MTJ-Pro

Emerging

1papers using it

2026first seen

MTJ-Pro is a challenging multi-turn progressive jailbreak dataset that contains seven advanced jailbreak strategies designed to evaluate the effectiveness of deceptive defenses against evolving multi-turn jailbreak attacks on large language models.

🔎 Find this dataset

Papers using MTJ-Pro (1)

HoneyTrap: Deceiving Large Language Model Attackers to Honeypot Traps with Resilient Multi-Agent Defense2026