Sign in to view source links and access this dataset
Description
Uka Cyber Dataset is a text dataset in ChatML format containing 81,958 entries for cybersecurity reasoning and tool calling. It includes 70,802 training entries, split into 63,216 question-answer pairs and 7,586 pentest tool call traces. The dataset was created by author 'hotdogs' and was last updated on July 1, 2026.
Use Cases
Fine-tuning language models for cybersecurity question-answering based on the described reasoning tasks.
Training models to generate or understand penetration testing tool call sequences based on the included tool call traces.
Developing agents that can reason about security vulnerabilities and execute appropriate tools based on the combined Q&A and tool call data.
Strengths
Contains 70,802 training entries, providing a substantial base for model training.
Includes a specific subset of 7,586 pentest tool call traces, which may be useful for specialized tool-calling tasks.
Provides separate validation (7,437 entries) and test (3,719 entries) splits for model evaluation.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Description metadata is limited; actual data quality requires manual inspection after download.
The dataset's source, collection method, and potential biases are not described.
Provenance
Source
huggingface user 'hotdogs'
Freshness
Last updated 2026-07-01 06:52:03; freshness should be verified.
License is unknown; users should verify permissions before use.