EMRA
Emerging1papers using it
2026first seen
The EMRA benchmark contains 5,200 adversarial samples across 8 attack types and is used to evaluate the effectiveness of defenses against evolving multi-round attacks on Large Language Models (LLMs).
The EMRA benchmark contains 5,200 adversarial samples across 8 attack types and is used to evaluate the effectiveness of defenses against evolving multi-round attacks on Large Language Models (LLMs).