And then some alignment, prompting structure, and task decomposition.
And praying that your desired output was embedded into the training data that was used to generate the model.
And praying that your desired output was embedded into the training data that was used to generate the model.