All datasets / hirundo-io
hirundo-io/secfid-agentic-refusal-qa
The declared licence (apache-2.0) is classified as permissive.
- Declared licence
- apache-2.0, from the tagEvidence: README.md at baa8545
- Licence files
LICENSE, detected asapache-2.0Evidence: LICENSE at baa8545- Declared sources
- None listed in the card metadata
- Access gated on Hugging Face
- No
Tasks
Citation text from the card
Run `python -m examples.build_agentic_refusal_dataset` from [llm-behavior-eval](https://github.com/Hirundo-io/llm-behavior-eval). `manifest.json` records source and generated-file checksums. The conversion uses all source rows, sorts by source ID, and emits clean then injected variants. No model-generated data or answers are added. SecFid is Apache-2.0. Its agentic extension derives from [InjecAgent](https://github.com/uiuc-kang-lab/InjecAgent) (MIT). The original LICENSE and THIRD_PARTY_LICENSES notices are included unchanged. Changes by Hirundo: Q&A serialization, explicit paired clean controls, reference answer rendering, and refusal-task metadata. Source benchmark examples may contain fictional identifiers and deliberately malicious instructions.
The same facts for your agent
A free MCP server at mcp.datalicense.uk/mcp. Connect it from an MCP client such as Claude or Cursor.
dataset_licence({ "id": "hirundo-io/secfid-agentic-refusal-qa" })
signal: "permissive"
stated_licence: "apache-2.0"
sources: 0