Jon Bishop
A trapdoor prompt is an input designed to trigger a specific output from a language model, without using any of the words in that output. It’s not a guess and not a coincidence. It’s a byproduct of how models memorize fragments of their training data and the way those fragments can be resurfaced with the […]
When I began exploring AI prompt engineering, I quickly realized that how I structured my prompts significantly impacted the quality of the AI’s output. A crucial aspect of this structuring was the use of delimiters, specific sequences of characters like —, ### and |||, to organize and clarify the instructions given to AI models like […]