面向生产环境的可信赖基础模型驱动软件指南
软件工程
2025-06-04 v2 人工智能
机器学习
摘要
基础模型（FMs）如大型语言模型（LLMs）正在重塑软件行业，通过构建将这些 FMs 作为核心组件的系统（即 FMware）实现其潜力。在本 KDD 2025 教程中，我们综合探讨了 FMware，将精选挑战目录与实际生产关切相结合。我们首先讨论构建 FMware 的研究与实践现状。我们进一步检视了选取合适模型、对齐高质量领域特定数据、工程化稳健提示词编写，以及编排自主代理等难题。我们随后阐述了从令人印象深刻的演示走向生产就绪系统的复杂路径，概述了系统测试、优化、部署以及与传统软件集成等问题。drawing on our industrial experience and recent research in the area, we provide actionable insights and a technology roadmap for overcoming these challenges. Attendees will gain practical strategies to enable the creation of trustworthy FMware in the evolving technology landscape.
引用
@article{arxiv.2505.10640,
title = {The Hitchhikers Guide to Production-ready Trustworthy Foundation Model powered Software (FMware)},
author = {Kirill Vasilevski and Benjamin Rombaut and Gopi Krishnan Rajbahadur and Gustavo A. Oliva and Keheliya Gallaba and Filipe R. Cogo and Jiahuei Lin and Dayi Lin and Haoxiang Zhang and Bouyan Chen and Kishanthan Thangarajah and Ahmed E. Hassan and Zhen Ming Jiang},
journal= {arXiv preprint arXiv:2505.10640},
year = {2025}
}