Oriel — AI & technology worth noticing, every day.

Reizxn/HolyCyber

Dataset · https://huggingface.co/datasets/Reizxn/HolyCyber

HolyCyber — defensive cybersecurity instruction dataset A long-form, fully grounded instruction dataset for a defensive cybersecurity assistant. Every row is a self-contained briefing of at least 2,000 tokens (user + assistant, tiktoken cl100k_base), written from authoritative public sources and shipped with the provenance needed to check it. At a glance Rows 51,805 (global) Tokens 127,643,325 (user + assistant) Domains 17 focused layers, each ≥2… See the full description on the dataset page: https://huggingface.co/datasets/Reizxn/HolyCyber.

First seen by Oriel on .

AI summary

HolyCyber 2.3.0 is a grounded defensive cybersecurity instruction dataset containing 51,805 long-form rows and 127,643,325 tokens across 17 independently usable domains.

  • Each JSONL record uses a three-turn chat format and includes measured token counts, core_share, appendix status, references, licensing, split identity, and a decontamination marker.
  • The dataset provides 40,998 training, 5,105 validation, and 5,702 test rows, with group_key-based splitting intended to prevent source-entity leakage.
  • Build checks enforce at least 2,000 tokens per row, unique questions and answers, authoritative-source grounding, defensive-only content, and no benchmark 25-gram overlap.
  • Reference appendices appear in 40,144 rows and contain reused supporting material; users can filter by core_share or mask labelled appendix blocks during fine-tuning.
  • Limitations include no row-by-row expert review, occasional RFC parsing errors, English-only content, uneven task diversity, and source-specific licensing obligations.

#cybersecurity #instruction tuning #datasets #data provenance

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Likes

Total likes reported by the source.

  • Hugging Face Datasets: 33, latest observation ; +23 over observed period since

Downloads (30d)

Rolling 30-day downloads. A change in this value is not new downloads during the observed period.

  • Hugging Face Datasets: 777, latest observation ; +777 over observed period since

Aliases

SourceExternal IDURL
Hugging Face DatasetsReizxn/HolyCyberhttps://huggingface.co/datasets/Reizxn/HolyCyber Read discussion