Building an AI Agent to help users study research papers published in Arxiv
Our agent generates a random topic and searches for articles in Arxiv based on that topic. It then downloads the article into a PDF file and gives the user a summary of the article, as well as a detailed explanation of the article and some MC Questions to help the user understand the article's topic.
A problem we faced while doing our project was making sure that the summary agent was accurately summarizing the article and not just taking a summary of the summary or abstract. We were able to solve this problem by rephrasing the goal of the summary agent and breaking down the goal into smaller tasks that the blogger agent can manage.
Another problem we faced was downloading the research articles. Because the articles were not getting downloaded correctly, the summarize agent only summarized the short abstract instead of the entire article.