Automatic Prompt Engineering with No Task Cues and No Tuning
Faisal Chowdhury, Nandana Mihindukulasooriya, Niharika S D'Souza, Horst Samulowitz, Neeru Gupta, Tomasz Hanusiak, Michal Kapitonow
TL;DR
This work addresses the brittleness and labor costs of traditional prompt engineering by introducing a simple, language‑adaptive automatic prompt engineering framework that requires no tuning, seed prompts, or task cues. It generates prompts from few example I/O pairs using a task‑agnostic meta‑prompt and randomized sampling, then ranks candidates with a lightweight, LLM‑free similarity score. Evaluated on cryptic column name expansion (CNE) tasks in English and German, the approach matches or surpasses more complex baselines while being easier to adopt and language‑agnostic. The study also provides a German CNE dataset, demonstrating practical impact for multilingual data discovery and readability in tabular contexts.
Abstract
This paper presents a system for automatic prompt engineering that is much simpler in both design and application and yet as effective as the existing approaches. It requires no tuning and no explicit clues about the task. We evaluated our approach on cryptic column name expansion (CNE) in database tables, a task which is critical for tabular data search, access, and understanding and yet there has been very little existing work. We evaluated on datasets in two languages, English and German. This is the first work to report on the application of automatic prompt engineering for the CNE task. To the best of our knowledge, this is also the first work on the application of automatic prompt engineering for a language other than English.