The rapid advancements in pre-trained Large Language Models (LLMs) and Large Multimodal Models (LMMs) have ushered in a new era of intelligent applications, transforming fields ranging from natural language processing to content generation. The LLM supply chain represents a crucial aspect of the contemporary artificial intelligence landscape. It encompasses the entire lifecycle of pre-trained models, from its initial development and training to its final deployment and application in various domains. This paper presents a comprehensive overview of the LLM supply chain, highlighting its three core elements: 1) the model infrastructure, encompassing datasets and toolchain for training, optimization, and deployment; 2) the model lifecycle, covering training, testing, releasing, and ongoing maintenance; and 3) the downstream application ecosystem, enabling the integration of pre-trained models into a wide range of intelligent applications. However, this rapidly evolving field faces numerous challenges across these key components, including data privacy and security, model interpretability and fairness, infrastructure scalability, and regulatory compliance. Addressing these challenges is essential for harnessing the full potential of LLMs and ensuring their ethical and responsible use. This paper provides a future research agenda for the LLM supply chain, aiming at driving the continued advancement and responsible deployment of these transformative LLMs.

翻译：预训练大型语言模型（LLMs）与大型多模态模型（LMMs）的快速发展，开启了智能应用的新纪元，推动了从自然语言处理到内容生成等多个领域的变革。LLM供应链作为当代人工智能生态系统中的关键环节，涵盖了预训练模型从初始开发训练到最终部署应用的全生命周期。本文对LLM供应链进行了全面综述，重点阐述其三大核心要素：1）模型基础设施，涵盖用于训练、优化与部署的数据集及工具链；2）模型生命周期，包括训练、测试、发布及持续维护；3）下游应用生态系统，支持将预训练模型集成至各类智能应用中。然而，这一快速发展的领域在数据隐私与安全、模型可解释性与公平性、基础设施可扩展性及合规监管等关键环节仍面临诸多挑战。解决这些问题对于充分发挥LLMs的潜力、确保其伦理与负责任使用至关重要。本文提出了LLM供应链的未来研究议程，旨在推动这些变革性大模型的持续进步与负责任部署。