mlfoundations-dev/instruction_filtering_scale_up_code_base_gemini_length_16K

该数据集包含多个字段，如指令种子(instruction_seed)、来源(source)、Gemini模型响应(gemini_response)等。数据集被划分为训练集(train)，包含16000个示例。每个示例都包含一系列的特征，如推理过程(reasoning)、DeepSeek解决方案(deepseek_solution)、最终推理跟踪(final_reasoning_trace)以及对话(conversations)信息。对话信息包括对话的发起者(from)和对话内容(value)。

The dataset includes multiple fields such as instruction_seed, source, gemini_response, etc. The dataset is split into a training set (train) containing 16,000 examples. Each example has a series of features, including reasoning process, DeepSeek solution, final reasoning trace, and conversation information. Conversation information includes the initiator (from) and content (value) of the conversation.