리미

https://aclanthology.org/2025.emnlp-main.725/ SHARP: Steering Hallucination in LVLMs via Representation EngineeringJunfei Wu, Yue Ding, Guofan Liu, Tianze Xia, Ziyue Huang, Dianbo Sui, Qiang Liu, Shu Wu, Liang Wang, Tieniu Tan. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.aclanthology.org Introduction기존 연구들은 환각의 원인을 크게 두 가지 관점에서 설명한다. 외부 요인은 학습 데이터..

https://arxiv.org/abs/2408.16314 ResVG: Enhancing Relation and Semantic Understanding in Multiple Instances for Visual GroundingVisual grounding aims to localize the object referred to in an image based on a natural language query. Although progress has been made recently, accurately localizing target objects within multiple-instance distractions (multiple objects of the same categarxiv.org (방법론..

https://arxiv.org/abs/2505.22334 Advancing Multimodal Reasoning via Reinforcement Learning with Cold StartRecent advancements in large language models (LLMs) have demonstrated impressive chain-of-thought reasoning capabilities, with reinforcement learning (RL) playing a crucial role in this progress. While "aha moment" patterns--where models exhibit self-correarxiv.org IntroductionLLM은 뛰어난 CoT 추..

https://www.youtube.com/watch?v=127lV0wcDmc&list=PLQIgLu3Wf-q_Ne8vv-ZXuJ4mztHJaQb_v&index=6 Text Splitter : 토큰 제한이 있는 LLM이 여러 문장을 참고해 답변할 수 있도록 문서를 분할하는 역할Chunk 하나당 하나의 Vector 생성대부분의 경우에 RecursiveCharacterTextSplitter를 통해 분할CharacterTextSplitter가장 간단한 텍스트 분할기로, 특정 구분자를 기준으로 텍스트를 여러 개로 분할from langchain.text_splitter import CharacterTextSplittertext_splitter = CharacterTextSplitter( separator =..

https://www.youtube.com/watch?v=tIU2tw3PMUE&list=PLQIgLu3Wf-q_Ne8vv-ZXuJ4mztHJaQb_v&index=5 RAG : 외부 데이터를 참조하여 LLM이 답변할 수 있도록 해주는 프레임워크 Document Loaders : 다양한 형태의 문서를 RAG 전용 객체로 불러들이는 모듈Page_content : 문서의 내용Metadata : 문서의 위치, 제목, 페이지 넘버 등 URL Document Loaderfrom langchain.document_loaders import WebBaseLoaderloader = WebBaseLoader("경로")data = loader.load()from langchain.document_loaders import..

https://arxiv.org/abs/2005.11401 Retrieval-Augmented Generation for Knowledge-Intensive NLP TasksLarge pre-trained language models have been shown to store factual knowledge in their parameters, and achieve state-of-the-art results when fine-tuned on downstream NLP tasks. However, their ability to access and precisely manipulate knowledge is still limarxiv.orgIntroduction사전 학습된 신경 언어 모델들은 데이터로부터..

https://arxiv.org/abs/1811.12560 An Introduction to Deep Reinforcement LearningDeep reinforcement learning is the combination of reinforcement learning (RL) and deep learning. This field of research has been able to solve a wide range of complex decision-making tasks that were previously out of reach for a machine. Thus, deep RL openarxiv.org 일반화는 다음 두 가지로 정의된다. 같은 환경에서의 일반화훈련 데이터는 일부만 있고, 전체 상태..

https://arxiv.org/abs/1811.12560 An Introduction to Deep Reinforcement LearningDeep reinforcement learning is the combination of reinforcement learning (RL) and deep learning. This field of research has been able to solve a wide range of complex decision-making tasks that were previously out of reach for a machine. Thus, deep RL openarxiv.orgPure model-based methods환경 모델이 명시적으로 주어지는 경우는 따로 학습할 필..

https://arxiv.org/abs/1811.12560 An Introduction to Deep Reinforcement LearningDeep reinforcement learning is the combination of reinforcement learning (RL) and deep learning. This field of research has been able to solve a wide range of complex decision-making tasks that were previously out of reach for a machine. Thus, deep RL openarxiv.org1. Stochastic Policy Gradient강화학습의 목표는 누적 보상을 최대화하는 최적..

https://arxiv.org/abs/1811.12560 An Introduction to Deep Reinforcement LearningDeep reinforcement learning is the combination of reinforcement learning (RL) and deep learning. This field of research has been able to solve a wide range of complex decision-making tasks that were previously out of reach for a machine. Thus, deep RL openarxiv.org 1. Q-learningQ-learning은 강화학습에서 최적 정책을 찾기 위해 행동의 가치를 ..