Show Your Work: Scratchpads for Intermediate Computation with Language Models
Large pre-trained language models perform remarkably well on tasks that can be done “in one pass”, such as generating realistic text (Brown et al., 2020) or synthesizing computer programs (Chen et al., 2021; Austin et al., 2021). However, they struggle with tasks that require unbounded multi-step computation, such as adding integers (Brown et al., 2020) or executing programs (Austin et al., 2021).
Surprisingly, we find that these same models are able to perform complex multistep computations—even in the few-shot regime—when asked to perform the operation “step by step”, showing the results of intermediate computations. In particular, we train Transformers to perform multi-step computations by asking them to emit intermediate computation steps into a “scratchpad”. On a series of increasingly complex tasks ranging from long addition to the execution of arbitrary programs, we show that scratchpads dramatically improve the ability of language models to perform multi-step computations.
事前に訓練された大規模な言語モデルは、現実的なテキストの生成(Brown et al., 2020)やコンピュータプログラムの合成(Chen et al., 2021; Austin et al., 2021)のような「1パスで」行えるタスクでは顕著に優れた性能を発揮します。しかし、整数の加算(Brown et al., 2020)やプログラムの実行(Austin et al., 2021)のような、束縛されない多段階の計算を必要とするタスクでは苦戦する。
>shodaiiiiii 段階的に考えてみましょう(Let’s think step by step)より、正しい答えが得られるように、これを段階的に解決していきましょう(Let’s work this out in a step by step way to be sure we have the right answer)の方がPromptの命令文としてパフォーマンスが高いらしい