ZennyKenny/russian_llm_response_chatgpt_distill

LLM Usage RU Dataset 是一个包含 50,000 条俄语人类-LLM 交互日志的合成数据集。每个样本包括用户查询、LLM 的响应、时间戳、用户反馈和会话元数据。该数据集旨在探索大型语言模型在俄语中的表现，这是一种在训练覆盖面上通常不如英语受关注的语言。查询和响应是从 GPT-4-turbo 中提炼出来的，模拟了多样化的用户提示和回复，以反映广泛的交互模式。尽管有些样本看起来自然流畅，但其他样本则故意展示 LLMs 在俄语能力方面的不足。

The LLM Usage RU Dataset is a synthetic dataset of 50,000 Russian-language human–LLM interaction logs, including user queries, LLM responses, timestamps, user feedback, and session metadata. It is designed to explore the performance of large language models in Russian, a language that typically receives less training coverage than English. The queries and responses are distilled from GPT-4-turbo, simulating a variety of user prompts and replies to reflect a wide range of interaction patterns. Some samples appear natural and fluent, while others deliberately highlight the weaknesses in LLMs Russian capabilities.