Answerable-or-Not is an open-source dataset for training deep learning models to determine prompt answerability. It contains 2,440 labeled textual prompts curated based on a hierarchical safety taxonomy, with 40 prompts per category. The dataset was created by kalyannakka and last updated on Hugging Face in April 2026.
Use Cases
- Training answerability classifiers based on labeled textual prompts.
- Evaluating AI safety taxonomies based on the hierarchical structure mentioned.
- Benchmarking prompt-filtering models based on the balanced yes/no label distribution.
- Studying the lower-level categories of safety taxonomies for content moderation.
Strengths
- Contains 2,440 labeled textual prompts.
- Balanced dataset with 20 YES and 20 NO labels per category.
- Structured around a hierarchical safety taxonomy.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
Provenance
- Source
- Hugging Face dataset by kalyannakka.
- Collection Method
- Curated based on the lower level of Do-Not-Answer's hierarchical safety taxonomy.
- Freshness
- Last updated 2026-04-17 02:23:51; freshness should be verified.