We will introduce Prompt Engineering and Finetune LLM with LoRA, along with the following paper.
LOLA: LLM-Assisted Online Learning Algorithm for Content Experiments | Marketing Science
Code for paper is available at https://github.com/DDDOH/LLM_News
One type of content experiment: given a article, choose the headline that leads to more click (click through rate, CTR)
Current methods: A/B test
Dataset available:
Can LLM replace A/B test? Let LLM act as the editor who select the headline.
<aside> 💡
Give ChatGPT multiple headline candidates of one article, let ChatGPT select the one that can lead to highest CTR.
</aside>
Prompt engineering is the practice of designing and refining inputs.
Rather than modifying the model’s parameters, we craft better instructions.