📄 中文摘要
六个月前，团队为一家120名员工的B2B SaaS公司构建内部支持工具，目标是开发一个能够回答内部流程问题的聊天机器人。面临选择：使用RAG还是微调模型。经过六个月的探索，团队在三个不同项目中同时运行这两种方法，发现大多数比较未能准确反映实际需求，特别是在特定场景下的适用性和潜在问题。最终的经验教训揭示了在选择技术时需要考虑的细节和实际应用中的挑战。
RAG与微调：我在构建大型语言模型应用六个月后学到的东西
📄 English Summary
RAG vs Fine-Tuning: What I Actually Learned After 6 Months of Building LLM Apps
Six months ago, a team was tasked with building an internal support tool for a B2B SaaS company with around 120 employees. The goal was to create a chatbot capable of answering questions about internal processes without fabricating information. Faced with the decision between using RAG or fine-tuning a model, the team spent six months exploring both approaches across three different projects. The findings revealed that most comparisons fail to address the specific needs of a situation, particularly regarding applicability and potential pitfalls. Ultimately, the lessons learned highlight the nuances and challenges involved in choosing the right technology for practical applications.