Learn PromptingRingkasan
Mencegah injeksi prompt bisa sangat sulit, dan hanya ada sedikit pertahanan kuat terhadapnya. Namun, ada beberapa solusi yang masuk akal. Misalnya, jika aplikasi Anda tidak perlu menghasilkan teks bebas, jangan izinkan keluaran semacam itu. Ada banyak cara berbeda untuk melindungi sebuah prompt. Kita akan membahas beberapa yang paling umum di sini.
Bab ini mencakup strategi-strategi umum tambahan seperti menyaring kata-kata. Bab ini juga mencakup strategi peningkatan prompt (pertahanan instruksi, pengiriman pasca, cara berbeda untuk melampirkan input pengguna, dan penandaan XML). Akhirnya, kita membahas menggunakan LLM untuk mengevaluasi output dan beberapa pendekatan yang lebih spesifik pada model.
Sander Schulhoff
Sander Schulhoff is the CEO of HackAPrompt and Learn Prompting. He created the first Prompt Engineering guide on the internet, two months before ChatGPT was released, which has taught 3 million people how to prompt ChatGPT. He also partnered with OpenAI to run the first AI Red Teaming competition, HackAPrompt, which was 2x larger than the White House's subsequent AI Red Teaming competition. Today, HackAPrompt partners with the Frontier AI labs to produce research that makes their models more secure. Sander's background is in Natural Language Processing and deep reinforcement learning. He recently led the team behind The Prompt Report, the most comprehensive study of prompt engineering ever done. This 76-page survey, co-authored with OpenAI, Microsoft, Google, Princeton, Stanford, and other leading institutions, analyzed 1,500+ academic papers and covered 200+ prompting techniques.
🟢 Penyaringan Kata / Frasa (Filtering)
🟢 Pertahanan Instruksi
🟢 Evaluasi LLM secara Terpisah
🟢 Pendekatan Lainnya
🟢 Post-Prompting
🟢 Random Sequence Enclosure
🟢 Sandwich Defense
🟢 XML Tagging
Footnotes
-
Crothers, E., Japkowicz, N., & Viktor, H. (2022). Machine Generated Text: A Comprehensive Survey of Threat Models and Detection Methods. ↩
-
Goodside, R. (2022). GPT-3 Prompt Injection Defenses. https://twitter.com/goodside/status/1578278974526222336?s=20&t=3UMZB7ntYhwAk3QLpKMAbw ↩