Select text on any page, add a note and tags, and it is saved with its source. Save the page to your Library and it becomes a citable, verifiable source. Follow the journals of your field in the same place.
Free plan · No credit card · Chrome and Firefox
The dominant sequence transduction models are based on complex recurrent or convolutional neural networks that include an encoder and a decoder. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely.
Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train.
The extension works on the live page in Chrome and Firefox. Every highlight carries a required comment, tags, a colour and a visibility: private, public or shared with a space. Reopen the page and the highlights are back in place.
The dominant sequence transduction models are based on complex recurrent or convolutional neural networks that include an encoder and a decoder. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely.
Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train.
9f2c…e41aDOI10.48550/arXiv.1706.03762CiteAPA 7 · MLA 9 · Chicago · BibTeXSave a page from the extension, paste a link, or import BibTeX, RIS, CSV or bookmarks. Each source keeps a snapshot with a SHA-256 hash, its metadata and your quotes, can be monitored for changes, and is cited in APA 7, MLA 9 or Chicago.
Group highlights, web pages, documents and images into collections by project, topic or thesis. Three collections on the free plan, unlimited on Pro; a space has its own collections tab.
The field turned on one architecture; [[Transformer]] links the definition note.
Complexity per layer: $O(n^2 \cdot d)$ for self-attention.
- [ ] Compare with recurrent baselines
- [x] Add the citation (APA 7)
Write in a clean editor with headings, lists, task lists, blockquotes, code blocks, images, math with KaTeX and Mermaid diagrams. Embed your saved highlights and articles, link notes with wiki-links, and export Markdown.
Add any RSS or Atom feed, or pick one from the catalogue, and read new articles in one stream. From an article: highlight it, save it to a project or a space, bookmark it, mark it read. Feeds are free and unlimited.
A space is a group with owner, admin and member roles. Share highlights, feed articles, collections and sources into it and talk it through in real-time chat. Spaces are part of the Pro plan.
Sign in once; the extension picks up your session. Capture in the browser, write in the app.
Search highlights by tag from the header, and your Library in full text, including the archived page text.
Flexible tags across highlights, quick captures and sources.
Highlights are private unless you make them public or share them to a space. All traffic to our servers uses HTTPS.
Markdown from notes; BibTeX, CSV and JSON from the Library; evidence packs on Pro.
Share a preserved source as a public page with its citation; revoke it any time.