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.

SENCA-st: Integrating Spatial Transcriptomics and Histopathology with Cross Attention Shared Encoder for Region Identification in Cancer Pathology

Automatic Grid Updates for Kolmogorov-Arnold Networks using Layer Histograms

The Path Not Taken: RLVR Provably Learns Off the Principals

Moral Susceptibility and Robustness under Persona Role-Play in Large Language Models

LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristics

Large Sign Language Models: Toward 3D American Sign Language Translation

Introducing A Bangla Sentence - Gloss Pair Dataset for Bangla Sign Language Translation and Research

SPEAR-MM: Selective Parameter Evaluation and Restoration via Model Merging for Efficient Financial LLM Adaptation

HQ-SVC: Towards High-Quality Zero-Shot Singing Voice Conversion in Low-Resource Scenarios

Designing LLM-based Multi-Agent Systems for Software Engineering Tasks: Quality Attributes, Design Patterns and Rationale

Binary Split Categorical feature with Mean Absolute Error Criteria in CART

Contrastive Integrated Gradients: A Feature Attribution-Based Method for Explaining Whole Slide Image Classification

Understanding Electro-communication and Electro-sensing in Weakly Electric Fish using Multi-Agent Deep Reinforcement Learning

Anatomy-VLM: A Fine-grained Vision-Language Model for Medical Interpretation

Interaction Dynamics as a Reward Signal for LLMs

Unifying Model and Layer Fusion for Speech Foundation Models

RAPTR: Radar-based 3D Pose Estimation using Transformer

Bid Farewell to Seesaw: Towards Accurate Long-tail Session-based Recommendation via Dual Constraints of Hybrid Intents

Text-based Aerial-Ground Person Retrieval

A Circular Argument : Does RoPE need to be Equivariant for Vision?

DPRM: A Dual Implicit Process Reward Model in Multi-Hop Question Answering

Extreme Model Compression with Structured Sparsity at Low Precision

Hybrid Quantum-Classical Selective State Space Artificial Intelligence

Towards Open-Set Myoelectric Gesture Recognition via Dual-Perspective Inconsistency Learning

HN-MVTS: HyperNetwork-based Multivariate Time Series Forecasting

LPPG-RL: Lexicographically Projected Policy Gradient Reinforcement Learning with Subproblem Exploration

Adaptive Multi-Agent Response Refinement in Conversational Systems

Improving the accuracy and generalizability of molecular property regression models with a substructure-substitution-rule-informed framework

Test-time Diverse Reasoning by Riemannian Activation Steering

Dual-Kernel Graph Community Contrastive Learning

Bi-Objective Evolutionary Optimization for Large-Scale Open Pit Mine Scheduling Problem under Uncertainty with Chance Constraints

ImagebindDC: Compressing Multi-modal Data with Imagebind-based Condensation

NERVE: Neighbourhood & Entropy-guided Random-walk for training free open-Vocabulary sEgmentation

Hierarchical Direction Perception via Atomic Dot-Product Operators for Rotation-Invariant Point Clouds Learning

Remodeling Semantic Relationships in Vision-Language Fine-Tuning

Real-Time Performance Analysis of Multi-Fidelity Residual Physics-Informed Neural Process-Based State Estimation for Robotic Systems

Benchmarking Educational LLMs with Analytics: A Case Study on Gender Bias in Feedback

2D Representation for Unguided Single-View 3D Super-Resolution in Real-Time

FedPoP: Federated Learning Meets Proof of Participation

MARC: Multimodal and Multi-Task Agentic Retrieval-Augmented Generation for Cold-Start Recommender System

Deep (Predictive) Discounted Counterfactual Regret Minimization

ProbSelect: Stochastic Client Selection for GPU-Accelerated Compute Devices in the 3D Continuum

Relation as a Prior: A Novel Paradigm for LLM-based Document-level Relation Extraction

SafeMIL: Learning Offline Safe Imitation Policy from Non-Preferred Trajectories

OTSNet: A Neurocognitive-Inspired Observation-Thinking-Spelling Pipeline for Scene Text Recognition

A robust methodology for long-term sustainability evaluation of Machine Learning models

PerspAct: Enhancing LLM Situated Collaboration Skills through Perspective Taking and Active Vision

StableMorph: High-Quality Face Morph Generation with Stable Diffusion

Beyond the Pixels: VLM-based Evaluation of Identity Preservation in Reference-Guided Synthesis

Dynamic Sparsity: Challenging Common Sparsity Assumptions for Learning World Models in Robotic Reinforcement Learning Benchmarks

BARD10: A New Benchmark Reveals Significance of Bangla Stop-Words in Authorship Attribution

Hierarchical Structure-Property Alignment for Data-Efficient Molecular Generation and Editing

Constrained and Robust Policy Synthesis with Satisfiability-Modulo-Probabilistic-Model-Checking

An Integrated Fusion Framework for Ensemble Learning Leveraging Gradient Boosting and Fuzzy Rule-Based Models

CLIP is All You Need for Human-like Semantic Representations in Stable Diffusion

Radar-APLANC: Unsupervised Radar-based Heartbeat Sensing via Augmented Pseudo-Label and Noise Contrast

Taming Identity Consistency and Prompt Diversity in Diffusion Models via Latent Concatenation and Masked Conditional Flow Matching

ProSona: Prompt-Guided Personalization for Multi-Expert Medical Image Segmentation

Multi-modal Deepfake Detection and Localization with FPN-Transformer

AVOID-JACK: Avoidance of Jackknifing for Swarms of Long Heavy Articulated Vehicles

Invisible Triggers, Visible Threats! Road-Style Adversarial Creation Attack for Visual 3D Detection in Autonomous Driving

DOA Estimation with Lightweight Network on LLM-Aided Simulated Acoustic Scenes

Sharp Eyes and Memory for VideoLLMs: Information-Aware Visual Token Pruning for Efficient and Reliable VideoLLM Reasoning

Self-Correction Distillation for Structured Data Question Answering

Hardware-Aware YOLO Compression for Low-Power Edge AI on STM32U5 for Weeds Detection in Digital Agriculture

State of the Art in Text Classification for South Slavic Languages: Fine-Tuning or Prompting?

NOTAM-Evolve: A Knowledge-Guided Self-Evolving Optimization Framework with LLMs for NOTAM Interpretation

Morphing Through Time: Diffusion-Based Bridging of Temporal Gaps for Robust Alignment in Change Detection

Reliable and Private Utility Signaling for Data Markets

USV Obstacles Detection and Tracking in Marine Environments

Balance Equation-based Distributionally Robust Offline Imitation Learning

Libra-MIL: Multimodal Prototypes Stereoscopic Infused with Task-specific Language Priors for Few-shot Whole Slide Image Classification

DiffRegCD: Integrated Registration and Change Detection with Diffusion Features

SpeechJudge: Towards Human-Level Judgment for Speech Naturalness

CNN-Based Automated Parameter Extraction Framework for Modeling Memristive Devices

Exploring the Underwater World Segmentation without Extra Training

Test-driven Reinforcement Learning

Statistically Assuring Safety of Control Systems using Ensembles of Safety Filters and Conformal Prediction

Toward Adaptive BCIs: Enhancing Decoding Stability via User State-Aware EEG Filtering

Generating Sketches in a Hierarchical Auto-Regressive Process for Flexible Sketch Drawing Manipulation at Stroke-Level

Intelligence per Watt: Measuring Intelligence Efficiency of Local AI

Meta-cognitive Multi-scale Hierarchical Reasoning for Motor Imagery Decoding

LoopLLM: Transferable Energy-Latency Attacks in LLMs via Repetitive Generation

LLM-Powered Fully Automated Chaos Engineering: Towards Enabling Anyone to Build Resilient Software Systems at Low Cost

A General Method for Proving Networks Universal Approximation Property

MURPHY: Multi-Turn GRPO for Self Correcting Code Generation

SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control

Sparse3DPR: Training-Free 3D Hierarchical Scene Parsing and Task-Adaptive Subgraph Reasoning from Sparse RGB Views

PRISM: Privacy-preserving Inference System with Homomorphic Encryption and Modular Activation

Judging by the Rules: Compliance-Aligned Framework for Modern Slavery Statement Monitoring

HybridGuard: Enhancing Minority-Class Intrusion Detection in Dew-Enabled Edge-of-Things Networks

Physical Consistency of Aurora's Encoder: A Quantitative Study

Semantic-Consistent Bidirectional Contrastive Hashing for Noisy Multi-Label Cross-Modal Retrieval

SALT: Steering Activations towards Leakage-free Thinking in Chain of Thought

Filtered-ViT: A Robust Defense Against Multiple Adversarial Patch Attacks

Auto-US: An Ultrasound Video Diagnosis Agent Using Video Classification Framework and LLMs

UltraGS: Gaussian Splatting for Ultrasound Novel View Synthesis

TurboSAT: Gradient-Guided Boolean Satisfiability Accelerated on GPU-CPU Hybrid System

Global Optimization on Graph-Structured Data via Gaussian Processes with Spectral Representations

ViPRA: Video Prediction for Robot Actions

EasyEdit2: An Easy-to-use Steering Framework for Editing Large Language Models

Created by

Haebom

저자

Ziwen Xu, Shuxun Wang, Kewei Xu, Haoming Xu, Mengru Wang, Xinle Deng, Yunzhi Yao, Guozhou Zheng, Huajun Chen, Ningyu Zhang

개요

EasyEdit2는 대규모 언어 모델(LLM)의 동작을 플러그 앤 플레이 방식으로 조절할 수 있도록 설계된 프레임워크입니다. 안전성, 감정, 개성, 추론 패턴, 사실성 및 언어 기능을 포함한 광범위한 테스트 시간 개입을 지원합니다. 이전 버전과 달리, 매개변수를 수정하지 않고도 조향 벡터를 자동으로 생성 및 적용하여 모델의 동작에 영향을 미칠 수 있도록 조향 벡터 생성기 및 조향 벡터 적용기와 같은 주요 모듈로 구성된 새로운 아키텍처를 특징으로 합니다. 사용 편의성이 장점이며, 단 하나의 예시만으로도 모델의 응답을 효과적으로 안내하고 조정할 수 있게 합니다. 다양한 LLM에 대한 모델 조향 성능을 실험적으로 보고하고 있으며, 소스 코드와 데모 비디오를 공개했습니다.

시사점, 한계점

•

시사점:

◦

LLM의 동작을 쉽고 효율적으로 조절할 수 있는 새로운 프레임워크 제공

◦

전문적인 지식 없이도 LLM의 안전성, 감정, 개성 등 다양한 측면을 제어 가능

◦

다양한 LLM에 적용 가능한 일반적인 접근 방식 제시

◦

소스 코드 공개를 통한 접근성 향상

•

한계점:

◦

논문에서 구체적인 성능 지표 및 비교 대상에 대한 정보 부족

◦

새로운 아키텍처의 구체적인 기술적 세부 사항 부족

◦

장기적인 안정성 및 확장성에 대한 검증 필요

◦

특정 유형의 개입에 대한 효과의 한계 존재 가능성

Made with Slashpage