Nemotron Sft Safety V2: Prompts and Responses for LLM Safety Alignment

A collection of prompts and synthetically generated responses designed to align large language models with safety and security values. The dataset was created by NVIDIA and was last updated on June 4, 2026. It likely contains a hybrid of open-source and synthetic prompts targeting various model vulnerabilities.

Use Cases

• Training safety classifiers based on the dataset's prompts designed to elicit vulnerabilities.
• Fine-tuning LLMs for robustness using the synthetically generated safety-aligned responses.
• Benchmarking model resilience against adversarial safety prompts mentioned in the description.
• Developing red-teaming tools based on the collection of prompts targeting different vulnerabilities.

Strengths

• Created by NVIDIA, a leading AI research organization.
• Last updated on June 4, 2026, indicating recent maintenance.
• Designed to address a variety of safety and security concerns for LLMs.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

NVIDIA

Collection Method

Collection includes a hybrid of open-source prompts and synthetically generated content.

Time Range

null

Freshness

Last updated 2026-06-04 04:49:04.

Geography

null