botp/shareAI_ShareGPT-Chinese-English-90k
收藏资源简介:
ShareGPT-Chinese-English-90k是一个中英文平行双语的人机问答数据集,覆盖了真实复杂场景下的用户提问,用于训练高质量的对话模型。该数据集提供中英文平行对照语料,所有问题均非人为臆想或通过API轮询拟造的假数据,更加符合真实用户场景的指令分布和提问表达。数据集由网友自发分享收集,经过天然过滤筛除了大部分体验不好的对话。数据收集于chatGPT还未表现出明显智力退化的时间点,可能由于官方替换模型或引入更多拒绝答复导致模型连接知识逻辑的程度退化。
ShareGPT-Chinese-English-90k is a parallel bilingual (Chinese-English) human-machine question answering dataset that covers user queries in real and complex scenarios and is used for training high-quality dialogue models. This dataset provides parallel aligned Chinese-English corpora. All the questions are neither artificially fabricated nor generated via API polling, which better aligns with the instruction distribution and query expressions of real user scenarios. The dataset was collected voluntarily by netizens, and most poorly performing dialogues were naturally filtered out. The data was collected at a time when ChatGPT had not yet exhibited obvious intellectual degradation, which may have been caused by official model replacements or the introduction of more refusal responses that led to the degradation of the model's ability to connect with knowledge logic.
ShareGPT-Chinese-English-90k 中英文双语人机问答数据集
概述
中英文平行双语优质人机问答数据集,覆盖真实复杂场景下的用户提问。用于训练高质量的对话模型。
特点
- 双语对照:同时提供意义表达完全相同的中英文平行对照语料,可进行双语对话模型训练。
- 真实性:所有问题均非人为臆想加上api轮询拟造的假数据,更加符合真实用户场景的指令分布和提问表达。
- 天然过滤:sharegpt数据集是由网友自发分享而收集到的,相当于有一层非常天然的过滤(通过人类感觉),筛除了大部分体验不好的对话。
补充信息
该数据收集于chatGPT还未表现出明显智力退化的时间点。



