Large Language Models can Learn Rules

arXiv:2310.07064v2 Announce Type: replace-cross
Abstract: When prompted with a few examples and intermediate steps, large language models (LLMs) have demonstrated impressive performance in various reasoning tasks. However, prompting methods that rely on implicit knowledge in an LLM often generate incorrect answers when the implicit knowledge is wrong or inconsistent with the task. To tackle this problem, we present Hypotheses-to-Theories (HtT), a framework that learns a rule library for reasoning with LLMs. HtT contains two stages, an induction stage and a …

abstract arxiv cs.ai cs.cl examples generate however intermediate knowledge language language models large language large language models learn llm llms performance prompting reasoning rules tasks type