MO-CAPO: Multi-Objective Cost-Aware Prompt Optimization
The paper introduces MO-CAPO, a novel algorithm for multi-objective prompt optimization in large language models. It aims to balance performance and inference cost, addressing limitations in existing methods. The evaluation shows that MO-CAPO outperforms traditional approaches while providing diverse performance-cost trade-offs.
- ▪MO-CAPO optimizes both performance and inference cost in prompt design for large language models.
- ▪The algorithm outperforms the NSGA-II baseline in 8 out of 12 cases based on the noisy R2 metric.
- ▪It allows practitioners to select from a set of prompts that offer various trade-offs between performance and cost.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Machine Learning arXiv:2605.18869 (cs) [Submitted on 15 May 2026] Title:MO-CAPO: Multi-Objective Cost-Aware Prompt Optimization Authors:Jan Büssing, Moritz Schlager, Timo Heiß, Tom Zehle, Matthias Feurer View a PDF of the paper titled MO-CAPO: Multi-Objective Cost-Aware Prompt Optimization, by Jan B\"ussing and 4 other authors View PDF Abstract:Large language models (LLMs) achieve strong performance across a wide range of tasks but are highly sensitive to prompt design, motivating the need for automatic prompt optimization. Existing methods predominantly focus on performance alone, ignoring competing objectives such as inference cost or latency.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
