Paper Review: Learning to Retrieve in Context Examples for Large Language Models

In this video, I discuss the importance of learning to retrieve in context examples for large language models. I explain how the choice of examples can significantly impact the quality of results. I also introduce a recent work by Microsoft's Furu Wei's lab, which focuses on training a reward-based model on LM feedback to evaluate candidates. The video covers the two-step process of training a reward model and using it for knowledge distillation to train a dense retrieval model. Various tasks are showcased to demonstrate the effectiveness of this approach. The work is publicly available for exploration and experimentation.