tutorqa
收藏资源简介:
TutorQA Benchmark数据集是用于评估推理、图理解和语言生成不同方面的6个任务的集合。每个任务都是一个独立的数据分割,包括关系判断、先决条件预测、路径搜索、子图补全、聚类和想法仓鼠(开放性问题,没有答案)。每个任务都包含问题和答案字段,除了想法仓鼠任务只有问题字段。
The TutorQA Benchmark is a collection of 6 tasks aimed at evaluating different dimensions of reasoning, graph understanding, and language generation. Each task forms an independent data split, covering relational judgment, prerequisite prediction, pathfinding, subgraph completion, clustering, and Thought Hamster (an open-ended question task with no answer provided). All tasks include both question and answer fields, with the sole exception of the Thought Hamster task, which only contains the question field.
TutorQA 数据集概述
基本信息
- 许可证: MIT
- 下载大小: 55,342 字节
- 数据集大小: 406,643 字节
数据集结构
- 特征:
question: 字符串类型answer: 字符串类型(task6除外)
- 任务划分:
task1: 250个样本,100,788字节task2: 250个样本,42,363字节task3: 250个样本,67,642字节task4: 250个样本,146,014字节task5: 100个样本,22,327字节task6: 100个样本,27,509字节
任务描述
- task1: Relation Judgment
- task2: Prerequisite Prediction
- task3: Path Searching
- task4: Subgraph Completion
- task5: Clustering
- task6: Idea Hamster(无答案,开放式问题)
使用示例
python from datasets import load_dataset
dataset = load_dataset("li-lab/tutorqa")
访问单个任务
task1 = dataset["task1"] task6 = dataset["task6"]
引用信息
bibtex @inproceedings{yang2025graphusion, title={Graphusion: A RAG Framework for Knowledge Graph Construction with a Global Perspective}, author={Yang, Rui and Yang, Boming and Feng, Aosong and Ouyang, Sixun and Blum, Moritz and She, Tianwei and Jiang, Yuang and Lecue, Freddy and Lu, Jinghui and Li, Irene}, booktitle={Proceedings of the NLP4KGC Workshop at The Web Conference 2025 (WWW25)}, year={2025}, url={https://arxiv.org/abs/2410.17600} }
相关资源




