Language Models Benefit from Preparation with Elicited Knowledge
This work addresses a specific bottleneck in language model performance for knowledge-intensive QA tasks, offering an incremental improvement over existing zero-shot CoT approaches.
The authors tackled the problem of question answering tasks that rely more on accessing relevant knowledge than on reasoning steps by introducing PREP, a prompting technique using two language model instances, which achieved consistently higher average accuracy across multiple datasets compared to other methods.
The zero-shot chain of thought (CoT) approach is often used in question answering (QA) by language models (LMs) for tasks that require multiple reasoning steps. However, some QA tasks hinge more on accessing relevant knowledge than on chaining reasoning steps. We introduce a simple prompting technique, called PREP, that involves using two instances of LMs: the first (LM1) generates relevant information, and the second (LM2) receives the information from the user and answers the question. This design is intended to make better use of the LM's instruction-following capability. PREP is applicable across various QA tasks without domain-specific prompt engineering. PREP is developed on a dataset of 100 QA questions, derived from an extensive schematic dataset specifying artifact parts and material composition. These questions ask which of two artifacts is less likely to share materials with another artifact. Such questions probe the LM's knowledge of shared materials in the part structure of different artifacts. We test our method on our parts-and-materials dataset and three published commonsense reasoning datasets. The average accuracy of our method is consistently higher than that of all the other tested methods across all the tested datasets.