The experimental data of FigStep.
CONTENT WARNING: Some file contains potential harmful content.
Some files have been removed because they contain harmful content. If you need the removed data, please contact the paper authors.
| Path | Description |
|---|---|
| question | SafeBench, a jailbreak evaluation dataset, generated by GPT-4 |
| response (removed) | The response generated by different and LLMs and VLMs |
| evaluation (removed) | The harmfulness of model response, evaluated by GPT-4 |
| result (removed) | The statistics of harmfulness |