Shallow Vs Deep Safety Alignment: Derivatives of HEx-PHI for LLM Tuning

A text dataset derived from the LLM-Tuning-Safety/HEx-PHI dataset, intended for research on large language model safety alignment. The dataset was created by Unispac and last updated on April 23, 2025. Its specific content and scale are not detailed in the provided metadata.

Use Cases

• Training safety classifiers based on the derived text content.
• Benchmarking safety alignment techniques using the HEx-PHI derivatives.
• Analyzing differences between shallow and deep safety alignment in model responses.

Strengths

• Derived from the established HEx-PHI dataset for LLM safety tuning.
• Last updated on April 23, 2025, indicating recent maintenance.

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 and file formats are unknown, which may limit suitability assessment.

Provenance

Source

Derived from LLM-Tuning-Safety/HEx-PHI dataset.

Freshness

Last updated 2025-04-23 03:49:36