Isotonic/agentinstruct-1Mv1-combined

Orca-AgentInstruct-1M-v1-combined是Microsoft发布的orca-agentinstruct-1M-v1数据集的编辑版本。这是一个完全合成的数据集，仅使用网络上公开的原始文本作为种子数据，并且是创建Orca-3-Mistral的完整AgentInstruct数据集的一个子集。与Mistral 7B Instruct相比，该数据集在AGIEval、MMLU、GSM8K、BBH和AlpacaEval等多个评估指标上分别有40%、19%、54%、38%和45%的改进。为了确保与大多数框架的兼容性，字符串被转换为字典列表，并且根据类别使用了自定义的系统提示。

Orca-AgentInstruct-1M-v1-combined is an edited version of the orca-agentinstruct-1M-v1 dataset released by Microsoft. It is a fully synthetic dataset using only raw text publicly available on the web as seed data, and it is a subset of the full AgentInstruct dataset that created Orca-3-Mistral. Compared to Mistral 7B Instruct, the dataset shows improvements of 40% on AGIEval, 19% on MMLU, 54% on GSM8K, 38% on BBH, and 45% on AlpacaEval. To ensure compatibility with most frameworks, strings were converted into lists of dicts, and custom system prompts based on the category were used.