Toxicity Debug Text Dataset

Toxicity Debug is a text dataset for evaluating language model safety, created by researcher tomekkorbak and hosted on Hugging Face. It was last updated in April 2022. The dataset's size is categorized as 'n1 K', indicating it contains over 1,000 entries.

Use Cases

• Fine-tuning a toxicity classifier using the provided text labels.
• Benchmarking language model outputs against annotated toxic content.
• Analyzing patterns in toxic text features for model debugging.

Strengths

• Contains over 1,000 text entries for analysis.
• Stored in Parquet format for efficient data processing.

Limitations

• Specific row count and column details are unknown.
• Data is from 2022 and may not reflect current language patterns.
• Geographic and demographic coverage is unspecified.

Provenance

Source

Hugging Face, author tomekkorbak.

Freshness

Last updated April 2022.

Geography

Region tag suggests US focus, but unspecified.