LLM Jailbreaking Guide主流大语言模型越狱指南

[ACL 2025] Guiding not Forcing: Enhancing the Transferability of Jailbreaking Attacks on LLMs via Removing Superfluous Constraints

Code of paper: xJailbreak: Representation Space Guided Reinforcement Learning for Interpretable LLM Jailbreaking"

Code for "When LLM Meets DRL: Advancing Jailbreaking Efficiency via DRL-guided Search" (NeurIPS 2024)

[AAAI-2026]MAJIC: Markovian Adaptive Jailbreaking. An automated black-box attack framework against LLMs that iteratively selects and fuses innovative disguise strategies guided by a dynamically updated Markov transition matrix.

A guide for every LLM jailbreaker. Learn, Test and Break!

The ultimate collection of bunch of LLM models for example GPT-4, Gemini, Claude prompts. Includes prompt engineering guides, productivity templates, developer modes, jailbreaks (JB), and system overrides for testing AI safety. - Educational Purposes Only!

LLM security threats and mitigations: prompt injection, data leakage, model poisoning, jailbreaking. Enterprise AI security checklist and on-premise deployment guide.

Defensive guide for testing and securing LLM-integrated applications against prompt injection, API misuse, data leakage, and jailbreak attempts.

This is a beginner-intermediate level report for people who are interested in LLM conditioning, probing, and general understanding of the fundamentals. This is NOT a guide on how to jailbreak LLMs. This report has a synopsis at the top - refer to that for a more detailed description.