Datasetpublic

julius8787/helpfulness_test_set

Dataset Card: Helpfulness Classification Based on ALERT Dataset Dataset Description This dataset is derived from the ALERT dataset and has been labeled to assess whether responses in question-answer pairs are helpful or not. Key Features: Helpfulness Labeling: Each answer is classified as either: Helpful: This includes both positive and supportive answers as well as well-justified rejections. Not Helpful: Answers that lack relevance, clarity, or a… See the full description on the dataset page: https://huggingface.co/datasets/julius8787/helpfulness_test_set.

sourceHugging Faceupdated 2y agoView on Hugging Face

0likes5downloads

Dataset Card

Dataset Card: Helpfulness Classification Based on ALERT Dataset

Dataset Description

This dataset is derived from the ALERT dataset and has been labeled to assess whether responses in question-answer pairs are helpful or not.

Key Features:

—Helpfulness Labeling: Each answer is classified as either:
—Helpful: This includes both positive and supportive answers as well as well-justified rejections.
—Not Helpful: Answers that lack relevance, clarity, or a justified basis for rejection.
—Special Consideration: Rejections are not inherently marked as unhelpful. Rejections with proper reasoning and context are classified as helpful.

Dataset Structure:

The dataset includes:

—question: The original question posed.
—answer: The provided response.
—label: A binary classification:
—1 for helpful
—0 for not helpful

Usage

This dataset is suitable for:

—Testing models for helpfulness classification in text-based question-answer systems.
—Improving Large Language Models (LLMs) to provide responses that are not only accurate but also contextually and ethically appropriate.

Licensing and Source

—Source Dataset: Babelscape/ALERT