Gemini-generated paraphrases

'Gemini-generated paraphrases' refers to a dataset containing 300 paraphrased sentences created using the Gemini model, which is used to evaluate the robustness and brittleness of defenses against paraphrased inputs in the context of LLM security threats.

Papers using Gemini-generated paraphrases (1)

Which Defense Closes Which Threat? Attributing OWASP-LLM-Top-10 Coverage and Its Brittleness Under Paraphrasing2026