Cyber Task Horizons: Human Study Data for AI Offensive Cybersecurity Capability

Offensive Cyber Task Horizons contains expert completion and estimation data from a human study described in a 2026 paper by Payne, Miller, and Peters. The dataset provides human-derived timing data that anchors the Item Response Theory (IRT) methodology for measuring AI offensive cybersecurity capability growth. For full reproducibility artifacts, including model logs and analysis pipelines, the dataset page references a GitHub repository.

Use Cases

• Benchmarking AI offensive cybersecurity capability growth rates based on human-derived timing data.
• Applying Item Response Theory (IRT) models to cybersecurity task performance data.
• Comparing expert human completion times against AI model performance on similar tasks.
• Reproducing the analysis from the 'Offensive Cyber Task Horizons' 2026 research paper.

Strengths

• Data is explicitly linked to a 2026 research paper, providing academic context.
• Dataset is hosted on Hugging Face with a CC BY 4.0 license, indicating open access.
• Platform tags indicate the data is optimized in Parquet format for efficient use.

Limitations

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

Provenance

Source

lyptus-research

Collection Method

Expert completion and estimation data from a human study.

Time Range

Data is associated with a 2026 publication; specific collection period is not stated.

Freshness

Last updated 2026-04-02 22:52:22.

Geography

Region is tagged as 'us'; specific geographic coverage is not detailed.