ArxivSwipe

About

ArxivSwipe — AI Paper Summarizer for arxiv & Hugging Face

Thousands of new ML and AI papers hit arxiv every week. Reading every abstract is impossible. ArxivSwipe drops a 3-sentence AI TL;DR right into your arxiv and Hugging Face Papers feed, so you triage hundreds of papers in minutes instead of hours — and only open the PDF for the ones actually worth your time.

WHAT IT DOES
• 3-sentence TL;DRs, generated by Claude, right on each paper card
• Optional deeper brief on demand when a paper deserves a closer look
• One-click BibTeX — copy a ready-to-paste citation (arxiv @misc format) for LaTeX, Overleaf or Zotero; press C or hit Cite on the card
• Code links — jump straight to a paper's implementation when one exists
• Explain a selection — highlight any term or sentence and get a plain-language explanation in place, no chat window
• Read TL;DRs in your language — pick a summary language and the card shows an AI translation, with the English original one click away
• Works on arxiv.org/list, arxiv.org/abs, and huggingface.co/papers
• Follow authors — never miss a new preprint from people you track
• Follow topics — keep an eye on categories like cs.AI, cs.LG, cs.CL, cs.CV, stat.ML
• Save for later — drop interesting papers into a reading queue
• Global summary cache — each paper is summarized once and served to everyone instantly
• Account-less cross-device sync — pair a second device with a short code
• Keyboard shortcuts: Alt+Shift+A to open the overlay, Alt+Shift+S to save, C to cite
• Light, Dark and System themes

WHO IT'S FOR
ML and AI researchers doing the daily arxiv triage, PhD students and postdocs keeping up with the literature, ML engineers tracking SOTA, authors maintaining a reading list, and anyone following Hugging Face Daily Papers who is drowning in the preprint stream.

HOW IT WORKS
Open arxiv or Hugging Face Papers and ArxivSwipe lays a card overlay over the feed with a TL;DR for each paper. Summaries are pulled from our backend's global Claude cache, so there's no wait for re-generation.

CITE & READING WORKFLOW
• Save — press Alt+Shift+S or the Save button to drop a paper into your queue
• Cite — press C or Cite for a ready-to-paste BibTeX entry (LaTeX / Overleaf / Zotero)
• Code — open the paper's implementation when a repo is linked
• Sync — enter a 6-digit pairing code on a second device; follows and saves sync, no account

WHY ARXIVSWIPE
• Free, no account, no paywall — no Pro tier and no cap on the inline TL;DRs, served from a shared cache. Heavy on-demand AI has a fair daily limit you can lift with your own API key — no account needed either way.
• Inline, right on arxiv — no copy-pasting links into yet another website
• Powered by Claude — clean 3-sentence TL;DRs and a deeper brief on demand
• Privacy-first — the only stored identifier is a locally generated device_id; there are no third-party trackers in the extension
• Scoped permissions — ArxivSwipe asks for access to arxiv.org and huggingface.co/papers ONLY. No <all_urls>. It never reads your history or traffic on other sites.
• MV3-native — built on the modern manifest

LOOKING FOR AN ALTERNATIVE?
If you've tried other paper-summary tools, ArxivSwipe is the free, no-account, privacy-first option that summarizes papers inline on arxiv and Hugging Face — with one-click BibTeX and code links — instead of sending you to another website.

USE CASES
summarize arxiv papers · keep up with machine learning research · daily arxiv triage · read research papers faster · AI paper digest · follow arxiv authors · track cs.AI / cs.LG / cs.CL · Hugging Face daily papers reader · export BibTeX from arxiv · cite an arxiv paper · find code for a paper · explain a research paper · build a reading list · literature review helper · NLP papers · computer vision papers · deep learning preprints

PRIVACY & PERMISSIONS
• storage — saves your follows, saved papers and settings, and syncs via a pairing code
• alarms — background timers (refreshing your follow feed, periodic tasks)
• scripting — renders the overlay cards on arxiv and Hugging Face Papers pages
• host access to arxiv.org and huggingface.co/papers only — to show the overlay there
We intentionally do NOT request <all_urls>. We intentionally do NOT read your browsing on other sites. The only identifier is a locally generated device_id, and there are no third-party trackers.