#machine learning
Dec 2025
Poodle: Seamlessly Scaling Down Large Language Models with Just-in-Time Model Replacement
This paper presents the vision for just-in-time model replacement (JITR), where, upon identifying a recurring task in calls to an LLM, the model is replaced transparently with a cheaper alternative that performs well for this specific task.
Nils Strassenburg, Boris Glavic, T. Rabl
· arXiv.org · 1 citation