📄 中文摘要
Few-Shot学习不仅仅是提供示例，而是提供正确的示例。通过展示输入和输出的示例，可以有效地引导模型完成任务。然而，随机选择示例与构建一个真正能引导模型的示例集合之间存在巨大差异。在修复支持票分类器的过程中，初次尝试使用的示例过于简单，未能覆盖边界案例，导致效果不佳。因此，选择合适的示例至关重要，这将直接影响模型的表现和分类准确性。
高级提示工程：我在生产环境中破坏后学到的
📄 English Summary
Prompt Engineering Avanzado: Lo que aprendí después de dos semanas rompiendo producción
Few-Shot learning is not just about providing examples, but about providing the right examples. By showing input-output pairs to the model, one can effectively guide it to perform tasks. However, there is a significant difference between randomly selecting examples and constructing a set of examples that truly orient the model. In the process of fixing the ticket classifier, the initial attempt used overly simplistic examples that did not cover edge cases, leading to poor performance. Therefore, selecting appropriate examples is crucial, as it directly impacts the model's performance and classification accuracy.