Gemini-flash-2.5

The 'Gemini-flash-2.5' dataset/benchmark is used to evaluate the effectiveness of safety alignment in Large Language Models (LLMs) against jailbreak attacks that exploit vulnerabilities in content generation and unsafety reasoning.

Papers using Gemini-flash-2.5 (1)

The Trojan Example: Jailbreaking LLMs through Template Filling and Unsafety Reasoning2025