Proof gold annotator

Work from the paper's PDF. For each proof, list what it uses, in the order the proof uses it, naming results exactly as the paper prints them. Entries save as you type and are kept under your name.

loading…
Add a paper

For an arXiv id the parser fetches the source and fills in the suggested proof and statement lists within about a minute; a PDF-only paper keeps the lists you build with "+ add proof".

What counts, what doesn't, and the exact rules

List: results in this paper the proof relies on (by printed number); definitions in this paper whose content the proof unpacks; results or definitions from other works, written as the paper cites them ([6, Thm 2.1]); common-knowledge facts the paper uses without citing, by their usual name (Cauchy–Schwarz) or in one short sentence when they have none. One row per distinct item, at its first use.

Missing from the list? Our parser suggests the proofs; if the PDF has a proof the chips don't show, type its label in "a proof the list is missing" and add it. If a chip's label is wrong, open it, fix "Proof of", save, and clear the old chip.

Don't list: the statement being proved or its hypotheses; claims the proof itself establishes; routine algebra; background concepts the proof only mentions.

Items: the "uses" box offers every numbered statement our parser found plus everything already entered for this paper, by anyone. Pick the existing entry when it is the same thing, so one item has one name; type a new one only when it is genuinely new.

Order: the order the proof first uses each item. Evidence: copy six to twelve words from the proof where the item is used; leave empty for implicit items with no textual cue. Nothing from any tool: only the paper.

Export (JSON, the scoring format)