Domain Knowledge Distillation from Large Language Model: An Empirical Study in the Autonomous Driving Domain
推荐理由
与当前方向有可迁移方法或背景价值，但不是本轮核心问题；注意：更像自动驾驶多模态/大模型背景材料，V2X 约束不足
核心判断
论文摘要（中文）
PDF 摘要：中文翻译尚未生成，请点击“显示原文依据”查看英文内容。
Engineering knowledge-based (or expert) systems require extensive manual effort and domain knowledge. As Large Language Models (LLMs) are trained using an enormous amount of cross-domain knowledge, it becomes possible to automate such engineering processes. This paper presents an empirical automation and semi-automation framework for domain knowledge distillation using prompt engineering and the LLM ChatGPT. We assess the framework empirically in the autonomous driving domain and present our key observations. In our implementation, we construct the domain knowledge ontology by “chatting” with ChatGPT. The key finding is that while fully automated domain ontology construction is possible, human supervision and early intervention typically improve efficiency and output quality as they lessen the effects of response randomness and the butterfly effect. We, therefore, also develop a web-based distillation assistant enabling supervision and flexible intervention at runtime. We hope our findings and tools could inspire future research toward revolutionizing the engineering of knowledge-based systems across application domains.
研究动机（中文总结）
全文第 2 页依据：中文翻译尚未生成，请点击“显示原文依据”查看英文内容。
Our contributions are as follows: • We are the first, to the best of our knowledge, to propose an empirical automation and semi-automation framework for domain knowledge distillation with LLMs. • We discuss our key observations and recommendations covering the entire distillation lifecycle in depth. • We present our web-based domain ontology distillation assistant to facilitate runtime human supervision, addressing the key challenges faced in the automatic ontology distillation experiment.
创新与贡献（中文总结）
全文第 2 页依据：中文翻译尚未生成，请点击“显示原文依据”查看英文内容。
In the next section, we present our empirical ontology distillation framework designed based on the required steps.
方法与证据
方法（中文总结）
全文第 2 页依据：中文翻译尚未生成，请点击“显示原文依据”查看英文内容。
In the next section, we present our empirical ontology distillation framework designed based on the required steps.
实验结果（中文总结）
全文第 5 页依据：中文翻译尚未生成，请点击“显示原文依据”查看英文内容。
Our repeated experiments show that while the ontologies of different experiment trials share many concepts, e.g., car, Pedestrian, and Driver, etc., those concepts have different hierarchical locations and relative distillation orders across the trails.