Skip to content

Commit 690ae19

Browse files
committed
Move publication to conference folder: LLM-guided multi-agent reinforcement learning
1 parent c6454cd commit 690ae19

2 files changed

Lines changed: 35 additions & 38 deletions

File tree

  • content/publication
    • 2025-cooperative-multi-agent-anomaly-detection-through-consensus-and-integration
    • conference/2025-llm-guided-multi-agent-reinforcement-learning-for-anomaly-detection

‎content/publication/2025-cooperative-multi-agent-anomaly-detection-through-consensus-and-integration/index.md‎

Lines changed: 0 additions & 38 deletions
This file was deleted.
Lines changed: 35 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,35 @@
1+
---
2+
title: LLM 기반 다중 에이전트 강화학습을 활용한 이상탐지 기법
3+
authors:
4+
- dongwook-kwon
5+
- minseong-oh
6+
- cheolsoo-park
7+
date: '2025-10-01T00:00:00Z'
8+
publishDate: '2025-10-01T00:00:00Z'
9+
publication_types:
10+
- paper-conference
11+
publication: 대한전자공학회 추계학술대회, October 2025
12+
summary: 대한전자공학회 추계학술대회, October 2025
13+
featured: false
14+
tags:
15+
- Anomaly Detection
16+
- Multi-Agent Systems
17+
- Reinforcement Learning
18+
- Large Language Models
19+
url_code: ''
20+
url_dataset: ''
21+
url_poster: ''
22+
url_project: ''
23+
url_slides: ''
24+
url_source: ''
25+
url_video: ''
26+
abstract: >
27+
본 논문은 다변량 시계열 이상탐지를 위해 대형 언어모델(LLM) 기반
28+
다중 에이전트 강화학습 프레임워크를 제안한다. 각 에이전트는 개별적으로
29+
이상을 탐지하며, 강화학습을 통해 에이전트 간 통신 구조를 효율적으로
30+
최적화한다. 일부 에이전트는 LLM을 활용하여 수치적 이상뿐 아니라
31+
의미적·맥락적 이상을 해석하며, 전체 시스템은 앙상블 융합과 불확실성
32+
보정 기법을 통해 탐지 성능을 향상시킨다. NASA Mars Science Laboratory
33+
데이터셋 실험 결과, 제안된 방법은 기존 개별 탐지 방식 대비
34+
정밀도, 재현율 및 F1-score에서 유의미한 성능 향상을 보였다.
35+
---

0 commit comments

Comments
 (0)