Reader
Open any arXiv paper as a PDF, move through it, and find text in it
The reader opens any arXiv paper as a PDF at /abs/<id>, with a toolbar for moving through it, finding text and downloading it, and a side panel for the abstract, your notes and, where it is configured, an assistant. Reading needs no account. Highlights and notes are on Highlights and notes, downloads on Export, and the optional question-and-answer tab on Reader assistant.
Open a paper
Paste an identifier or link into the field on the front page and press Read, or use the magnifier in the header of any other page. The same forms are accepted everywhere a paper is asked for, including the Library’s save box.
| Accepted form | Example |
|---|---|
| A bare identifier | 1706.03762 |
| A versioned identifier | 1706.03762v7 |
| An old-style identifier | hep-th/9901001 |
| With the arXiv: prefix or a .pdf suffix | arXiv:1706.03762 |
| An abstract, PDF or HTML link on arxiv.org | https://arxiv.org/pdf/1706.03762 |
| The same paths on export.arxiv.org or alphaxiv.org | https://alphaxiv.org/abs/1706.03762 |
Anything else is refused with Enter an arXiv ID or an arxiv.org paper link. An address that looks like an identifier but is not one shows the Page unavailable page. The page says Opening the paper… while it prepares, then Loading PDF… while the file streams from arXiv through Substrate. If the PDF cannot be fetched, the reading pane says Couldn’t load the PDF and offers Try again. HTML links only resolve to the PDF; there is no HTML reading mode.
The header
The header shows the paper’s title, its first six authors and a Paper badge. arXiv opens the abstract page on arxiv.org in a new tab. The magnifier opens another paper; Rooms, Library and, except on narrow screens, Docs lead elsewhere in the app. Show Panel and Hide Panel toggle the side panel, which starts hidden on narrow screens.
Signed in, two more buttons appear before the arXiv link. The bookmark, Save to Library, saves the paper to your shelf and becomes Saved to Library. The folder, Save to a collection, opens a Collections list with a check mark beside each Collection that already holds the paper; clicking one files or unfiles it, and filing also saves. A New collection field at the bottom creates a Collection and files the paper in one step. See The Library.
The toolbar
| Control | What it does |
|---|---|
| Page number | Type a page and press Enter to jump there. The count after the slash is the number of pages. |
| Zoom out, Zoom in | Change the zoom by 10% per press, between 50% and 300%. |
| Zoom percentage | Shows the current zoom. Clicking it resets to 100%. |
| Fit width | Also resets the zoom to 100%. It does not compute a different width. |
| Highlight count | Appears once the paper has at least one highlight. |
| Find in paper | Opens the Find bar. Also ⌘F or Ctrl+F. |
| Download Paper | A menu with Without annotations and With annotations, described on the Export page. |
Find in the paper
Press ⌘F (Ctrl+F on Windows and Linux) or the toolbar button. Find searches the text of the PDF as the reader extracts it. Typing shows a running current/total counter; ↵ moves to the next match and ⇧↵ to the previous one. Three switches sit beside the counter: Highlight All (on by default), Match Case and Whole Words. Esc closes the bar.
- A query of fewer than two characters matches nothing.
- At most 500 matches are collected; further ones are not counted.
The side panel
| Tab | What it holds |
|---|---|
| Overview | Title, authors, the abstract, a Publications section (a link Publish a claim or finding, then the cited claims and findings that reference this paper) and a Related research section. A Write a note button switches to My notes. |
| My notes | Your private note for the whole paper, the save status line, the Export notes & highlights button, and the list of your highlights. |
| Assistant | Shown only when the server has an assistant key configured and you are signed in. |
The Publications list is public and comes from the paper’s associations in Rooms; when there are none it says No publications here yet. If you opened a versioned identifier such as 1706.03762v7, a collapsed Citations with an unspecified revision section lists publications whose authors did not declare which revision they cited. What a publication is, and how to make one from this paper, is on Findings and cited claims.
Below it, Related research Research about a paper: Live opens two lists: the hypotheses and the experiments that named a cited claim of this paper as a premise, each showing the path by which it reaches the paper. A versioned identifier adds a third list for research that reached the paper through a citation with no declared revision. Only an explicit premise puts anything here; see Hypotheses and experiments.
Where the paper details come from
The title, authors and abstract Paper metadata: Live come from arXiv’s OAI-PMH interface, which describes the latest version of a paper; a versioned link still opens that version’s PDF. A good answer is kept for a day, so a paper anyone has opened recently does not ask arXiv again. The page does not wait for the answer: the header first shows arXiv:<id> and the Overview tab Loading paper details…, and both fill in when arXiv answers, usually before the PDF does.
If arXiv refuses or does not answer within the limit, one retry follows after a short pause. If arXiv answers that it does not describe the paper, which happens for a few hours after a new paper is announced, Substrate asks arXiv again at once without the kept answer, so a stale answer never stands in for a paper that exists. After that, a paper that anyone has saved to a Library is named from Substrate’s own record, without an abstract, and the Overview tab says The abstract is unavailable from arXiv; the title and authors come from Substrate’s own record of this paper. Otherwise the header keeps arXiv:<id> and the tab says Metadata is unavailable from arXiv. The PDF may still load; reload to retry. The PDF is fetched separately, so reading usually still works. The limits are on Limits.
Second metadata source: IdeaSecond metadata source
When arXiv’s own interfaces fail, the citation tags of the abstract page or DataCite’s record of the arXiv DOI would name the paper, so the reader depends on no single arXiv service.
Remembered metadata failures: IdeaRemembered metadata failures
A refusal or an outage would be remembered for a minute, so that reloads during it show the same answer without asking arXiv again.
Stored paper metadata: IdeaStored paper metadata
Substrate would keep every paper’s abstract and categories in its own database, next to the title and authors it already stores for saved papers, so metadata outlives the cache and each deployment.
Signed out, everything you do in the reader stays in this browser. Signing in keeps notes and highlights with your account instead; see Accounts and privacy.