How to Read a Research Paper

About: Reading research papers is challenging; they are often dense, assuming significant prior knowledge about the field and related work. It is important to:

1. Expect that it will be a learning process – It might take a while at first, but with practice, you will improve!
2. Adopt and practice paper-reading strategies.

This guide outlines a general strategy for paper reading. We hope that you will find it useful, and encourage you to ask peers, colleagues, and mentors for more insights about how they read papers, especially in your field.

Acknowledgements: This guide is adapted from Harvard’s Fall 2024/Spring 2025 course “CS2901: Seminar on Effective Research Practices & Academic Culture” (Archived). Other inspirations include Margo Seltzer’s course on OS Research and Weiwei Pan’s course on Stochastic Methods.

Paper-Reading Strategies

Research papers are not meant to be read front-to-back like a novel.

Instead, consider these steps:

1. Skim the Paper

• Focus on the high-level ideas.
• Jot down high-level questions you may have about the paper’s narrative.
• Avoid getting bogged down by details.

2. Deeper Read (if relevant)

• Don’t read front-to-back – instead, read to answer the “guiding questions” (see below).
• Answering these questions helps you construct the story or narrative of the paper.

3. Read Actively

• Compare what the paper says with what you already know.
• Draw pictures or diagrams to check your understanding.
• Translate general confusion into specific questions to improve clarity.

4. Tips for Getting Unstuck

• You might be missing assumed knowledge.
• Look for related papers, Wikipedia pages, blog posts, or lectures to fill gaps.
• Going back and forth between related materials and the original paper can help you get unstuck.

5. Stay Organized

• Take notes to save time when revisiting the paper.
• Methods include:
  • Writing notes on the paper itself.
  • Keeping a searchable document (e.g., Evernote) to summarize the paper.
  • Using a citation manager (e.g., Zotero) to save paper lists and easily generate bibliographies.
• For every paper you read, you should be recording:
  • The Big Idea of the paper.
  • Any relevant buzzwords, jargon, or technical terms of interest.
  • Broad strokes of the story/narrative, answers to our guiding questions.

6. Ask Many Questions

• Ask questions about the paper.
• Ask questions about the paper’s context.
• Ask questions about the paper’s implications.
• Nobody knows how to read a paper automatically; it’s a skill that takes time to develop.
• Research papers are very dense and assume prior knowledge. It is crucial to ask questions and seek answers to them as this will help you understand the paper better (and also help you understand the field better).
• When entering a new field or subject area, it is normal for it to take several hours to read a single paper. This is because you are learning the language of the field, the context, and the background knowledge that the paper assumes you have. Do not get discouraged by this; it is a normal part of the learning process.

Guiding Questions

We recommend of thinking of “paper reading” as the process of answering the following guiding questions. Whether you are skimming or deep-reading, these guiding questions will help you reconstruct the narrative told by the paper. By practicing answering these questions for every paper you read, you will practice quickly honing in on the meaning behind the paper instead of being bogged down by details.

In other words, these guiding questions give you a framework for understanding any research paper you’re reading. The more you practice answering these, the faster you’ll be at skim-reading and understanding the main ideas of a paper.

What kind of paper are you reading?

By understanding the type of the paper, you can anticipate its structure and evaluation, which will help you determine which parts to on during your initial skim.

Some paper types include:

• Big idea
• Unifying theme
• Small idea with evaluation (most papers)
• Measurement
• Comparison
• Retrospective or experience paper
• Proposes a new model or type of problem for study
• Solves a known open problems
• Clarifies or simplifies our understanding of a known result or phenomenon
• New techniques for proofs

What is the motivation and potential impact of the paper?

• What is the problem the paper is trying to solve?
• Why is the problem important?
• For theory papers, what are the formal definitions? Do they accurately capture the problem?

What is the relevant related work and what is this paper’s contribution?

• How is the problem traditionally solved?
• Why are existing approaches insufficient?
• How does the paper address the problem?

What are the results (Theory/Experiments)?

• Theory:
  • What are the main theorems?
  • What are the proof techniques?
  • How do the results relate to earlier results?
• Experiments:
  • How does the paper define success?
  • How do they measure/quantify success?
  • What approaches do they compare against?
  • What are the results? (high-level)

What are the implications of the paper? (Broader Impact)

• Identify socio-technical systems that might be affected by the technology.
  • Where could this technology be deployed?
  • What kind of technology is informed by this research?
  • How would this technology be used?
• Identify the stakeholders
  • Who would be the users?
  • Who are the affected communities? Are these the same as the users?
• What positive benefits can this technology realize?
  • What kinds of needs do these users/businesses/communities have?
  • What kinds of constraints do these users/businesses/communities have?
• What negative consequences can this technology realize?
  • What happens if this kind of technology fails? What are its failure modes?
  • What direct harms could this technology cause? Which are born of the failure modes?
  • What kinds of harm can be caused by the broader, socio-technical system?

How to Critique a Paper

After answering the guiding questions, critique the paper by considering these:

The Ideas

• Do the ideas make sense?
• Are the ideas well-motivated?

The Methodology

• Are the high-level ideas behind the approach sound? (i.e., does the approach appear plausible?)
• For theory papers: Do the theorems support the claims?
  • Are the theorems proved by the authors the “right ones” to prove to support the claims? If not, do the authors discuss why they chose the theorems they did?
  • Do the proofs appear to be complete? Are they clear in their exposition?
  • Is the theory adequately motivated? Are the definitions and theorems situated well within the context of existing work?
• For experimental papers: Do the results support the claims?
  • Are the problem instances studied representative of the problem space?
  • Do the results generalize? Or is this case-study specific?
  • Do the baselines reflect the current state-of-the-art for the problem?
  • Are the evaluation metrics accurate and appropriate?
  • Do the experiments provide evidence that the innovations introduced are responsible for the improvements observed?

Presentation

• Is the paper well-written?

More information:

• Teaching Materials
• SIGCSE Paper