annotations_creators: []
language:
- en
- kn
- th
language_creators: []
license: []
multilinguality:
- multilingual
pretty_name: adaption-cyber_sec_code_instruct
size_categories:
- 1K<n<10K
source_datasets: []
tags:
- adaption
- instruction-tuning
- code
- governance
- technology
task_categories: []
task_ids: []
This dataset is a remastered version prepared using Adaption's Adaptive Data platform.
adaption-cyber_sec_code_instruct
This dataset consists of instruction-response pairs focused on cybersecurity incident analysis, threat detection, and secure software engineering. Tasks span multiple programming languages, including Python, JavaScript, C, SQL, and Rust, covering vulnerability assessment, remediation, and log investigation. Prompts range from straightforward coding tasks to complex incident scenarios with missing context, evaluating technical reasoning and cross-language handling.
Dataset size
There are 7,959 data points in this dataset. This is an instruction tuning dataset.
Quality of Remastered Dataset
The final quality is A, with a relative quality improvement of 2.2%.
Domain
- Code (92%)
- Governance (6%)
- Technology (2%)
Language
- English (12%)
- Kannada (10%)
- Thai (8%)
Tone
- Technical (90%)
- Educational (4%)
- Urgent (2%)
Evaluation Results
Quality Gains:
Grade Improvement:
Percentile Chart:
