Reading and Reviewing Papers › Module 1: Finding papers
Where papers live
Lesson 1 of 3 · 3 min
Research is published in a few main places. Knowing them helps you search well and judge what you find.
Journals
A journal publishes papers after peer review: other experts read the paper and advise the editor whether to accept it. In most sciences, journals are the main place for finished work.
Conferences
In computer science, and especially in AI, top conferences are often more important than journals. Papers are peer reviewed, and accepted papers appear in the conference proceedings. Well-known AI venues include NeurIPS, ICML and ICLR (machine learning), CVPR (computer vision) and ACL (language).
Preprints
A preprint is a paper shared publicly before, or without, peer review. arXiv is the main preprint server for physics, mathematics, computer science and related fields. It has basic moderation, but moderation is not peer review. Many important AI papers appear on arXiv months before a conference.
Kinds of papers
Papers also differ in what they do. The kind tells you what to expect, and how much to trust it.
| Kind | What it is | Typical length |
|---|---|---|
| Journal article | Finished, peer-reviewed work; often the fullest version | 8–30 pages |
| Conference paper | Peer-reviewed work presented at a conference; the main kind in CS and AI | 8–10 pages + appendix |
| Short paper / letter | A small, quick result | 2–5 pages |
| Workshop paper | Early or in-progress work at a smaller event beside a conference; lighter review | 4–8 pages |
| Survey / review | Sums up a whole area; the best place to start a new topic | 15–50 pages |
| Systematic review / meta-analysis | A review with a strict search method; a meta-analysis also pools the numbers | varies |
| Position paper | Argues a view or a direction for the field; little or no new data | 4–10 pages |
| Dataset or benchmark paper | Introduces new data or a new test | 8–10 pages |
| Thesis | A PhD or master's dissertation; long, detailed, with lots of background | 50–300 pages |
| Technical report | Released by a lab or company, often without peer review | varies |
- Journal article
- What it is: Finished, peer-reviewed work; often the fullest version
- Typical length: 8–30 pages
- Conference paper
- What it is: Peer-reviewed work presented at a conference; the main kind in CS and AI
- Typical length: 8–10 pages + appendix
- Short paper / letter
- What it is: A small, quick result
- Typical length: 2–5 pages
- Workshop paper
- What it is: Early or in-progress work at a smaller event beside a conference; lighter review
- Typical length: 4–8 pages
- Survey / review
- What it is: Sums up a whole area; the best place to start a new topic
- Typical length: 15–50 pages
- Systematic review / meta-analysis
- What it is: A review with a strict search method; a meta-analysis also pools the numbers
- Typical length: varies
- Position paper
- What it is: Argues a view or a direction for the field; little or no new data
- Typical length: 4–10 pages
- Dataset or benchmark paper
- What it is: Introduces new data or a new test
- Typical length: 8–10 pages
- Thesis
- What it is: A PhD or master's dissertation; long, detailed, with lots of background
- Typical length: 50–300 pages
- Technical report
- What it is: Released by a lab or company, often without peer review
- Typical length: varies
RememberNew to a topic? Start with a recent survey, then read the papers it cites most.
Where to search
| Tool | Good for |
|---|---|
| Google Scholar | Searching almost everything; "Cited by" links |
| Semantic Scholar | AI-assisted search, short summaries, citation graphs |
| arXiv | The newest preprints in CS, maths and physics |
| dblp | Complete lists of computer science papers by author and venue |
| PubMed | Medicine and life sciences |
| OpenAlex | A free, open index of research papers, authors and institutions |
- Google Scholar
- Good for: Searching almost everything; "Cited by" links
- Semantic Scholar
- Good for: AI-assisted search, short summaries, citation graphs
- arXiv
- Good for: The newest preprints in CS, maths and physics
- dblp
- Good for: Complete lists of computer science papers by author and venue
- PubMed
- Good for: Medicine and life sciences
- OpenAlex
- Good for: A free, open index of research papers, authors and institutions
Getting access
Many papers are free to read. If one is behind a paywall, try:
- the arXiv version, or the author's own website
- an open-access copy found through your library or tools like Unpaywall
- your college library
- emailing the author: most researchers are glad to send a copy
RememberUse pirate sites at your own risk. They are illegal in many countries, including India, and there is almost always a legal route.