Topics
Sample efficiency is the ability of a statistical method or algorithm to make accurate predictions or inferences using a minimal amount of sample data. It’s also a measure of how well samples are used to train a model.
High sample efficiency implies that a model can quickly assimilate knowledge, needing fewer data points to learn. This idea becomes especially striking when comparing the learning abilities of humans with those of large language models (LLMs) ⎯ Humans are remarkably sample-efficient learners.