keyword

prompt generation methods

Prompt generation methods are systematic techniques in artificial intelligence and natural language processing used to construct, discover, or optimize input templates and instructions that guide language models to complete tasks or elicit stored information. Rather than relying solely on manual trial and error, these approaches often employ automated or semi-automated strategies, such as mining text corpora for relational patterns, paraphrasing existing templates, discrete token search, or continuous vector optimization. By identifying high-performing, diverse, and contextually appropriate formulations, prompt generation methods reduce the need for extensive human engineering, mitigate model sensitivity to specific phrasing, and enable more accurate evaluations of the knowledge and capabilities embedded within pretrained models.

1 item

How Can We Know What Language Models Know?

Zhengbao Jiang, Frank F. Xu, Jun Araki, Graham Neubig

OrganizationsCarnegie Mellon UniversityRobert Bosch Research and Technology Center

Why you should read this

Demonstrates that manual prompts underestimate the factual knowledge stored in language models and introduces automated mining, paraphrasing, and ensembling methods to substantially improve relation extraction accuracy on the LAMA benchmark.

Recent work has presented intriguing results examining the knowledge contained in language models (LM) by having the LM fill in the blanks of prompts such as "Obama is a _ by profession". These prompts are usually manually created, and quite possibly sub-optimal; another prompt such as "Obama worked as a _" may result in more accurately predicting the correct profession. Because of this, given an inappropriate prompt, we might fail to retrieve facts that the LM does know, and thus any given prompt only provides a lower bound estimate of the knowledge contained in an LM. In this paper, we attempt to more accurately estimate the knowledge contained in LMs by automatically discovering better prompts to use in this querying process. Specifically, we propose mining-based and paraphrasing-based methods to automatically generate high-quality and diverse prompts, as well as ensemble methods to combine answers from different prompts. Extensive experiments on the LAMA benchmark for extracting relational knowledge from LMs demonstrate that our methods can improve accuracy from 31.1% to 39.6%, providing a tighter lower bound on what LMs know. We have released the code and the resulting LM Prompt And Query Archive (LPAQA) at this https URL.

Source

https://arxiv.org/pdf/1911.12543

Added

2026-09-24