Skip to Main Content

Structured prompts enable GPT-3.5-class models to match GPT-4 output quality on 78% of classification tasks, at 1/30th t.Khattab et al., 'DSPy: Compiling Declarative Langu…