|
Download README.md from HYUKI5/tasksource__nlgraph: direct link, hf CLI and curl.
- Browser
- Download file 4.73 kB
-
https://huggingface.co/datasets/HYUKI5/tasksource__nlgraph/resolve/main/README.md
- Command line
-
hf download hf://datasets/HYUKI5/tasksource__nlgraph/README.md
-
curl -L -o README.md https://huggingface.co/datasets/HYUKI5/tasksource__nlgraph/resolve/main/README.md
4.73 kB
metadata
license: other
pretty_name: tasksource/nlgraph (mirror)
source_datasets:
- tasksource/nlgraph
tags:
- laya
- laya-finetune
- mirror
- domain-math
- use-unclear
- genre-other
- label-text
- annot-unknown
- lang-en
dataset_info:
features:
- name: question
dtype: string
- name: answer
dtype: string
- name: difficulty
dtype: string
- name: task
dtype: string
splits:
- name: train
num_bytes: 6593495
num_examples: 5022
- name: test
num_bytes: 1270263
num_examples: 1000
download_size: 1448275
dataset_size: 7863758
tasksource/nlgraph(複製)
データの内容
言語モデルが自然言語のグラフ問題を解けるかを調べる Wang らの研究(arXiv:2305.10037)のデータ。1 行が 1 問(question)で、答え answer、難易度 difficulty、課題の種類 task を持つ。
データの概要
| 項目 | 値 |
|---|---|
| 行数 | 6,022(1 構成・2 分割の合計) |
| 列数 | 4 |
| 容量 | 1.45 MB(ファイル 4 本) |
| ファイルの形式 | parquet 2 |
| 公開 | public |
構成と分割
| 構成(config) | 分割(split)ごとの行数 | 列数 |
|---|---|---|
default |
train 5,022、test 1,000 |
4 |
列(構成 default・分割 train)
| 列 | 型 | 先頭の行の値 |
|---|---|---|
question |
string |
In an undirected graph, the nodes are numbered f… |
answer |
string |
The answer is: node 0: [0,1] node 1: [1,3] node … |
difficulty |
string |
easy |
task |
string |
GNN |
行数と列は HF のデータセットビューアの集計(元のデータセットの現在の版で数えたもの)。全データセットの一覧は HYUKI5/dataset-catalog にある。
複製について
tasksource/nlgraph をリビジョン e8a405c296cf9c3be1c995d9460bf1e4ab04b9b6 の時点でそのまま複製したもの。中身は変えていない。判断モデル Laya(convaiinnovations/laya)の微調整に使う候補として、調査メモ(Laya 学習用データセット調査)に載せたデータを集めている。
出典
- 元のデータセット: https://huggingface.co/datasets/tasksource/nlgraph
- 複製したリビジョン:
e8a405c296cf9c3be1c995d9460bf1e4ab04b9b6(2023-05-23 07:36:04+00:00) - 元のライセンス表記:
other
論文
アノテーションの方法
元のカードにアノテーションの節が無かった。下の「調査メモの記述」と「元のカード」を参照。
微調整に使うときの注意
- 調査メモの台帳での区分: D(記載なし・other・unknown・資料間の食い違い)。区分は記載の文言で機械的に付けたもので、法的な判定ではない。学習に入れる前に元のカードとライセンスの原文で確かめる。
- Laya の入力は 2,048 token まで(ModernBERT-large / mmBERT)。長い本文は先頭から黙って切られるので、切れる行は事前に除くか、要約・切り出しをしてから入れる。
- 分割は記事・文書単位で行い、同じ本文から作った別の問いを学習側と較正側に分けない。温度と閾値は学習に使っていない分割で別々に当てる。
調査メモの記述
| メモ | 節 | ライセンス(記載) | 確認 | 注記 |
|---|---|---|---|---|
jev_sources_tasksource_bekko.csv |
CSV(tasksource-jev-typed-decisions) | unspecified / license_use=unspecified | 確認済 | license_use=unspecified / [Jev 元データ source] source_annotation_converted(deterministic recast; no model r |
元のカード(日本語訳)
以下は元のデータセットのカードの本文を日本語に訳したもの(機械翻訳)。訳に疑問があるときは、末尾の原文を確かめる。
データセットカード: "nlgraph"
@article{wang2023can,
title={Can Language Models Solve Graph Problems in Natural Language?},
author={Wang, Heng and Feng, Shangbin and He, Tianxing and Tan, Zhaoxuan and Han, Xiaochuang and Tsvetkov, Yulia},
journal={arXiv preprint arXiv:2305.10037},
year={2023}
}