Phi-3 安全后训练：通过“破坏-修复”循环对齐语言模型
Phi-3 Safety Post-Training: Aligning Language Models with a "Break-Fix" Cycle
AI总结：
本文提出针对Phi-3系列语言模型的安全后训练方法论，采用“破坏-修复”循环执行多轮数据构建、安全训练、基准测试与红队测试，有效提升了模型在单多轮场景及多语言能力上的安全对齐性能。
摘要：
语言模型训练的最新创新表明，创建性能极高且小到足以在智能手机上运行的模型是可能的。随着这些模型在越来越多的领域部署，确保它们与人类偏好和安全考量保持一致至关重要。在本报告中，我们介绍了对 Phi-3 系列语言模型进行安全对齐的方法论。我们采用了“破坏-修复”循环，执行多轮数据集构建、安全后训练、基准测试、红队测试和漏洞识别，以覆盖单轮和多轮场景中的各种危害领域。我们的结果表明，这种方法在各种负责任 AI 基准上迭代地提升了 Phi-3 模型的性能。最后，我们包含了用于测试 Phi-3.5-mini 和 Phi-3.5-MoE 安全行为的额外红队策略和评估，这些模型针对多语言能力进行了优化。
英文摘要：
Recent innovations in language model training have demonstrated that it is possible to create highly performant models that are small enough to run on a smartphone. As these models are deployed in an increasing number of domains, it is critical to ensure that they are aligned with human preferences and safety considerations. In this report, we present our methodology for safety aligning the Phi-3 series of language models. We utilized a "break-fix" cycle, performing multiple rounds of dataset curation, safety post-training, benchmarking, red teaming, and vulnerability identification to cover a variety of harm areas in both single and multi-turn scenarios. Our results indicate that this approach iteratively improved the performance of the Phi-3 models across a wide range of responsible AI benchmarks. Finally, we include additional red teaming strategies and evaluations that were used to test the safety behavior of Phi-3.5-mini and Phi-3.5-MoE, which were optimized for multilingual capabilities.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
