Recently, I attended the reading group of our TELOS lab, and I think the previous method of my paper reading has something incorrect. Here comes two questions:
- How can you summarize the paper after reading it, and how can the summary reminding you later?
- How can AI help you when you reading?
The first question is simple: reading with questions. In the reading group, Diyu developed an article in five steps:
- Target problem: The problem this article targets for. Every paper, excluding the new are purposing papers, must correspond with a problem.
- Prior work: Why this paper occurs? The prior work must have something that didn’t work well enough.
- Challenges: To solve this problem, there must have some challenges where prior work can’t deal.
- Design: This paper purposes some design to solve the challenges.
- Evaluation: Something that not that important.
- Thoughts: Is this a good problem? Is this a good paper? Think it fundamentally.
The second question that need some thoughts. If we directly ask AI to generate the answer above, we can’t promise that AI has a good comprehensive of this paper. So I think the best is read the paper line by line by yourself, and then ask AI where confused you. Finally, submit your answer and let AI check. This time you will review too. So after the complicated pipeline, I think you will finally understand it.