Iteratively Prompt Pre-trained Language Models for Chain of Thought
Boshi Wang, Xiang Deng Huan Sun The Ohio State University, Columbus, OH
Abstract
While Pre-trained Language Models (PLMs) internalize a great amount of world knowledge, they have been shown incapable of recalling these knowledge to solve tasks requiring complex & multi-step reasoning. Similar to how humans develop a “chain of thought” for these tasks, how can we equip PLMs with such abilities? In this work, we explore an iterative prompting framework, a new prompting paradigm which progressively elicits relevant knowledge from PLMs for multi-step inference. We identify key limitations of existing prompting methods, namely they are either restricted to queries with a single identifiable relation/predicate, or being agnostic to input contexts, which makes it difficult to capture variabilities across different inference steps. We propose an iterative context-aware prompter, which addresses these limitations by learning to dynamically synthesize prompts conditioned on the current step’s contexts. Experiments on three datasets involving multi-step reasoning show the effectiveness of the iterative scheme and the context-aware prompter design.111Our source code is available at https://github.com/sunlab-osu/IterPrompt.
中文速览
大型预训练语言模型（PLM）虽然内化了大量世界知识，却在需要多步推理的复杂问题上屡屡碰壁——即便模型"记得"每一条单步事实，也难以将它们串联成完整的推理链。为此，研究者提出了一种"迭代上下文感知提示（Iterative Context-Aware Prompter）"框架：在每个推理步骤，由一个轻量级的提示生成器根据当前查询和已召回的中间知识动态合成提示，引导冻结参数的PLM逐步回忆出下一条所需知识，从而像人类一样构建出完整的"推理链"。在2WikiMultiHopQA、R4C和LoT三个多步推理数据集上的实验表明，迭代方案本身以及上下文感知的提示设计都带来了显著的性能提升，优于静态提示调优、前缀调优等现有方法。这项工作的价值在于，它无需微调庞大的模型参数便能激活PLM的多步推理潜力，为未来随模型规模持续扩大而高效利用其内部知识提供了一条可行路径。
原文 arXiv:2203.08383;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2203.08383v3