Prompt Optimization KDD2025:

Welcome to KDD 2025 workshop on Prompt Optimization!

The rapid development of large language models (LLMs) has enabled us to achieve very strong performance across a wide range of tasks, including text generation, summarization, and question answering. However, the effectiveness of these LLMs heavily depends on the quality of the prompts used. Furthermore, LLMs are known to exhibit unpredictable sensitivity to input factors like the task description, ordering, choice of delimiters, etc. Prompt optimization has become a critical step to elicit desired responses for various tasks. Despite its importance, AI practitioners often rely on trial-and-error methods, leading to inefficiencies, and sub-optimal performance. These challenges are further compounded by the growing demand for prompt optimization in low-resource settings, multimodal applications, agentic and multi-agentic systems, and ethical AI deployment.

This workshop aims to address these gaps by providing a forum for researchers and practitioners to share their latest findings, tools, and methodologies in prompt optimization and engineering. The event will focus on prompt optimization including hard/discrete prompts (human-readable) and soft prompts (embeddings). Topics of interest include best practices, theoretical understanding, adversarial prompting, robustness and generalization, interplay between exemplar and instruction optimization, and applications in domains such as healthcare and finance.

This workshop will highlight the latest advancements in prompt engineering and optimization, provide a platform for discussing challenges and opportunities, and also encourage collaboration between researchers and industry professionals. We invite submissions from a diverse range of perspectives, including theoretical insights, empirical studies, and practical applications. By bringing together experts from NLP, ML, and related fields, this workshop will play an important role in shaping the future of prompt engineering and optimization.

Call for Contributions

This workshop will cover a wide range of research topics related to prompt engineering and optimization, focusing on both prompt techniques in real-world applications and theoretical understanding of these methods. Below is a detailed list of important research topics that the workshop will cover:

• Prompt engineering methods
• Prompt optimization for different applications such as RAG, Agents, and Multi-agent systems (MAS)
• Automatic discrete prompt optimization
• Prompt tuning
• Prompt optimization for multilingual and low-resource settings
• Prompt optimization methods to mitigate bias, fairness, and ethical concerns
• Multi-modal prompting
• Adversarial prompting
• Theoretical foundations of prompt optimization
• Framework and tools used for prompt optimization