zhou-2023-ape-text-davinci-003-instruction-gen
IN premise — summaries/2026/08/24/zhou-2023-context-faithful-prompting-sR-references.md
Created 2026-08-25T02:59:10+00:00
Because smaller LLMs failed to produce meaningful instructions, text-davinci-003 was used to generate instructions via APE across all model scales tested, with top-3 instructions generated per dataset and the top-1 selected for experiments.
Summary
Because the smaller models under test couldn't write their own effective instructions, a single fixed generator was used to produce them for every model, keeping instruction quality constant so that performance differences reflect each model's ability to follow prompts rather than to create them.