遇见数据集

llm-jp-instructions-jculture-mcq

收藏
魔搭社区2026-05-14 更新2026-08-02 收录
官方服务:

资源简介:

# Japan-Specific Multiple-Choice Question Evaluation Dataset A multiple-choice variant of [llm-jp/llm-jp-instructions-jculture](https://huggingface.co/datasets/llm-jp/llm-jp-instructions-jculture). Each item asks a question about information specific to Japan, with four candidate answers; some items have a single correct answer and some have multiple correct answers. ## Definition of "Japan-specific" This refers to situations where the answer to a question differs between Japan and other countries. For example, whether cars drive on the left or the right side of the road: although some places outside Japan also have left-side driving, the rule is determined by each country, and there are countries where right-side driving is standard. Therefore, this can be considered Japan-specific. The dataset targets questions for which answers may differ between Japan and other countries due to social, cultural, or legal reasons. ## Dataset Statistics - **Total items**: 200 - **Single-correct-answer items**: 169 - **Multi-correct-answer items**: 31 - **Number of choices per item**: 4 The `id` field is aligned with the source `llm-jp-instructions-jculture` dataset (1..200). ## Schema | Field | Type | Description | | --- | --- | --- | | `id` | int64 | 1-indexed identifier matching the source `llm-jp-instructions-jculture` dataset | | `question` | string | The question stem with choices removed | | `choices` | list[string] | Four answer candidates with original labels stripped | | `answer` | list[int64] | Zero-indexed positions of the correct choice(s); length 1 for single-correct items | | `rationale` | string | Annotator-written explanation of why the answer is correct | | `reference_url` | string | Source URL(s) backing the answer (official) | ## License CC BY 4.0, inherited from the source dataset. ## Send Questions to llm-jp(at)nii.ac.jp ## Model Card Authors Shintaro Horie, Namgi Han, Kouta Nakayama, Takashi Kodama, Yusuke Miyao.

提供机构:
maas
创建时间:
2026-05-13
二维码
社区交流群
二维码
科研交流群
商业服务