Reading research papers is an indispensable part of doing research. However, learning how to read papers effectively is far from easy.

Let me first list what I believe are the most important purposes of reading papers:

• First, to build a solid research foundation and develop relevant background knowledge.
• Second, to understand the research trends and challenges in the field.
• The third one is mentioned less often, but it is just as important: learning how to judge the quality and value of a paper.

Although reading more papers usually helps more with research, I would like to strongly emphasize one very important point: not every paper is worth reading in detail. This is so important so I will repeat it a few more times.

Not every paper is worth reading in detail.
Not every paper is worth reading in detail.
Not every paper is worth reading in detail.

I often see students treat papers like textbooks, reading every related paper word by word, trying to understand every detail, and even adding extensive annotations. There is nothing wrong with this level of seriousness, but in research, time is the most limited resource. More importantly, you must learn how to allocate your time effectively and focus your energy on papers that truly matter and are most helpful to your research.

In other words, you need to learn how to quickly judge the quality of a paper and decide which ones are actually useful.

This is especially critical in today’s AI research landscape. Hundreds of new AI papers appear every day, with wildly varying quality. In fact, only a small fraction of them are worth investing serious time and effort in.

Similarly, you should not blindly assume that a published paper is necessarily correct. In this era of rapid AI paper production, many studies lack sufficient rigor, sometimes the theory is flawed, sometimes the assumptions do not hold, and sometimes even the experimental design or results are problematic. Therefore, maintaining a critical mindset while reading papers is absolutely essential.

Below, I would like to share my own process for reading research papers.

1️⃣ Check the authors and their affiliations

Before I start reading a paper, I usually look at where the authors come from. This is not an absolute criterion for judging quality, but in my experience, well-known research groups tend to produce more consistent work and are more likely to offer new insights. By checking the authors and institutions, I can form an initial sense of the paper’s reliability.

2️⃣ Skim through the abstract and introduction

I begin by quickly scanning the abstract and introduction. On the one hand, this helps me confirm whether the topic aligns with my interests; on the other hand, I check whether the authors clearly articulate the paper’s main contributions. Based on how substantial and appealing these contributions are, I roughly decide how much time I am willing to spend reading later.

3️⃣ Predict where the paper is going

This is, in my opinion, the most important stage of reading a paper. Before diving into the methods section, I try to infer the possible research design based on the introduction. At the same time, I ask myself: If I were to be convinced by the authors’ claims, what kind of evidence would I need to see? What would convince me that the proposed approach is reasonable and effective?

4️⃣ Quickly read the methods

Next, I read the methods section to see whether my predictions were correct. If the approach largely matches what I expected, I won’t spend much time on the remaining details. If it differs significantly from my expectations, I then read it more carefully and compare the authors’ design choices with the alternatives I had in mind.

5️⃣ Check whether the experimental results support the claims

Finally, I examine whether the experimental results actually support the claimed contributions. I check whether each major claim is backed by corresponding and sufficient evidence.

6️⃣ Connect the paper’s core ideas to my own research

After getting a high-level concept of the paper, I think about how helpful it is for my current research. If certain ideas seem promising or applicable, I go back and reread the relevant sections in detail and think more deeply about them.

This is how I read research papers. The most important thing is not to “read a paper from beginning to end,” but to learn how to skim efficiently, grasp the core ideas, and continuously think, predict, and evaluate a paper’s value while reading. If a paper turns out not to be very helpful, once you have understood its main idea, you should move on and save your time for works that truly delivers insights and helps you learn something new.