Every paper, woven together
Search 3.1 million papers by meaning, trace the researchers behind them, and ask questions across the ones you keep.
The latest papers
Environment Alignment and Redundant Record Formation in Imperfect-CNOT Quantum Darwinism
Coding Agents with an Obstacle-Aware Harness for Safe Robot Manipulation
Embedding Models Measure in Peculiar Ways
Workspace Models: Lightweight Robotic Memory via Saliency-Driven Supervision
Can 4D Foundation Models Remember?
SplashSplat: Reconstructing Splashing Liquids from Real-World Multi-View Videos
6,745 more from the last 7 days
Sign in to see the week filtered to the fields you actually read, and to keep the ones worth returning to.
What you can do here
Search by meaning
Describe the idea, not the keywords. Results rank on semantic similarity, so papers that never use your words still surface.
Try a searchTrace the people
Author records built from co-authorship: what someone published, with whom, and where the work moved next.
Browse researchersAsk across papers
Put a handful of papers in context and question them together. Every answer cites the section it came from.
Open the assistantThe pulse
Last 30 daysBefore you start
wovepaper is a search and reading surface over the arXiv corpus. It indexes paper metadata, embeds every title and abstract for meaning-based search, resolves authors into disambiguated researcher profiles and links those to institutions, then lets a signed-in reader question the papers they keep.
What is wovepaper?
A search and reading surface over the arXiv corpus. It indexes paper metadata, embeds every title and abstract so search can work by meaning rather than keywords, resolves author names into disambiguated researcher profiles, links those to institutions, and lets a signed-in reader ask questions across the papers they keep.
Do I need an account?
No, not for the corpus. Search, paper pages, researcher profiles and institution pages are public, unmetered and need no sign-in. An account only unlocks the parts that are personal or that cost money to run: a saved library, followed topics and researchers, and the assistant that reads papers and answers questions. Signing in is by emailed link or short code, with no password.
How is this different from searching arXiv directly?
arXiv's own search matches the words you type against the words in a paper. wovepaper compares the meaning of your description against the meaning of every title and abstract, so a paper that solves your problem in unfamiliar vocabulary still surfaces. It also builds the layer arXiv leaves out: which person wrote what, which institution they were at, and what a body of work adds up to.
Where does the data come from?
Paper metadata and abstracts come from arXiv's official OAI-PMH feed, with links back to arXiv for the canonical abstract and PDF. Researcher records and citation counts come from Semantic Scholar and OpenAlex, which is also the source of institution identifiers. Summaries and topic tags are generated from the paper's own text by a language model.
Can I trust the AI summaries?
Treat them as a reading aid, not as the record. They are generated from the paper's own text and they can be wrong or flatten a nuance, which is why every paper page shows the abstract, links the arXiv original and the PDF, and cites the section an assistant answer came from. If a summary misrepresents a paper, reporting it gets it regenerated.
Is wovepaper open to AI agents and crawlers?
Yes, deliberately. Every public page is server-rendered, robots.txt allows the whole corpus, a sitemap index lists every indexable URL, and any page will serve clean markdown instead of HTML to a client that asks for it with an Accept header. The public JSON API needs no key for corpus reads, and llms.txt describes when to use the site and which endpoints to call.
Chat with the papers
Sign in to get the assistant, a feed filtered to your fields, and a library that remembers what you read.
Get started