KDD 2025 Workshop on Prompt Optimization

摘要

Prompt engineering plays a critical role in enabling the effective use of large language models (LLMs). LLMs exhibit unpredictable sensitivity to various input factors that may result in a performance gap between prompts that are semantically indistinguishable. Therefore, LLM researchers and practitioners often optimize their prompts in an ad hoc manner due to the lack of systematic methods. Prompt optimization remains an open problem due to the rapidly evolving landscape of NLP tasks, target LLMs, and associated best practices. To address this gap, we propose the first KDD workshop on Prompt Optimization. This workshop aims to bring together researchers and practitioners working on prompt design, automatic optimization, and evaluation, fostering the exchange of ideas and methodologies. By exploring topics such as discrete and soft prompt tuning, low-resource applications, we aim to establish best practices, identify challenges, and drive future research in this critical area.