📄 中文摘要
作者分享了自己在使用 AI 技术进行模型训练的经历，尽管并非数据科学家，但在处理简单分类问题时获得了信心。随后，朋友请求帮助解决更复杂的问题，涉及从 CRM 数据和电话录音中预测结果。作者介绍了 AUC 指标的基本概念，并描述了在尝试多种方法（如 XGBoost 和特征工程）后，模型的表现始终停留在 0.58 的 AUC 分数，反映出该任务的挑战性和复杂性。
Karpathy 的 Autoresearch 为我解锁的内容
出处: What Karpathy's Autoresearch Unlocked for Me
发布: 2026年3月31日
📄 English Summary
What Karpathy's Autoresearch Unlocked for Me
The author shares their experience with AI technology in model training, highlighting their confidence gained from handling simple classification problems despite not being a data scientist. A friend's request for assistance with a more complex task, involving predicting outcomes from CRM data and call recordings, led the author to explore various methods. They introduce the AUC metric and describe their attempts with techniques like XGBoost and feature engineering, ultimately finding their model's performance plateauing at an AUC score of 0.58, illustrating the challenges and complexities of the task.