In-context Learning (ICL) is one of the key methods for enhancing the performance of large language models on specific tasks by providing a set of few-shot examples. However, the ICL capability of different types of models shows significant variation due to factors such as model architecture, volume of learning data, and the size of parameters. Generally, the larger the model's parameter size and the more extensive the learning data, the stronger its ICL capability. In this paper, we propose a method SLEICL that involves learning from examples using strong language models and then summarizing and transferring these learned skills to weak language models for inference and application. This ensures the stability and effectiveness of ICL. Compared to directly enabling weak language models to learn from prompt examples, SLEICL reduces the difficulty of ICL for these models. Our experiments, conducted on up to eight datasets with five language models, demonstrate that weak language models achieve consistent improvement over their own zero-shot or few-shot capabilities using the SLEICL method. Some weak language models even surpass the performance of GPT4-1106-preview (zero-shot) with the aid of SLEICL.

Two years ago a novelist and as-yet-unproduced screenwriter named Nic Kelman went to work for Wizards of the Coast, the company that makes the popular collectible card game Magic: The Gathering. Kelman's job, though he might not put it this way, was to write a grimoire--a kabbalistic story bible. "Rules for magic out of the rules for Magic," as Kelman says. The company needed that grimoire because it was going to try to cast a spell in the real world--to transform a popular albeit niche game, complicated and nerdy, into a cross-media franchise. That has happened for comic books, for literature, even for toys, heaven help us.