Incremental Utility: A Novel Approach to Improve Few-Shot In-Context Learning with Large Language Models
Introducing a novel method called "incremental utility" to estimate how much additional knowledge a demonstration brings to a large language model for few-shot in-context learning tasks, and showing its effectiveness compared to previous utility estimation approaches.