郭志義

COMET: Cross-domain Optimization of Meta-prompts for Efficient Target-domain Prompt Synthesis

智慧計算與科技研究所

提示工程、小型語言模型、跨領域遷移、元提示、行為對齊、Prompt Engineering、Small Language Models、Cross-domain Migration、Meta-prompt、Behavioral Alignment

CogMem: A Cognitive Memory System with Graph-Based Reasoning for LLM Agents

智慧計算與科技研究所智慧物聯網產業碩士專班

大型語言模型代理、個人化記憶系統、長期記憶、知識圖譜、混合式檢索、個人化 PageRank、多跳推理、LLM Agents、Personalized Memory Systems、Long-term Memory、Knowledge Graph、Hybrid Retrieval、Personalized PageRank、Multi-Hop Reasoning

Automated Prompt Optimization for Intervention Timing and Response Strategy Control in Tool-Use Multi-Party Conversational Agents

智慧系統與應用研究所・共同指導：魏澤人

多方對話代理人、雙層代理人架構、自動化提示最佳化、競技場評估框架、遺傳帕雷托提示演化、Multi-Party Conversational Agent、Dual-Layer Agent Architecture、Automated Prompt Optimization、Arena Evaluation Framework、Genetic-Pareto Prompt Evolution

Improving Startup Phase Quality in Hyperparameter Optimization via LLM-Guided Search Space Refinement

智慧系統與應用研究所

超參數最佳化、序列式模型導向最佳化、搜尋空間收縮、固定預算最佳化、初始設計、聯邦式學習、啟動階段、貝葉斯最佳化、hyperparameter optimization、sequential model-based optimization、search space shrinkage、fixed-budget optimization、initial design、federated learning、startup phase、Bayesian optimization

SPICE：Shapley-guided Proximate and Interpretable Counterfactual Explanations

智慧與綠能產學研究所

可解釋人工智慧、反事實解釋、Explainable Artificial Intelligence、counterfactual explanation

REMIND: Recall Enhanced Memory Integration for Natural Language Dialogue Systems

智慧計算與科技研究所

語言模型、對話記憶系統、混和記憶搜尋、多輪對話系統、個人化AI、Memory-Augmented Dialog Systems、Large Language Models (LLMs)、Hybrid Memory Retrieval、Multi-Turn Dialogue Systems、Personalized Memory Systems

SwiftAgg++: A Communication-Efficient Secure Aggregation Protocol with Privacy Guarantee for Federated Learning

智慧計算與科技研究所

聯邦式學習、安全匯集運算、可驗證隨機函數、資料隱私、通訊效率、SwiftAgg、Federated Learning、Secure Aggregation、Verifiable Random Functions、Privacy、Communication Efficiency

FedCLO: Federated Learning with Clustered Layer-wise Communication Load Optimization

智慧計算與科技研究所

聯邦學習、分層模型聚合、模型壓縮、集群化聯邦學習、分佈歐式距離、Federated Learning、Layer-wise model aggregation、Model compression、Clustered Federated Learning、Distribution Euclidean Distance

Continual Learning with Out-of-Distribution Data Detection for Fine-grained Defect Classification

智慧系統與應用研究所

持續學習、細粒度影像辨識、分布外數據偵測、continual learning、fine-grained image categorization、out-of-distribution detection

Fast Adaption Classifier: A Framework for Overcoming Challenges in AI Model Training for Smart Manufacturing Defect Classification

智慧與綠能產學研究所

類別間樣本不平衡、小樣本學習、細粒度分類、開放集、Class-imbalance、Few-shot learning、Fine-grained、Open-set

Gated Tri-Tower Transformer (GT3) - An Inflated Power Generation Attack Detector for Microgrids

智慧計算與科技研究所

發電量浮報攻擊、區塊鏈、閘控 Transformer、閘控三塔 Transformer、微電網、Inflated Power Generation Attack、Blockchain、Gated Transformer、Gated Tri-Tower Transformer、Microgrid

Improving Federated Learning Communication Efficiency with Global Momentum Fusion for Gradient Compression Schemes

智慧系統與應用研究所

聯邦學習、通訊流量、梯度壓縮、記憶機制、融合全局動量、Federated Learning、Communication Overheads、Gradient Compression、Memory Mechanism、Global Momentum Fusion

Adaptive Federated Proximal Optimization

智慧與綠能產學研究所

聯邦學習、近端參數項、自適應近端參數項、統計異質性、系統異質性、目標不一致性、Federated Learning、Proximal Term、Adaptive Proximal Term、Statistical Heterogeneity、System Heterogeneity、Objective Inconsistency