Attribution & licence

The archived Papers with Code data

The paper, code-link, benchmark, dataset and method records on this site were created by the Papers with Code project (paperswithcode.com, operated by Meta AI until the site was shut down in July 2025) and preserved after the shutdown by the pwc-archive community mirror on Hugging Face. The original site's footer stated: "Papers With Code is a free resource with all data licensed under CC-BY-SA."

The material is licensed under the Creative Commons Attribution-ShareAlike 4.0 International licence (CC BY-SA 4.0): https://creativecommons.org/licenses/by-sa/4.0/legalcode. Snapshot date: 2025-07-28 (the mirror's stated last public retrieval). The tables used, each with its licence tag on the mirror:

• pwc-archive/papers-with-abstracts — papers, abstracts, authors, dates, per-paper tasks and methods
• pwc-archive/links-between-paper-and-code — paper ↔ repository links with the official / mentioned flags
• pwc-archive/evaluation-tables — tasks, benchmarks and leaderboard rows
• pwc-archive/datasets — the dataset catalogue
• pwc-archive/methods — the method catalogue

The material has been modified

Syntology changed the archived material as follows. The exact counts for the current build are on what is not shown.

• Reshaped the five tables into one catalogue; joined leaderboard rows to papers by URL, arXiv id and title; joined benchmark dataset names to the dataset catalogue by name, full name and variant.
• Removed method rows identified as spam (rows hung off five dummy paper entries, and names matching a customer-service pattern), and merged the method records embedded in the paper catalogue with the method table.
• Kept one leaderboard per benchmark URL where the archive reached the same table from several task nodes.
• Rewrote Markdown links that pointed at paperswithcode.com to this site's routes, and dropped images hosted on the original site's media server, which no longer resolves.
• Rendered no page for papers that have no code link, no benchmark row, no introduced dataset or method, and no code in Syntology's graph.
• Added, clearly marked Syntology, records that are not part of the archived material (below).

What is Syntology's own

Repository links marked "found in paper text by Syntology", the harvested code samples, every per-sample execution label, behaviour fingerprint and licence state, every paper marked "added by Syntology", and, in a build that shows them, results listed under "Since the archive: results placed by Syntology" on a benchmark page or "Placed on leaderboards by Syntology" on a paper page (numbers Syntology extracted from the paper's own tables; what is not shown states whether this build shows any) come from Syntology's graph and extraction runs and are not derived from the archived dataset. Syntology offers these records on this site under the same CC BY-SA 4.0 licence so that the whole site can be shared under one licence.

arXiv metadata

Two kinds of page show text from arXiv's own metadata, which arXiv releases under CC0 1.0 (https://info.arxiv.org/help/license):

• Papers newer than the archive snapshot: the title, abstract, author list and date on the paper page and in the home feed.
• Archive papers marked “links table only”, whose catalogue row the archive never published: the authors, date and abstract on the paper page, when it shows them, read from the Kaggle arXiv metadata snapshot where its title matched the archive's. The title stays the archive's.

Each such page says so once. The papers themselves remain under their authors' licences and are linked, not reproduced.

Abstracts

An abstract is its authors' text, and the CC BY-SA 4.0 licence on the archived data does not relicense it. On an archive paper's page, the abstract is the one the archive recorded, unless the page is marked “links table only” (below); the page's “Paper” button links to where the archive says the paper is published.

• Where that link goes to arxiv.org, the abstract is the paper's arXiv abstract as the archive recorded it; arXiv releases its abstracts as part of its metadata under CC0 1.0 (https://info.arxiv.org/help/license).
• Where it goes anywhere else, such as a publisher's or a proceedings site, the site cannot show that the abstract came from arXiv, so it remains under that publisher's or its authors' own terms, not under the CC BY-SA 4.0 label that covers the rest of the archived data, and it is shown here as the archive recorded it, with that link to its source.
• An archive paper marked “links table only” has no abstract in the archive; any authors, date and abstract it shows are from arXiv's metadata (CC0 1.0), as its page states (above).
• Papers newer than the snapshot show an abstract only from arXiv's metadata (above).

To have an abstract removed, write to legal@syntology.ai.

ShareAlike

This site shares a substantial portion of the archived database. The compilation served here is therefore offered under CC BY-SA 4.0, with no additional terms or technical measures restricting the rights that licence grants: no account, no token and no payment stands between you and any archive row, and the terms impose nothing beyond the licence. Third-party code linked on paper pages remains under its own licence; a sample's licence state is stated beside it. Abstracts the site cannot show came from arXiv remain under their publishers' or authors' terms (above).

No affiliation

Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror. Nothing here implies that the use of the archived material is connected with, sponsored or endorsed by, or granted official status by, its creators.

Syntology's own attribution page, covering the same dataset as used inside the Syntology graph: syntology.ai/attribution.