LLM Service Outages and Incidents Dataset
收藏资源简介:
该数据集由阿姆斯特丹自由大学的研究团队创建,旨在分析大型语言模型(LLM)服务的故障和恢复过程。数据集涵盖了8个常用的LLM服务,包括OpenAI的ChatGPT、DALL·E、Playground,Anthropic的Claude、Console,以及Character.AI的服务。数据集包含从2021年至2024年的长期故障和恢复数据,记录了故障次数、故障持续时间、故障恢复时间等关键指标。数据来源主要为LLM服务提供商公开的故障报告和用户反馈。该数据集的应用领域包括LLM系统的可靠性分析、故障恢复优化以及服务质量提升,旨在帮助研究人员和开发者更好地理解和改进LLM系统的稳定性与性能。
This dataset was developed by a research team at Vrije Universiteit Amsterdam to analyze the failure and recovery processes of Large Language Model (LLM) services. It covers eight widely utilized LLM services, including OpenAI's ChatGPT, DALL·E, and Playground, Anthropic's Claude and Console, as well as services provided by Character.AI. The dataset contains long-term failure and recovery data spanning from 2021 to 2024, recording key metrics such as the number of failures, failure duration, and recovery time. The data is primarily sourced from public failure reports and user feedback released by LLM service providers. The application areas of this dataset include reliability analysis of LLM systems, failure recovery optimization, and service quality enhancement, with the goal of assisting researchers and developers in better understanding and improving the stability and performance of LLM systems.




