PISA-E is a prompt-adaptive router that selects an ensemble of models to balance response quality and inference cost. It uses soft cluster memberships and model-specific predictors to adjust ensemble composition at inference time without retraining.
HOW THIS AFFECTS YOU
●
builderYou can use this to switch between model ensembles based on a specific cost-quality preference for each request.
●
founderYou can potentially lower your inference bills by routing easy queries to cheaper model combinations automatically.