by LLM-jp
AnswerCarefully is LLM-jp's Japanese instruction dataset for the safety and appropriateness of LLM output. It pairs hand-written questions with reference safe answers.
DatasetFree
Step 1: Sign in to Hugging Face and accept the dataset's terms of use on its page.
Step 2: Install the datasets library.
Step 3: Call load_dataset with the repository name llm-jp/AnswerCarefully, a version such as v3.0, and a split such as dev or test.
Step 4: Use the question and reference answer pairs to fine-tune or evaluate a model's safety behavior.
Step 5: If you publish derived data, state that it was built from this dataset and keep the original samples out of it.
It is meant for improving the safety of LLMs in Japanese and other languages. The terms forbid using it for anything other than safety improvement, such as getting around safety measures.
No, redistribution of the data is not allowed. You may create and share derived data such as translations or synthetic prompts if it does not contain the original data and credits this dataset.
The card says the dataset is public for its safety purpose, including commercial use, under its own terms of use. Access requires accepting those terms on Hugging Face.