s-nlp/Llama-3.1-8B-Instruct-TriviaQA-HighlyKnown
收藏Hugging Face2025-02-24 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/s-nlp/Llama-3.1-8B-Instruct-TriviaQA-HighlyKnown
下载链接
链接失效反馈官方服务：
资源简介：
这是一个基于TriviaQA的数据集，用于研究在不损害大型语言模型（LLM）的情况下，能在LoRA适配器中填充多少知识。数据集包含问题、问题来源、预问问题、贪婪答案以及答案的相关信息，如别名、匹配的维基实体名称、标准化别名、标准化匹配的维基实体名称、标准化值、类型和值。数据集分为完整集，共有90981个示例。
This dataset is based on TriviaQA and is used for studying how much knowledge can be packed into a LoRA Adapter without harming Large Language Models (LLM). It contains information such as questions, sources, primed questions, greedy answers, and details about the answers including aliases, matched wiki entity names, normalized aliases, normalized matched wiki entity names, normalized values, types, and values. The dataset is split into a full set with a total of 90981 examples.
提供机构：
s-nlp
