GPT-4 can often generate usable spaced repetition prompts for declarative knowledge from explanatory text with guidance

By combining the following insights, I’ve had success with GPT-4 in Using machine learning to generate good spaced repetition prompts from explanatory text, usually on the first try:

• In prompt generation, choosing reinforcement targets and writing prompts for those targets are two separate problems. In these examples, I choose the targets (i.e. by pointing to phrases within a larger passage). The model’s job is to generate effective prompts given those specific targets.
• In prompt generation, LLMs may perform better when given prompt-writing principles; I provide those principles.
• In prompt generation, LLMs often need extra hints about what angle to reinforce; I provide those hints.
• In prompt generation, LLMs may perform better with ample context; I provide that context.

This works fairly well for simple statements, but not for more conceptual material: In prompt generation, LLMs lack prompt-writing patterns for complex conceptual material

An example prompt: 20230614123022