Conducting Research

Every piece of research you do stands on the shoulders of giants — the accumulated work of countless researchers who came before you. Your job is not to reinvent the wheel, but to understand what has already been built well enough to know where the next step lies. This guide will help you navigate the academic publishing ecosystem, find and evaluate relevant work, and situate your thesis within the broader literature.

Publication Venues

In computer science and AI, research is primarily disseminated through conferences and journals. Understanding the landscape of venues helps you judge the significance of a paper and, eventually, where to aim your own work.

Venue type	Typical timeline	What you'll find
Workshops	Co-located with conferences	Early-stage ideas, position papers, ongoing work. Lower acceptance bar, great for feedback.
Conferences	6–12 months from submission to proceedings	The primary venue for mature contributions in AI/ML/Robotics. Top venues: NeurIPS, ICML, ICLR, CVPR, ECCV, ICCV, ICRA, IROS, etc.
Journals	1–2+ years from submission	Comprehensive, extensively reviewed work. Top journals: TPAMI, JMLR, IJCV, IJRR, Science Robotics, etc.

For your thesis literature review, focus primarily on top-tier conference papers (the last 3–5 years) and journal papers for foundational methods.

Understanding Conference and Journal Rankings

Not all venues carry the same weight. Use dedicated ranking resources to assess the prestige of a conference or journal before deciding how much to trust its papers:

Conferences — use the CORE Conference Portal to check the tier (A*, A, B, C) of a venue. A* and A conferences (e.g., NeurIPS, CVPR, ICRA) represent the gold standard; B venues are still reputable but less selective.
Journals — use SCImago Journal Rank (SJR) to look up a journal's quartile (Q1–Q4) within its subject area. Q1 journals are the most prestigious; Q4 journals warrant extra scrutiny.

Venue prestige is a useful heuristic, not a verdict. The content of a paper is always the primary thing to evaluate: a preprint or a workshop paper can contain a genuinely important idea; conversely, even top-tier venues occasionally publish flawed work. Read critically regardless of where a paper is published, evaluating its quality according to the criteria outlined in Evaluating Paper Quality.

Types of Papers

Learning to distinguish what kind of paper you need at any given moment will save you a lot of reading time. Here are the main types of research papers you'll encounter:

Type	Description	How to identify it
Full paper	The standard research contribution. Presents a problem, method, experiments, and analysis.	8–16 pages; peer-reviewed at a main conference or journal.
Short paper	A focused contribution — same rigour, narrower scope.	4–6 pages; often marked explicitly.
Survey / Review	Systematically covers an entire research area. Excellent entry point for new topics.	Titles like "A Survey of…", "Recent Advances in…".
Preprint	Not yet peer-reviewed. Fast to share, but treat with appropriate scepticism.	arXiv ID (e.g., arXiv:2312.12345). Check if a peer-reviewed version exists.

Finding Papers

The first step in any research project is to understand the landscape of existing work. This means finding and reading papers that are relevant to your topic. Here are the best search tools:

Tool	Best for
Google Scholar	General search; citation counts; "Cited by" to find follow-up work.
Scopus	A comprehensive research database with a query-based search engine for journal articles. Keyword queries retrieve articles containing those words. Requires a PoliMI account.

Starting from scratch

When you have a topic but no specific paper to begin from, use this sequence to build a solid understanding efficiently:

Search for a recent survey paper on your topic. This gives you vocabulary, keywords, and an overview of open problems.
Identify the 5–10 most-cited papers in the area from the survey's references. These are the foundational methods.
For each foundational paper, follow its "Cited by" chain on Google Scholar to find the latest work that builds on it. Also browse the conference proceedings of relevant top-tier venues to catch important papers that might not surface through citation chains.
Stay updated on LinkedIn and X (Twitter) by following prominent researchers, lab accounts, and conference pages. Announcements, code releases, and early discussion of new results surface here before anywhere else.

Starting from an existing paper

When your supervisor gives you a specific paper as a starting point, use that paper as a hub and expand outward:

Read the paper carefully and identify the key problem it solves, the method it proposes, and its main claims.
Mine its reference list for foundational and directly related work — these are the papers you need to understand first.
Use the "Cited by" chain on Google Scholar to find papers that build on or challenge it. This reveals the trajectory of the field after this work was published.
Look for a survey paper that covers the same area — it will contextualise the paper within the broader literature and surface related threads you might otherwise miss.
Stay updated on LinkedIn and X (Twitter) by following the paper's authors and relevant lab accounts to track their latest work and ongoing discussions.

Evaluating Paper Quality

Not all published work is equally rigorous. Before spending significant time on a paper, do a quick quality check:

Signal	What to check
Ablation studies	Does the paper isolate contributions? A method without proper ablations may result in unreliable claims.
Baselines	Are the comparisons fair and against strong, recent baselines?
Evaluation	Are the findings exhaustively evaluated across multiple datasets and metrics? Are the results and limitations clearly discussed?
Code availability	Open-source code strongly suggests the results are reproducible.

Organising Papers

A well-organised library pays dividends at every stage of your thesis — from literature review to writing the final draft. The standard tool we recommend is Zotero (free, open-source, with browser integration).

Recommended plugins

Better BibTeX — improves BibTeX/BibLaTeX export with stable citation keys and automatic library sync. When adding a paper, prefer importing its .bib entry directly from the publisher page or arXiv rather than letting Zotero parse the URL — this avoids metadata errors caused by link-parsing failures.
Better Notes — adds a richer, Markdown-based note-taking experience inside Zotero, including templates and bi-directional linking between notes and papers.
Zotero Reading List — lets you tag papers as "to read", "reading", or "read" and track your progress through the library.

Getting started with Zotero

Save papers in one click — install the Zotero Connector browser extension. When you visit a paper page (Google Scholar, arXiv, ACM DL, etc.) the connector detects the metadata and saves the full record — including the PDF — directly to your library.
Annotate PDFs inside Zotero — highlight passages and add notes directly in the built-in PDF reader.
Sync your library across devices — By default, Zotero stores PDFs in its own cloud storage with very limited space. The ZotMoov plugin lets you automatically move attachments to a folder of your choice — such as your PoliMI OneDrive — keeping your PDFs backed up and accessible across devices. Follow the instructions in the ZotMoov GitHub repo to set it up.

Take notes — relentlessly. The single best habit you can build during your thesis period is taking notes as you go. Every paper you read, every experiment you run, every insight from a meeting — write it down. When the time comes to write the actual thesis, having a rich record of your thinking means you are assembling material rather than trying to reconstruct it from memory. The more faithfully you record your journey, the easier — and better — your thesis will be.