Questions deserve clear answers.
Understand how search works, what the ranking means, where the data comes from, and where human judgment still matters.
Start with search, inspect the explanation on each result, then save useful papers to a library.
Start searching →Browse by question or topic.
Showing all questions
No answer found
Try a shorter term, or explore the categories below.
Getting started
The essentials for finding and organizing literature.
What is PaperMetrix?
PaperMetrix is a research platform for discovering, explaining, and organizing scientific literature. It is also an experimental environment for evaluating retrieval and recommendation methods reproducibly.
Do I need an account to search?
No. Public search and paper details are available without an account. An account is needed for personal libraries, saved papers, research interests, and personalized features.
How should I write a good search query?
Describe the research problem with a few distinctive concepts. Include exact technical terms when they matter; you do not need to write a full question or Boolean expression.
Ranking and recommendations
What the order means and how to interpret it responsibly.
Does a higher rank mean a paper is scientifically better?
No. Rank estimates relevance to the current query. It is not a judgment of methodological quality, truth, novelty, or scientific impact. Researchers should still assess the paper itself.
How does PaperMetrix rank search results?
The live baseline combines lexical evidence such as BM25 with scholarly semantic similarity through Reciprocal Rank Fusion. The interface records retrieval components and timing so the ranking remains inspectable.
Why might two similar searches return different papers?
Small wording changes can alter exact-term and semantic matches. The corpus also grows as metadata is updated. PaperMetrix records request details for controlled experiments, but public search is a live discovery service.
Are citation counts used as proof of relevance?
No. Citations can provide structural or descriptive signals, but they reflect age, field, visibility, and social processes. They are never treated as an unbiased label of quality or relevance.
Data and access
Coverage, metadata sources, and links to full text.
Where does the scholarly metadata come from?
PaperMetrix reconciles complementary records from sources such as OpenAlex, Semantic Scholar, Crossref, arXiv, and Unpaywall. Provider identifiers and missing-data reasons are retained for provenance.
Does PaperMetrix cover every scientific paper?
No scholarly index is complete. Coverage depends on the registered corpus scope, dates, languages, document types, and available metadata. Missing papers should not be interpreted as irrelevant.
Can PaperMetrix give me the full paper?
When an open-access location is known, PaperMetrix links to it. Otherwise it links to the publisher or scholarly record. The platform does not host copyrighted articles or bypass access restrictions.
What should I do if metadata is incomplete or wrong?
Check the DOI, publisher page, and source provider before citing the record. Metadata can be incomplete, duplicated, or delayed; the canonical paper remains the authoritative object to inspect.
AI, privacy, and research use
How experimental models and interaction data are handled.
Does an LLM decide which papers I see?
Not in the public baseline. LLM query expansion is an isolated, switchable experimental channel with a fixed model and prompt, budget controls, and separate evaluation. Public LLM traffic remains disabled until evidence supports a controlled rollout.
What interaction data does PaperMetrix record?
For evaluation, the platform can associate impressions, clicks, saves, and explicit relevance feedback with a retrieval request. These signals support analysis but are not automatically treated as ground-truth relevance.
Can I use PaperMetrix for a systematic review?
Use it as a complementary discovery tool, not as the sole search source or screening decision-maker. A systematic review requires a registered search strategy, appropriate databases, reproducible screening, and documented inclusion decisions.
Can results from PaperMetrix be used in research publications?
Yes, when the corpus, query set, judgments, configuration, code revision, and statistical protocol are frozen and reported. Live interface observations alone are not a publication-ready evaluation.