Institution profile

Macau University of Science and Technology

Academic institutionasia · mo
Official website
Research library196linked papers
Opportunities0open roles
Selected work

Representative Papers

Towards Understanding Retrieval Accuracy and Prompt Quality in RAG Systems

Nov 29, 2024arXiv.org

The impact of key design decisions—RAG activation, retrieval granularity, and knowledge integration strategy—on RAG system performance remains poorly understood. Method: We conduct systematic ablation studies across three code/qa benchmarks and two state-of-the-art LLMs, quantitatively evaluating how document type, recall rate, document selection strategy, and prompt engineering jointly affect answer correctness and confidence via multi-dimensional analysis, cross-model/dataset comparison, and joint prompt-retrieval analysis. Contribution/Results: We identify precise interaction patterns and operational boundaries among these factors and propose nine actionable, empirically grounded guidelines for diagnosing and optimizing RAG failures. Our findings significantly improve RAG system stability, debuggability, and reliability, offering rigorous empirical evidence and a principled methodology to support the engineering deployment of LLM-augmented systems.

7 citationsRead paper

HiMem: Hierarchical Long-Term Memory for LLM Long-Horizon Agents

Jan 10, 2026arXiv.org

Existing long-term memory systems struggle to simultaneously achieve adaptability, scalability, and self-evolution in continuous interaction scenarios. To address this challenge, this work proposes HiMem—a cognitively inspired hierarchical long-term memory framework that dynamically constructs, retrieves, and updates memories by establishing semantic associations between event memory and note memory. The core innovations include a topic-aware dual-channel segmentation strategy that distinguishes routine events from unexpected ones, and a conflict-aware memory reconsolidation mechanism, both integrated with multi-stage information extraction and a hybrid retrieval strategy. Experimental results demonstrate that HiMem significantly outperforms current state-of-the-art methods on long-term dialogue benchmarks, achieving leading performance in accuracy, consistency, long-range reasoning capability, and computational efficiency.

1 citationsRead paper

Large Language Models for Mobile GUI Text Input Generation: An Empirical Study

Apr 13, 2024arXiv.org

This study investigates the capability of large language models (LLMs) to autonomously generate effective textual inputs for Android GUI testing. Addressing 114 open-source apps, we propose a UI context-aware prompt engineering method to guide nine mainstream LLMs—including LLaMA and GPT variants—to produce inputs capable of triggering navigation transitions. We conduct the first large-scale empirical evaluation in this domain, revealing a statistically significant positive correlation between UI context completeness and page transition success rate. The best-performing model achieves a success rate of 50.58%–66.67%, substantially outperforming the invalid-input baseline (<23% defect detection). Contextual enhancement is empirically validated as critical to input quality. Our contributions include: (1) six actionable insights for applying LLMs in mobile testing practice, and (2) a reusable, context-driven framework for generating test-relevant textual inputs.

1 citationsRead paper
Recent publications

Latest Papers