Google Research describes ToolGrad, which starts from successful tool-use chains and then generates matching prompts. This reverses the common prompt-first process and uses execution feedback to improve the quality of training examples.
From the linked article:
ToolGrad generates successful tool-use chains before generating prompts, yielding a high pass rate.
Source: @GoogleResearch. This note summarizes the linked material; images belong to their respective creators.