Skip to Content
BlogResearch & WritingReading Papers

Reading Research Papers Effectively

Effective paper reading is a critical research skill. The three-pass method allows you to efficiently evaluate and understand papers at different depths.

The Three-Pass Method

Pass 1: The Quick Scan (5-10 minutes)

Purpose: Decide if the paper is relevant to your work.

What to read:

  1. Title, abstract, introduction
  2. Section and subsection headings
  3. Conclusion
  4. References (scan for familiar papers)

Questions to answer:

  • What category? (Methodology, application, survey, etc.)
  • What is the core contribution?
  • Is it relevant to my work?

Pass 2: Detailed Reading (30-60 minutes)

Purpose: Grasp the content, but skip detailed proofs and derivations.

How to read:

  1. Examine figures and tables carefully
  2. Mark relevant references for later reading
  3. Identify assumptions and limitations
  4. Note questions for later investigation

Key question: Can you summarize the paper to someone else?

Pass 3: Deep Understanding (2-4 hours)

Purpose: Virtual re-implementation - understand the work deeply enough to reproduce it.

What to do:

  • Challenge every assumption
  • Work through the mathematical derivations
  • Think about how you would implement it
  • Compare their approach to yours
  • Identify what you would do differently

Key question: Can you reproduce this work?

Paper Reading Tracker

Use this template to organize your paper notes:

class PaperNotes:
    def __init__(self, paper_title):
        self.title = paper_title
        self.pass1 = {
            'category': '',
            'core_contribution': '',
            'relevant': True  # or False
        }
        self.pass2 = {
            'key_figures': [],
            'assumptions': [],
            'limitations': [],
            'relevant_refs': []
        }
        self.pass3 = {
            'implementation_notes': '',
            'comparison_to_my_work': '',
            'questions_for_authors': []
        }

Application to Your Research

For thesis work, you should aim for:

  • Pass 3 for papers directly related to your thesis (e.g., ETHOS, CLIP, key baseline papers)
  • Pass 2 for all papers you cite in your related work
  • Pass 1 for exploratory reading to understand the broader field

Best Practices

  1. Start broad, go deep: Begin with Pass 1 on many papers, then focus on the most relevant
  2. Take notes: Document your understanding at each pass
  3. Build a bibliography: Use reference managers like Zotero or Mendeley
  4. Compare papers: Create comparison tables for related work
  5. Read actively: Question assumptions, identify gaps, think critically

Critical Reading Questions

When reading papers, ask yourself:

  • Validity: Are the claims supported by evidence?
  • Assumptions: What assumptions are made? Are they reasonable?
  • Limitations: What are the weaknesses? What wasn’t tested?
  • Reproducibility: Could I reproduce these results?
  • Relevance: How does this relate to my work?
  • Impact: What are the broader implications?

Creating a Literature Map

As you read papers, build a conceptual map of the field:

  1. Core papers: Foundational work that everyone cites
  2. Recent advances: Latest improvements and techniques
  3. Related work: Papers adjacent to your topic
  4. Baselines: Methods you’ll compare against
  5. Gaps: What’s missing that you can address

This map will form the basis of your Related Work section.

Organizing Your Reading

By Research Phase

Exploration Phase (Months 1-2):

  • Pass 1 on 50-100 papers
  • Pass 2 on 20-30 most relevant
  • Pass 3 on 5-10 key papers

Implementation Phase (Months 3-4):

  • Pass 3 on baseline papers you’re comparing against
  • Pass 2 on papers with relevant techniques
  • Pass 1 for staying current with new work

Writing Phase (Months 5-6):

  • Re-read key papers (Pass 2-3) to ensure accurate citations
  • Scan new papers (Pass 1) to update related work

Paper Categories for Healthcare AI Research

For a healthcare AI thesis, organize papers into categories:

  1. Healthcare Foundation Models: ETHOS, BEHRT, Med-BERT
  2. Multimodal Learning: Multimodal foundations, CLIP
  3. Transformers: Attention is All You Need, GPT, BERT
  4. Healthcare Applications: EHR analysis, medical imaging, clinical NLP
  5. Interpretability: Attention visualization, SHAP, clinical validation

Time Management

Don’t get stuck in “reading forever” mode:

  • Set limits: 2 weeks maximum for literature review
  • Focus on recent: Last 3-5 years for most papers
  • Follow citations: Read papers cited by key work
  • Stop when saturated: Diminishing returns after ~30-40 papers

Key Takeaways

  1. Three passes: Quick scan → Detailed reading → Deep understanding
  2. Be strategic: Not every paper needs Pass 3
  3. Take notes: Document your understanding at each level
  4. Build connections: Create a literature map showing relationships
  5. Read critically: Question assumptions and identify gaps
  6. Manage time: Don’t read forever - focus on most relevant papers
OverviewResearch Questions