Following the findings from Min et al. (2022)(opens in a new tab), here are a few more tips about demonstrations/exemplars when doing few-shot:
Prompting GPT-4 for multi-visual interactive dashboard creation
. The author experiments with the model, asking it to add a walrus to a prompt, and is surprised to find that the model can maintain consistency between images with a slightly altered prompt using a "seed" number. The author also delves into the underlying prompt engineering of DALL-E 3, revealing policies and guidelines that govern the model's image generation, including diversity and inclusivity guidelines.
training GPT to query with few-shot prompting.
