erfnzdeh/MCI-LLM-Security-Hackathon

Contest diary of team Prompt Police at MCI's LLM-security hackathon (Quera, Nov 2025): jailbreaking safety-guarded models across 14 forbidden domains via cross-lingual, transliteration & framing attacks.

语言：Jupyter Notebook · GitHub 星标：100 · Fork：0

话题：adversarial-attacks · ai-safety · farsi · hackathon · jailbreak · llm · llm-security · persian

在 GitHub 上查看原仓库