Li gëna néew-ba-li gëna am solo

Least-to-Most prompting, introduced by Zhou and colleagues at Google in 2022, has two stages. First, the model is prompted to decompose a complex question into an ordered list of easier subquestions. Second, it solves those subquestions one at a time, appending each solved answer to the context so later steps can build on earlier ones. This differs from chain-of-thought, which reasons in a single pass without explicit decomposition. The authors evaluated symbolic manipulation, math reasoning, and SCAN, a compositional-generalization benchmark built from natural-language commands. For the specific SCAN setup using GPT-3 code-davinci-002 and 14 exemplars, Least-to-Most reached at least 99% accuracy across the tested splits, including the length split, compared with 16% for chain-of-thought prompting. This is a result for that model, prompt setup, and benchmark, not a general accuracy guarantee.