LLM-Tuning-Safety/HEx-PHI
Text GenerationEN
LLM-Tuning-Safety/HEx-PHI is a text generation-focused dataset in EN distributed in Parquet format.
About LLM-Tuning-Safety/HEx-PHI
HEx-PHI: Human-Extended Policy-Oriented Harmful Instruction Benchmark
This dataset contains 330 harmful instructions (30 examples x 11 prohibited categories) for LLM harmfulness evaluation.
In our work "Fine-tuning Aligned Language Models Compr...
Details
- Task
- Text Generation
- Language
- EN
- Format
- Parquet
- Rows / instances
- N/A
- Creator
- LLM-Tuning-Safety
- Year
- 2023