Datasetpublic

jasonwu1017/s1k-1.1-text-kg-dataset

S1K Text-KG Dataset This dataset contains questions with: Knowledge graphs generated using GPT-4o Thinking trajectories from DeepSeek Formatted answers Dataset Structure The dataset contains 630 examples with the following fields: question: The original question deepseek_thinking_trajectory: Step-by-step reasoning from DeepSeek deepseek_attempt: The answer from DeepSeek deepseek_grade: Evaluation grade ("Yes" for all examples in this dataset) gpt-4o_graph:… See the full description on the dataset page: https://huggingface.co/datasets/jasonwu1017/s1k-1.1-text-kg-dataset.

sourceHugging Faceupdated 1y agoView on Hugging Face

0likes11downloads

Dataset Card

S1K Text-KG Dataset

This dataset contains questions with:

—Knowledge graphs generated using GPT-4o
—Thinking trajectories from DeepSeek
—Formatted answers

Dataset Structure

The dataset contains 630 examples with the following fields:

—question: The original question
—deepseek_thinking_trajectory: Step-by-step reasoning from DeepSeek
—deepseek_attempt: The answer from DeepSeek
—deepseek_grade: Evaluation grade ("Yes" for all examples in this dataset)
—gpt-4o_graph: Knowledge graph generated using GPT-4o, with nodes and edges
—text: Formatted text that combines all elements in a chat format

Example Usage

python

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("s1k-text-kg-dataset")

# Access the first example
example = dataset["train"][0]

# Print the question
print(example["question"])

# Get the knowledge graph
kg = example["gpt-4o_graph"]

# Get the formatted text
text = example["text"]

Dataset Creation

This dataset was created by extracting entities and relations from questions and answers using GPT-4o. The knowledge graphs represent the conceptual relationships needed to answer the questions.