Reference

Glossary of whitepaper terms

Whitepapers borrow words from research, engineering, law, and marketing, and they do not always use them in the same way. These entries give the sense each word carries when it appears in a serious paper.

Claims and evidence

Claim

A statement the paper presents as true and asks you to accept, wholly or in part, because of its evidence. Claims appear in the abstract, the introduction, and the conclusion. Write each one in your own sentence before you judge it.

Evidence

The data, proof, measurement, or case offered in support of a claim. Evidence is useful only when you can say which claim it supports and how.

Provenance

Where a figure, dataset, or quotation came from, and how it reached the paper. Provenance decides whether a number traces back to a measurement or only to another author's summary of it.

Citation chain

The path from a claim through a series of references to the original source. Long chains tend to lose context at each link, so check the last link before you trust the first.

Baseline

The comparison point for a result, such as an earlier system, a simpler method, or no intervention at all. A result without a fair baseline shows a change, not an improvement.

Method and reproduction

Methodology

The procedure used to produce the evidence, including how data was gathered, measured, filtered, and analyzed. Describe it in your own words before you evaluate it.

Assumption

A condition the argument needs in order to hold, which the paper may not state directly. Examples include honest participants, stable prices, or a test population that resembles the people the method will serve.

Benchmark

A standard test, dataset, or task used to compare methods. Check whether the benchmark resembles the use case the paper says it serves.

Sensitivity analysis

A test of how much a result moves when an input or assumption is varied. A conclusion that flips under small changes is fragile, even when it is correct on the page.

Reproducibility

Whether a careful second reader can follow the described method and reach the same or a consistent result. It depends on the text as much as on the experiment.

Scope and disclosure

Scope

The range of cases a paper claims to cover. Conclusions should not reach past it, and a careful reader notices when they do.

Threat model

In security and systems papers, the adversaries, capabilities, and failures a design is meant to resist. Anything outside the threat model is outside the paper's claims.

Limitation

A boundary the authors admit. Limitations are often the most candid part of a paper. Check whether the conclusions respect them.

Conflict of interest

A relationship, such as employment, ownership, or funding, that could bias the conclusions. Disclosures often sit in a footnote, an acknowledgments section, or the final page.