A summary tells you what a paper says. Around it, you also want to know which papers it builds on, which ones came after, which ones are close without citing it, and whether the version you read is the last one. Every paper page on SummarizePaper gathers this. The summaries guide covers the summary itself; this one covers what is around it.

Our examples: the neighbourhood of Bergez-Casalou & Kral (2024), on planetary gaps in the gas of debris discs, and the versions of Morf & Helled (2026), on the interiors of the sub-Neptunes K2-18 b and TOI-270 d, three versions on arXiv.

1. The tools of a paper page

Under the summary, Tools for this paper gathers what you can do with it: explain it at another level or in another language, save it to a collection with a reading status, export it (BibTeX, RIS, Obsidian, Notion), and the Premium tools: send it to Zotero, follow its citations, find who works on its topic, prepare a journal club, see what changed between its versions, or prepare its press kit. Each tool links its guide.

Tools for this paper: explain it differently, save to a collection, reading status, export, and the advanced tools Zotero, citation alerts, who works on it, journal club, versions and press kit, with links to their guides.
Everything you can do from a paper page, with the guide of each tool.

2. The papers around it

Further down, Papers around this one lists up to three groups, each paper with its closeness (very close, close, related), its authors, year and journal, its citation count, and whether it is already summarized here:

Closeness compares titles and abstracts, so a close paper that uses other words still comes up. A paper both close and cited stays among the closest, with both facts: for our paper, the close work it cites is all there, so only two groups show.

Papers around this one: the closest papers, such as The vertical structure of debris disks and the impact of gas (Olofsson et al. 2022), each marked Very close, with its authors, year, journal and whether it is summarized.
The closest papers of a paper on the gas of debris discs.

Tick several papers to chat with them together, or put them in a collection. Not relevant takes a paper out of your list.

3. The map, by date and closeness

A map of these papers by date and closeness, a link under the list, places them by date (left is older) and closeness (higher is closer); lines join papers close to each other, and groups of lines are sub-topics. The dashed line is the paper itself: on the left, the work it builds on; on the right, in green, the papers that cite it. Select a point to see why that paper is there.

The map of the papers around Observing planetary gaps in the gas of debris disks: points by year and closeness, from Kral 2017 and Pericaud 2016 to Marel 2026, with the papers that cite it in green after 2025.
The papers around one, by year and closeness; the papers that cite it in green.

4. What changed between versions (Premium)

An arXiv paper can have several versions: the one accepted by a journal often differs from the first. Versions → What changed between versions compares two of them. Choose the older and the newer version, then compare; each comparison is kept, for each pair of versions and language.

What changed for Composition gradients in sub-Neptunes: K2-18 b and TOI-270 d as case studies, with the older version v1, the newer version v3, and a saved comparison v1 to v3.
Two versions to compare, and the comparisons already made.

The result gives a verdict, then each change with its kind (results, conclusions, method, data, an error fixed, figures, references, writing) and its importance. For our paper, the authors' comment only says "v2: accepted after moderate revisions; v3: language compliance update", but the comparison finds that the results changed: version 1 announced that composition gradients could raise the maximum hydrogen-helium fraction of K2-18 b up to fivefold; version 3 gives about threefold at 300 K and tenfold at 600 K, and adds TOI-270 d.

From v1 to v3: limited comparison, abstracts and comments only; the verdict that quantitative results changed; then two changes of results, one of conclusions and one of writing, each explained.
What changed from v1 to v3: two results, one conclusion, and the writing.

The page says what was compared. Here arXiv did not have the HTML of every version, so only the abstracts and comments were read: changes elsewhere in the paper cannot be seen, and the verdict says so. When the full text can be read, only the paragraphs that differ are given to the model.

When you follow a paper's citations, or watch a collection, you also get an email when it gets a new version, with what changed (citation alerts guide).

Limits

FAQ

How are the close papers chosen?
By the closeness of their title and abstract to those of the paper, on one scale: the closest papers of our whole index and those Semantic Scholar recommends, the papers it cites and the papers that cite it. Each one says how close it is, whether it is published and how often it is cited.
Can I remove a paper that has nothing to do with it?
Yes: Not relevant takes it out of your list, and out of everyone's once several readers agree.
Does the version comparison read the full text?
When the license of the paper allows it and arXiv has the HTML of both versions, the paragraphs that differ are compared. Otherwise only the titles, abstracts and comments of the versions are, and the page says so.

Sources