

Unveiling the Potential of Quantization with MXFP4: Strategies for Quantization Error Reduction

Let's Verify Math Questions Step by Step

Think Before You Lie: How Reasoning Improves Honesty

The Confidence Gate Theorem: When Should Ranked Decision Systems Abstain?

PathMem: Toward Cognition-Aligned Memory Transformation for Pathology MLLMs

MedMASLab: A Unified Orchestration Framework for Benchmarking Multimodal Medical Multi-Agent Systems

Influencing LLM Multi-Agent Dialogue via Policy-Parameterized Prompts

LCA: Local Classifier Alignment for Continual Learning

Quantifying the Necessity of Chain of Thought through Opaque Serial Depth

World2Mind: Cognition Toolkit for Allocentric Spatial Reasoning in Foundation Models

AutoAgent: Evolving Cognition and Elastic Memory Orchestration for Adaptive Agents

Does the Question Really Matter? Training-Free Data Selection for Vision-Language SFT

OOD-MMSafe: Advancing MLLM Safety from Harmful Intent to Hidden Consequences

EsoLang-Bench: Evaluating Genuine Reasoning in Large Language Models via Esoteric Programming Languages

Logics-Parsing-Omni Technical Report

MiniAppBench: Evaluating the Shift from Text to Interactive HTML Responses in LLM-Powered Assistants

PRECEPT: Planning Resilience via Experience, Context Engineering & Probing Trajectories A Unified Framework for Test-Time Adaptation with Compositional Rule Learning and Pareto-Guided Prompt Evolution

Context Engineering: From Prompts to Corporate Multi-Agent Architecture

Enhancing Debunking Effectiveness through LLM-based Personality Adaptation

Vibe-Creation: The Epistemology of Human-AI Emergent Cognition

GenePlan: Evolving Better Generalized PDDL Plans using Large Language Models

Telogenesis: Goal Is All U Need

An Empirical Study and Theoretical Explanation on Task-Level Model-Merging Collapse

AI Act Evaluation Benchmark: An Open, Transparent, and Reproducible Evaluation Dataset for NLP and RAG Systems

Test It Before You Trust It: Applying Software Testing for Trustworthy In-context Learning

Created by

Haebom

Category

Empty

저자

Teeradaj Racharak, Chaiyong Ragkhitwetsagul, Chommakorn Sontesadisai, Thanwadee Sunetnanta

개요

본 논문은 대규모 언어 모델(LLM)의 인 컨텍스트 학습(ICL)의 신뢰성을 평가하기 위한 소프트웨어 테스트 기반 프레임워크인 MMT4NL을 제안합니다. MMT4NL은 적대적 섭동과 소프트웨어 테스트 기법을 활용하여 ICL의 다양한 언어적 능력을 평가합니다. LLM을 소프트웨어로 간주하고, 테스트 세트로부터 변형된 적대적 예시를 생성하여 ICL 프롬프트의 버그를 정량화하고 특정짓는 것을 목표로 합니다. 감정 분석과 질의응답 과제에 대한 실험을 통해 최첨단 LLM의 다양한 언어적 버그를 드러냅니다.

시사점, 한계점

•

시사점:

◦

소프트웨어 테스트 기법을 활용하여 LLM의 ICL 신뢰성을 평가하는 새로운 프레임워크(MMT4NL) 제시.

◦

적대적 섭동을 이용하여 LLM의 취약점과 예측 불가능한 동작을 탐지.

◦

다양한 언어적 버그를 발견하고, ICL의 신뢰성 향상에 기여할 수 있는 통찰력 제공.

•

한계점:

◦

제안된 프레임워크의 일반화 가능성에 대한 추가적인 연구 필요.

◦

다양한 LLM 아키텍처 및 과제에 대한 확장성 검증 필요.

◦

MMT4NL을 통해 발견된 버그의 심각도 및 영향에 대한 정량적 분석 부족.

Made with Slashpage