Random Numbers Improve Output Diversity in Language Models

ACL ARR 2025 February Submission4772 Authors

16 Feb 2025 (modified: 09 May 2025)ACL ARR 2025 February SubmissionEveryoneRevisionsBibTeXCC BY 4.0
Abstract:

A frequent roadblock in AI research and its real-world applications is that there are only so many potential answers one can get from a single prompt. In this paper we present RESt – a prompting technique yielding diverse outputs from a single prompt without any human intervention. We explore AI's proven divergent thinking capabilities and supplement them with the addition of random numbers, which spark association between different concepts. We show that, just like humans, machines can be creative by drawing inspiration from external stimuli.

Paper Type: Long
Research Area: Generation
Research Area Keywords: human evaluation, automatic evaluation, few-shot generation
Contribution Types: NLP engineering experiment, Data resources, Data analysis
Languages Studied: English
Submission Number: 4772
Loading